At a glance
- Latest
- X · 29 Aug 2026 Baker notes that a 50% price cut in GPT 5.6 drove 14x volume increase, calling it wild.
Themes 8 themes
last 90 days
-
Baker argues disaggregated GPU architectures with SRAM will outperform custom ASICs and sees open-source models as net positive for infrastructure.
postBaker says future is multi-model with specialized open-source models on customer data working with frontier models behind custom router
“@IbrahimAjami Yes. The acceleration in OpenAI/Anthropic as open-source takes share is so interesting. The future is multi-model with specialized open-source models that were post-trained on customer data working together with frontier models behind a customer specific router. Firework Nexus”
postBaker says Atreides internal AI spend will be 100x higher in August 2026 versus March 2026 and still doubling monthly
“Suspect we will start to hear about a “Pareto optimal” balance of computationally efficient humans, cheaper open-source tokens and frontier tokens. Our internal AI spend @Atreidesmgmt will be roughly 100x higher in August 2026 vs. March 2026. Still roughly doubling every month.”
post“More data than open-source AI is taking share from OpenAI and Anthropic. Open source has gone from 28% token share to 62% token share @vercel over the last 2 months. Chart from @rauchg Super impressive given that the sum of OpenAI and Anthropic accelerated in July. So net https://t.co/wjDtlcl6ic”
post“@Shaughnessy119 @TMTBreakout 2 of the 3 parts of your equation are correct (first and last, not the middle). An open-source token costs just as much compute and energy to produce as a frontier token for a model of comparable size. And they are often less token efficient.”
postBaker would not take under on 250 billion for Anthropic unless they stumble or fail to secure compute
“@altcap Love the expectations management! Let me say this: I would not take the “under” on $250 billion *unless* they stumble or fail to secure enough compute. Maybe you can discuss with @DarioAmodei at @TomorrowXSummit before I interview you about public policy!”
postBaker estimates his personal AI usage is up 100x and calls Grok Bot another Claude Code moment
“I think @bot is another “Claude Code” moment for AI. I would estimate my personal AI usage is up something like 100x. And for everyone who reached out about how to build a “podcast summarizer” it took me about 15 seconds in Grok Bot and is better than what I had before. https://t.co/AaoeHbEqtl”
postBaker says Grok 4.6 matches Fable 5 Max at 85% discount, 80% cheaper input and 88% cheaper output, Pareto dominant
“Grok 4.6 is roughly the same performance as Fable 5 Max at an 85% discount. 80% cheaper for input tokens and 88% cheaper for output tokens. Pareto dominant. Grok 4.7 will be significantly better as is a much larger model with the Cursor and SpaceX data included in pretraining. https://t.co/4e8CXbzs2D”
post“Market is overreacting to hyperscale credit spreads widening from my perspective. TL;DR Spot pricing for renting GPU compute materially above contracted rates implies hyperscalers are underearning while operating cash flow acceleration is an underestimated source of funds for AI https://t.co/D9FSHXQ8F8”
40 more posts
postBaker notes Jalapeño participated in third-party benchmarks before TPUs did, suggesting competitive confidence
“Also pretty funny that Jalapeño participated in a third party benchmark *before* TPUs! Tells you a lot!”
postBaker says real comparison will be Jalapeño versus Rubin Ultra and Feynmann given current shipping timelines
“Should note the real comparison will be with Rubin Ultra given Rubin shipping today so kinda have to wait. And Feynmann coming fast. But still impressive debut from Jalapeño.”
postBaker observes different chips hit different points on interactivity versus throughput tradeoff for model sizes
“@ChrisMurphyHQ Seem to hit different points of interactivity vs. throughput tradeoff for different size models. Basically everyone plays to their traditional strengths is early read. Really good AFFN disaggregation could change this.”
postBaker credits Jalapeño as first good ASIC outside TPU and Trainium but says disaggregated GPU plus SRAM will outperform
“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
postBaker says frontier labs claim their tokens use less compute than open-source tokens and have more intelligence density
“@JoshPurtell @arakharazian Frontier labs would say their tokens use less compute than comparable open-source tokens but that would break people’s brains. And for sure they have more intelligence density per token which compounds their efficiency advantage.”
postBaker argues open-source gaining share is net positive for everyone except model companies
“@arakharazian Markets do irrational stuff all the time! But open-source gaining share is net positive for everyone except for the model companies.”
postBaker argues cheaper open-source tokens are net positive for AI infrastructure because they cost same compute as frontier tokens
“@arakharazian Respectfully this is not correct from my perspective. Cheaper open-source tokens are net positive for the AI infrastructure names, although maybe not for the model companies. An open-source token broadly costs the same amount of compute (watts and wafers) as a frontier token to”
postBaker calculates Ultrafast timeline from December deal through wafer and server production to August debut
“@DratchCap @anni_sen @stefanveliki137 @thsottiaux I am also curious about this. Along with why Ultrafast mode is capacity constrained. Deal struck in late December, 4.5 months to make a wafer, 6 weeks to make a server, a month to energize a server. So logical that Ultrafast debuted in August and would think it would be less”
post“I literally have no idea why anyone is concerned about the FT data on Fable 5. Just shows customer preference for ZDR. Actually surprised so many are still using it after Opus 5 became available. https://t.co/u5Rzi13JmV https://t.co/GCqX5bMgYg”
postBaker says more interesting question is where teacher models stack up versus frontier and where they will be in months
“@firesidealpha This was a good podcast. The more interesting question is where those teacher models stack up vs. the frontier. And where the teacher models will be in a few months. Those ASICs sure were fun to develop!”
postBaker asks whether OpenAI or production constraints are limiting Ultrafast capacity
“@anni_sen @stefanveliki137 @thsottiaux You think OpenAI is constraining the capacity or production volume constraints elsewhere are constraining the capacity? Genuinely curious.”
post“@StevanBoljevic @insane_analyst @IanCutress Yes also disaggregated inference is the future IMO.”
postBaker argues Ultrafast is flagship model from second largest frontier lab and OpenAI wants it as default mode
“@anni_sen @stefanveliki137 Hmm. This looks like the flagship big model from the second largest frontier lab. And IIRC @thsottiaux from OpenAI said they wanted this currently capacity limited Ultrafast mode to be the default mode 6 days ago. I’m not sure I would describe “default mode” for the flagship https://t.co/OQqwwW8CD9”
postBaker says GPU price increase means anyone spending on Blackwell and Rubin before January 2027 gains competitive advantage
“The most obvious implication of GPU price increase is that anyone who is spending a lot on Blackwell and Rubin before January 31, 2027 is in a good relative position vs. competitors. Super curious to know if GPU prices are going up more than rack prices.”
postBaker views Grok Bot as a great router and harness
“@iabdul_wasey Grok Bot feels like a great router and harness to me.”
postBaker has high conviction that large enterprises will use frontier harnesses for Pareto dominance while keeping data safe
“High conviction hypothesis especially for large enterprises. This is the best path to Pareto dominance while keeping data safe. SMBs and individuals probs just use the frontier harnesses. https://t.co/4petLDY4kt”
postBaker calls AFFN disaggregation the holy grail but notes it is a hard networking problem
“@AccBalanced @Atreidesmgmt AFFN disaggregation is the holy grail just such a hard networking problem.”
postBaker explains open-source lowers compute spend by ratio of frontier to open-source margins because frontier marks up compute more
“@itsvlady Well it lowers compute spend by the ratio of the margins of the frontier tokens to the margins of the open-source tokens - frontier tokens mark up underlying compute dramatically more than open-source.”
postBaker says ratio of frontier inference margins to open-source margins will determine the outcome
“@JosephJacks_ The ratio of frontier inference margins to average open source inference margin will determine the result.”
postBaker thinks Anthropic likely building in acceleration for roadshow and year-end but less true for OpenAI
“@jweil130 @vercel @rauchg Likely true of Anthropic IMO - likely building in an acceleration for roadshow and end of year. Probably less true for OpenAI.”
postBaker says big models are expensive and Opus 5 is a Mythos or Fable distillation doing well
“@F1NewsWire @TMTBreakout Big models are expensive. I don’t think super important one way or another. Opus 5 is a Mythos/Fable distillation and doing well.”
postBaker says Citadel probably added AI exposure elsewhere after reducing position
“@hitraderjoe @CNBC Yes. And probs means they added AI exposure elsewhere. He is super strategic.”
postBaker agrees and thinks Poolside news is important indicator for future of American open-source
“@DavidSacks @sarthakgh Yes. Although I do think the Poolside news is an important indicator for the future of American open-source.”
postBaker says sum of OpenAI and Anthropic accelerated QTD with sharper August acceleration while Grok and open-source grow faster
“Yes. And the sum has accelerated slightly QTD with a sharper acceleration in August vs. July. And note that Grok Bot, Grok and Open Source are all likely growing faster than both of the two leading frontier labs. But at some point going to have to move past second derivatives https://t.co/L4r1aExjrI”
postBaker explains Anthropic and OpenAI have more leverage over infrastructure providers than open-source does over cloud providers
“@Shaughnessy119 Y is higher than X - Anthropic and Open AI have more leverage over the infrastructure/cloud providers. For a given token over time all else equal, the cloud and infrastructure players charge more for open source tokens vs. what Anthropic and OpenAI pay per token b/c open source”
postBaker says Stripe is growing net revenue and FCF almost 2x faster than Adyen with declining share count
“Empirically impressive metrics from Stripe in the Axios report about their acquisition of Openrouter. Growing net revenue and FCF almost 2x faster than Adyen YTD with the share count declining over the last 3 years. I think Stripe will be a good steward of Openrouter and it is https://t.co/mnxU0XqP0N”
postBaker says Fireworks and Baseten are two largest inference clouds and first time seeing SpaceXAI on fastest growing list
“Fireworks and Baseten are the two largest inference clouds and I think of them as *an* index for open-source AI adoption. Also first time seeing SpaceXAI on the list of “fastest growing” IIRC. SambaNova! Will wonders never cease. https://t.co/HlmFsKRtVk”
postBaker suggests Anthropic may have shifted from gross to net ARR reporting making 65 billion more conservative than 47 billion
“Good chart from @TMTBreakout July accelerated both MoM and YoY for the sum of OpenAI and Anthropic. Some chance Anthropic shifted from gross to net reporting for ARR. Meaning the 65b is a more conservative metric than the $47. Smart thing to do before an IPO. Openrouter data https://t.co/FOwqXizXIL”
postBaker thinks at least two other labs believe they might be creating God
“@firesidealpha Funny and true. I do think at least two other labs think they might be creating God.”
post“From yours truly on Friday’s @theallinpod “On the margin, Anthropic is losing share to OpenAI, to open source, and to Grok, and they're still growing so fast.” If the $65b is correct, then someone leaked it to help Anthropic on their forthcoming IPO roadshow. Basically, they https://t.co/oIYSS1PZhT”
postBaker acknowledges having more sophisticated use cases than podcast summarizer but cannot post them
“@goldilock3din Laughing I do like my “podcast summarizer” There are much more sophisticated uses cases I promise but can’t post those here!!!”
postBaker shares GrokBot origin story and notes they lack sharing feature and iPad app but both coming soon
“The GrokBot origin story as told by @bot Sadly they lack both a sharing feature and an iPad app but I’ve been told they are both coming soon: https://t.co/4YaISohWW6”
postBaker says future is multiple models working together, lower model margins mean more AI infrastructure demand
“Important. The future is almost certain to be multiple models working together. Lower margins at the model layer whether from open-source or vertically integrated players = more AI infrastructure demand. https://t.co/YBWlCNnrDD”
postBaker announces interviewing Lin Qiao of Fireworks AI at Tomorrow X Summit November 17-18, praises specialized intelligence vision
“Super excited that I will be interviewing @lqiao CEO and founder of @FireworksAI_HQ at the @tomorrowxsummit in Austin, TX November 17-18. Lin’s vision of specialized intelligence is compelling and she is making it a reality. Fireworks is at the forefront of helping enterprises https://t.co/MK7CFgAsAU”
postBaker announces Antonio Gracias interviewing Jon Gray of Blackstone at Tomorrow X Summit November 17-18, notes Blackstone leaning into AI
“Looking forward to listening to @AntonioGracias interview Jon Gray, President and COO of @blackstone at the @tomorrowxsummit November 17-18 in Austin, TX. Blackstone is leaning into AI more than most of its peers and they are integral part of financing it. So many topics: their https://t.co/6xK1XtNKl1”
postBaker says Terafab will create over 3000 high quality jobs in heart of Texas, AI good for America
“AI is good for America, episode 1001. Terafab will create over 3000 high quality jobs right in the heart of Texas. https://t.co/FA2KHqMd11”
post“Really impressive from @cursor_ai The future is trending hybrid. Next iteration will likely be allowing customers to use their own proprietary data to post-train an open-source model. Excited for Grok 4.6. https://t.co/ZTE2VAskTG”
postBaker says Crusoe CEO will have interesting perspective being vertically integrated from dirt through electrons through tokens
“@benitoz @ChaseLochmiller @CrusoeAI @TomorrowXSummit He will have an interesting perspective on the buildout. Vertically integrated from dirt through electrons through tokens.”
postBaker announces Chase Lochmiller of Crusoe as second Tomorrow X Summit speaker, jokes about first time seeing him in suit
“Happy to announce @ChaseLochmiller CEO and Founder of @crusoeai as the second confirmed speaker at the @TomorrowXSummit This photo is my first time ever seeing Chase in a suit! https://t.co/mL4kKOTfwf”
postBaker acknowledges real yields rising and spreads widening but argues less credit needed given operating cash flow acceleration
“@dampedspring Of course. Real yields going up and spreads widening is undeniable. I am saying we may need less credit than expected given a likely continued acceleration in operating cash flow.”
-
Baker says multiple sources believe Anthropic wants to remain the only private AI company, consistent with Dario Amodei's public messaging.
postBaker advises a founder not to publicly complain about VC criticism and to focus on execution, referencing Starlink
“@PhilipJohnston Literally don’t even post about it. Move on. You are getting Starlink lasers for a reason.”
postBaker says Dario's public commentary makes it believable he said Anthropic might be the only private company
“@sashadem @theallinpod I believe both you and Sholto. This issue is that Dario’s own public commentary makes it believable to many serious people in Silicon Valley that Dario might have said something similar to “Anthropic might end up being the only private company.” I would love nothing more than to https://t.co/NF7MaO8wl1”
postBaker says multiple serious Silicon Valley people believe Anthropic wants to be only private company, consistent with Dario's messaging
“@_sholtodouglas Sholto, thank you for setting the record straight. Larger issue is that multiple very serious people in Silicon Valley have heard some variation of this and believe it to be true. And the reason it is believable to so many is that it is consistent with Dario’s public messaging”
postBaker notes FT reports three trillion dollar Anthropic valuation, says public markets struggle to value this high growth
“$3 trillion for Anthropic in the FT. I feel like I’ve heard that somewhere else. Although public markets generally struggle to accurately value growth when it is this high. https://t.co/2R80q18GUH”
post“@jukan05 4 gigawatts this year and another 4 next year per this report. Spot price for compute has gone up since they signed their Anthropic and Google deals.”
postBaker asks where Anthropic's and OpenAI's combined 150b ARR came from, disputes that leveraging up is only funding path
“@dampedspring Where has Anthropic’s and OpenAI’s combined 150b of ARR come from? I don’t think the idea that leveraging up is the only way to fund this is at all correct.”
postBaker argues 'I don't know' are the three most important words in investing, not 'margin of safety'
“@travisb1224631 I tell myself this every day. “I don’t know” are the three most important words in investing not “margin of safety.””
-
Baker defends AI infrastructure buildout and argues hyperscalers are underearning given spot compute prices now exceed contracted rates significantly.
postBaker agrees that Jensen is competing with OpenAI and predicts American open-weight models will approach the frontier.
“@dnapway Directionally accurate imo. I think we are going to see American open-weight models approach the frontier one way or another.”
postBaker hypothesizes tomorrow's Ultrafast will be GPUs or Jalapeño with CS-4 or CS-5
“@dylan522p @maxkan And they also said this followed by “Cerebras FTW.” I would be surprised if their head of Codex was unaware of their roadmap. Reasonable to hypothesize tomorrow’s Ultrafast will be either GPUs or Jalapeño working with CS-4 or CS-5. We shall see! https://t.co/QM0vs809x7”
postBaker argues game theory suggests memory LTAs are durable and Nvidia may have significant cloud revenue shares
“Game theory suggests memory LTAs are likely to be durable. And yes, Nvidia may have a significant cloud business via revenue shares. https://t.co/GFN1kDQmSk”
postBaker predicts vertically focused AI native companies will accelerate due to routers, open-source models and specialized post-training
“We are going to see a lot of vertically focused AI native companies accelerate. Routers, open-source models and specialized post-training enabled by companies like @FireworksAI_HQ have all made dramatic advances and the combination of the three is driving accelerating growth. https://t.co/f8qJdKSciI”
post“@jukan05 4 gigawatts this year and another 4 next year per this report. Spot price for compute has gone up since they signed their Anthropic and Google deals.”
post“A series of fairly wild thoughts from Dwarkesh. I thought I was bullish but even I am not assuming that the price to rent compute continues to go up! I will say that I am in SV this week and Dwarkesh is capturing the zeitgeist. Net, net public markets would probably be trading https://t.co/PzIxxnzbiS”
postBaker asks where Anthropic's and OpenAI's combined 150b ARR came from, disputes that leveraging up is only funding path
“@dampedspring Where has Anthropic’s and OpenAI’s combined 150b of ARR come from? I don’t think the idea that leveraging up is the only way to fund this is at all correct.”
post“Market is overreacting to hyperscale credit spreads widening from my perspective. TL;DR Spot pricing for renting GPU compute materially above contracted rates implies hyperscalers are underearning while operating cash flow acceleration is an underestimated source of funds for AI https://t.co/D9FSHXQ8F8”
6 more posts
postBaker endorses data centers as beneficial for America, reacting to Quincy Washington's economic transformation
“Data centers are awesome for America in every way. https://t.co/gSe3Gf2kYp”
postBaker suggests NVIDIA accelerators have residual value that distinguishes them from first-party ASICs, responding to concerns about financing OpenAI hardware
“@dylan522p @TShirtnJeans2 @maxkan Some accelerators (the green ones) are not like the other accelerators.”
postBaker speculates Ultrafast will use GPUs plus CS-4/5, notes Jalapeño not financeable at GPU rates
“@dylan522p @maxkan Maybe GPUs plus CS-4/5 for Ultrafast at scale. Jalapeño not financeable at comparable rates to GPUs and probably not that big of a wafer allocation next 2 years. Jalapeño still a really impressive achievement from OpenAI. I remember being sad when they hired Thomas away from https://t.co/a0jIXXahXN”
postBaker questions whether AFD is compatible with data locality, disputes OpenAI bet against disaggregation
“@maxkan Hmm. I might ask @dylan522p if Attention-FFN disaggregation (AFD) is compatible with data locality and not moving the KV cache. That constraint is what OpenAI focused on with Jalapeño. They absolutely did not make an “explicit bet that disagg is not the way to go.” There are https://t.co/2lVjH2M5vS”
postBaker notes compute brought on in June may show Q3 acceleration
“@armandorvila They brought on a lot of compute in the month of June so may be more of a Q3 acceleration there - we shall see very shortly”
post“@NoonHillCapMgmt Always wise to be skeptical. Nothing wrong with the analogy, just the barriers to entry are higher, quality of companies is materially higher and structurally harder to bring on capacity given shortages.”
-
Baker insists data centers deliver economic benefits including job creation and lower electricity costs, calling opponents unqualified for office.
postBaker links to lower electricity costs in Louisiana related to data centers.
“Last one for today. Lower electricity costs in Louisiana: https://t.co/fZoBpgvYvr”
post“Yes. Love to see the unions fighting for datacenters. https://t.co/Tkq95Etcdn”
postBaker claims Georgia, Indiana, Michigan, and California are cutting electricity prices because of data centers.
“Georgia, Indiana, Michigan, California cutting electricity prices because of data centers. https://t.co/QoVbyIYNxK”
postBaker says more examples of data center benefits will emerge.
“Going to see more and more of these: https://t.co/zcdyY3KBtD”
postBaker declares politicians opposed to data centers unqualified, citing job creation and lower electricity costs.
“Nature is healing. Any politician opposed to data centers is unqualified to hold public office. Data centers are revitalizing small towns all over America, a godsend for blue collar workers, lowering electricity costs, accelerating the transition to sustainable energy and use https://t.co/fTH3qZTP2V”
postBaker endorses data centers as beneficial for America, reacting to Quincy Washington's economic transformation
“Data centers are awesome for America in every way. https://t.co/gSe3Gf2kYp”
postBaker questions whether AFD is compatible with data locality, disputes OpenAI bet against disaggregation
“@maxkan Hmm. I might ask @dylan522p if Attention-FFN disaggregation (AFD) is compatible with data locality and not moving the KV cache. That constraint is what OpenAI focused on with Jalapeño. They absolutely did not make an “explicit bet that disagg is not the way to go.” There are https://t.co/2lVjH2M5vS”
-
Baker says NVIDIA accelerators retain residual value unlike first-party ASICs and predicts every accelerator company will adopt 3D DRAM variants.
postBaker suggests NVIDIA accelerators have residual value that distinguishes them from first-party ASICs, responding to concerns about financing OpenAI hardware
“@dylan522p @TShirtnJeans2 @maxkan Some accelerators (the green ones) are not like the other accelerators.”
postBaker notes Jalapeño participated in third-party benchmarks before TPUs did, suggesting competitive confidence
“Also pretty funny that Jalapeño participated in a third party benchmark *before* TPUs! Tells you a lot!”
postBaker says real comparison will be Jalapeño versus Rubin Ultra and Feynmann given current shipping timelines
“Should note the real comparison will be with Rubin Ultra given Rubin shipping today so kinda have to wait. And Feynmann coming fast. But still impressive debut from Jalapeño.”
postBaker credits Jalapeño as first good ASIC outside TPU and Trainium but says disaggregated GPU plus SRAM will outperform
“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
postBaker says GPU price increase means anyone spending on Blackwell and Rubin before January 2027 gains competitive advantage
“The most obvious implication of GPU price increase is that anyone who is spending a lot on Blackwell and Rubin before January 31, 2027 is in a good relative position vs. competitors. Super curious to know if GPU prices are going up more than rack prices.”
postBaker predicts every accelerator company will do some variant of 3D DRAM or zHBM discussed at HotChips
“Every accelerator company, and I mean every accelerator company, is likely going to do some variant of 3D DRAM or the “zHBM” which were both discussed at HotChips today. The larger companies will probably run (are running) parallel programs, but these two memory variants”
postBaker says Jensen was number one on Glassdoor best CEOs, roughly half Nvidia employees worth over twenty-five million, 75% millionaires
“Jensen was #1 on Glassdoor’s list of the best CEOs to work for. Still the same guy he was when I met him 27 years ago. Maybe slightly calmer. The ultimate testament is that roughly half of the employees at Nvidia are worth over $25 million, 75% are millionaires and essentially https://t.co/iS2zgYE0jI”
postBaker argues game theory suggests memory LTAs are durable and Nvidia may have significant cloud revenue shares
“Game theory suggests memory LTAs are likely to be durable. And yes, Nvidia may have a significant cloud business via revenue shares. https://t.co/GFN1kDQmSk”
1 more post
post“Delighted to announce @JensenHuang as the first confirmed speaker for the @TomorrowXSummit and even happier that I will be interviewing him. We will have Chicken Fried Steak at the summit. Sawyer’s is my favorite place in Austin but we will of course have Denny’s. https://t.co/4QKBMGh4TU”
3 more topics
-
Baker dismisses SRAM scaling concerns as settled and emphasizes memory per FLOP as the key driver of tokens per watt.
postBaker credits Jalapeño as first good ASIC outside TPU and Trainium but says disaggregated GPU plus SRAM will outperform
“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
postBaker dismisses SRAM scaling concerns as settled for anyone close to logic, memory and packaging roadmaps at wafer scale
“@IanCutress Luckily for me, I have never been near a cliff or a bridge when I have read a post about how “SRAM cannot scale.” This topic has largely been put to bed for anyone remotely close to logic/memory/packaging roadmaps especially at wafer scale.”
postBaker says no one thought someone was crazy and everyone will likely do 3D DRAM or zHBM
“@bubbleboi No one thought you were crazy. Everyone likely going to do this or zHBM. *Everyone*”
postBaker predicts every accelerator company will do some variant of 3D DRAM or zHBM discussed at HotChips
“Every accelerator company, and I mean every accelerator company, is likely going to do some variant of 3D DRAM or the “zHBM” which were both discussed at HotChips today. The larger companies will probably run (are running) parallel programs, but these two memory variants”
postBaker announces Micron executive Sumit Sadana as speaker and says memory per FLOP drives tokens per watt
“Happy to announce Sumit Sadana EVP and Chief Business Officer of @MicronTech as a confirmed speaker at the @TomorrowXSummit in November. Memory per FLOP drives tokens per watt. Looking forward to hearing Sumit’s thoughts on how AI is changing the business model of memory https://t.co/k1Pdk94PDu”
post“Looking forward to interviewing @andrewdfeldman CEO and founder of @cerebras at the @TomorrowXSummit in Austin November 17-18. Scale-up networking and SRAM acceleration are the best solutions to the DRAM shortage from my perspective. Given that Nebius is monetizing at $20 to $25 https://t.co/ZUAzwy8A5d”
post“Jitendra Mohan, CEO and co-founder of Astera Labs, is our third confirmed speaker for the @TomorrowXSummit. Really looking forward to discussing the role of scale-up networking in a memory constrained world and the intersection of their switch franchise with optics. https://t.co/stwGpJ8zox”
postBaker argues game theory suggests memory LTAs are durable and Nvidia may have significant cloud revenue shares
“Game theory suggests memory LTAs are likely to be durable. And yes, Nvidia may have a significant cloud business via revenue shares. https://t.co/GFN1kDQmSk”
-
Baker defends Cerebras economics and suggests critics review the company's patent filings before questioning yield or unit economics publicly.
postBaker speculates Ultrafast will use GPUs plus CS-4/5, notes Jalapeño not financeable at GPU rates
“@dylan522p @maxkan Maybe GPUs plus CS-4/5 for Ultrafast at scale. Jalapeño not financeable at comparable rates to GPUs and probably not that big of a wafer allocation next 2 years. Jalapeño still a really impressive achievement from OpenAI. I remember being sad when they hired Thomas away from https://t.co/a0jIXXahXN”
postBaker hypothesizes tomorrow's Ultrafast will be GPUs or Jalapeño with CS-4 or CS-5
“@dylan522p @maxkan And they also said this followed by “Cerebras FTW.” I would be surprised if their head of Codex was unaware of their roadmap. Reasonable to hypothesize tomorrow’s Ultrafast will be either GPUs or Jalapeño working with CS-4 or CS-5. We shall see! https://t.co/QM0vs809x7”
postBaker says Cerebras sells racks not chips and questions claim about bad unit economics
“@insane_analyst @andrewdfeldman @cerebras @TomorrowXSummit They sell racks not chips and why do you say the unit economics are bad?”
postBaker warns questioner to review Cerebras patent filings before he asks their questions publicly
“@insane_analyst @andrewdfeldman @cerebras @TomorrowXSummit I would not recommend this course of action but I am happy to ask your questions. Just be sure you want me to ask and attribute to you *after* you have carefully reviewed their patent filings. Always looking out for you as appreciate your contributions both here and IRL.”
postBaker offers to discuss Cerebras yields and points to patent filings to settle the issue
“@masayoshisson @andrewdfeldman @cerebras @TomorrowXSummit Happy to discuss yields to put this silly issue to bed forever. In the meantime, I would point you to their patent filings in the public domain.”
post“Looking forward to interviewing @andrewdfeldman CEO and founder of @cerebras at the @TomorrowXSummit in Austin November 17-18. Scale-up networking and SRAM acceleration are the best solutions to the DRAM shortage from my perspective. Given that Nebius is monetizing at $20 to $25 https://t.co/ZUAzwy8A5d”
-
Baker argues monopolar frontier AI model control is dangerous and worries negative regulatory visions are winning with unpredictable consequences.
postBaker agrees cyber is defense-dominant and bio offense-dominant, with bio being the best pro-regulation argument
“@_sholtodouglas Ok. Becky just brought me another gin and tonic and is watching our daughter. So: I think the framing of cyber as defense dominant and bio as offense dominant is sensible and the bio point is probably the best pro-regulation argument. And the USG should absolutely have all”
postBaker worries the negative vision is winning with unpredictable consequences for everyone
“@sashadem @theallinpod Likewise appreciate your response and generally agreed. Am just a little worried that the negative vision is “winning” right now with very unpredictable consequences for everyone. Reasonable minds can minds can disagree about the best path forward for the most important”
postBaker says Dario's public commentary makes it believable he said Anthropic might be the only private company
“@sashadem @theallinpod I believe both you and Sholto. This issue is that Dario’s own public commentary makes it believable to many serious people in Silicon Valley that Dario might have said something similar to “Anthropic might end up being the only private company.” I would love nothing more than to https://t.co/NF7MaO8wl1”
postBaker says multiple serious Silicon Valley people believe Anthropic wants to be only private company, consistent with Dario's messaging
“@_sholtodouglas Sholto, thank you for setting the record straight. Larger issue is that multiple very serious people in Silicon Valley have heard some variation of this and believe it to be true. And the reason it is believable to so many is that it is consistent with Dario’s public messaging”
postBaker argues monopolar or bipolar frontier model world is dangerous, open-source provides necessary diversity reflecting human variation
“A monopolar or bipolar world of frontier models is a very dangerous world for humans. This is why open-source is so important - we need a rich diversity of models that reflect the richness of human variation and let us all find the model that is best for each of us. Also why so https://t.co/xxMglmQZJC”
Then and now 50 pairs
-
consistent data centers · 1 days apartThen post
“Data centers are awesome for America in every way. https://t.co/gSe3Gf2kYp”
Now post“Last one for today. Lower electricity costs in Louisiana: https://t.co/fZoBpgvYvr”
-
consistent data centers · 0 days apartThen post
“Going to see more and more of these: https://t.co/zcdyY3KBtD”
Now post“Last one for today. Lower electricity costs in Louisiana: https://t.co/fZoBpgvYvr”
-
consistent data centers · 0 days apartThen post
“Georgia, Indiana, Michigan, California cutting electricity prices because of data centers. https://t.co/QoVbyIYNxK”
Now post“Last one for today. Lower electricity costs in Louisiana: https://t.co/fZoBpgvYvr”
-
consistent data centers · 0 days apartThen post
“Nature is healing. Any politician opposed to data centers is unqualified to hold public office. Data centers are revitalizing small towns all over America, a godsend for blue collar workers, lowering electricity costs, accelerating the transition to sustainable energy and use https://t.co/fTH3qZTP2V”
Now post“Last one for today. Lower electricity costs in Louisiana: https://t.co/fZoBpgvYvr”
-
consistent data centers · 0 days apartThen post
“Yes. Love to see the unions fighting for datacenters. https://t.co/Tkq95Etcdn”
Now post“Last one for today. Lower electricity costs in Louisiana: https://t.co/fZoBpgvYvr”
-
consistent nvidia · 2 days apartThen post
“The most obvious implication of GPU price increase is that anyone who is spending a lot on Blackwell and Rubin before January 31, 2027 is in a good relative position vs. competitors. Super curious to know if GPU prices are going up more than rack prices.”
Now post“@dylan522p @TShirtnJeans2 @maxkan Some accelerators (the green ones) are not like the other accelerators.”
-
consistent inference · 0 days apartThen post
“Should note the real comparison will be with Rubin Ultra given Rubin shipping today so kinda have to wait. And Feynmann coming fast. But still impressive debut from Jalapeño.”
Now post“Also pretty funny that Jalapeño participated in a third party benchmark *before* TPUs! Tells you a lot!”
-
consistent inference · 0 days apartThen post
“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
Now post“Also pretty funny that Jalapeño participated in a third party benchmark *before* TPUs! Tells you a lot!”
-
consistent memory · 12 days apartThen post
“Looking forward to interviewing @andrewdfeldman CEO and founder of @cerebras at the @TomorrowXSummit in Austin November 17-18. Scale-up networking and SRAM acceleration are the best solutions to the DRAM shortage from my perspective. Given that Nebius is monetizing at $20 to $25 https://t.co/ZUAzwy8A5d”
Now post“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
-
consistent memory · 8 days apartThen post
“Happy to announce Sumit Sadana EVP and Chief Business Officer of @MicronTech as a confirmed speaker at the @TomorrowXSummit in November. Memory per FLOP drives tokens per watt. Looking forward to hearing Sumit’s thoughts on how AI is changing the business model of memory https://t.co/k1Pdk94PDu”
Now post“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
-
consistent memory · 1 days apartThen post
“@bubbleboi No one thought you were crazy. Everyone likely going to do this or zHBM. *Everyone*”
Now post“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
-
consistent memory · 1 days apartThen post
“Every accelerator company, and I mean every accelerator company, is likely going to do some variant of 3D DRAM or the “zHBM” which were both discussed at HotChips today. The larger companies will probably run (are running) parallel programs, but these two memory variants”
Now post“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
-
consistent memory · 0 days apartThen post
“@IanCutress Luckily for me, I have never been near a cliff or a bridge when I have read a post about how “SRAM cannot scale.” This topic has largely been put to bed for anyone remotely close to logic/memory/packaging roadmaps especially at wafer scale.”
Now post“Impressive that Jalapeño outperforms the comparable TPU and is in the mix with Rubin. Credit where credit is due - first good ASIC outside of TPU/Trainium. However, will likely significantly underperform a disaggregated GPU/Trainum plus SRAM accelerator setup. Especially with”
-
consistent spacex · 15 days apartThen said
“What it was is they saw SpaceX have a big installed base of compute and sell some big trading optimized clusters into the market at a truly massive premium to these contracted rates.”
Now post“Fireworks and Baseten are the two largest inference clouds and I think of them as *an* index for open-source AI adoption. Also first time seeing SpaceXAI on the list of “fastest growing” IIRC. SambaNova! Will wonders never cease. https://t.co/HlmFsKRtVk”
-
consistent regulation · 2 days apartThen post
“A monopolar or bipolar world of frontier models is a very dangerous world for humans. This is why open-source is so important - we need a rich diversity of models that reflect the richness of human variation and let us all find the model that is best for each of us. Also why so https://t.co/xxMglmQZJC”
Now post“@_sholtodouglas Ok. Becky just brought me another gin and tonic and is watching our daughter. So: I think the framing of cyber as defense dominant and bio as offense dominant is sensible and the bio point is probably the best pro-regulation argument. And the USG should absolutely have all”
-
consistent anthropic · 87 days apartThen said
“And these three companies employ thousands of people, tens of thousands collectively. They've all spent ten years building their businesses and Anthropic added their combined businesses in one month.”
Now post“@sashadem @theallinpod I believe both you and Sholto. This issue is that Dario’s own public commentary makes it believable to many serious people in Silicon Valley that Dario might have said something similar to “Anthropic might end up being the only private company.” I would love nothing more than to https://t.co/NF7MaO8wl1”
-
consistent anthropic · 87 days apartThen said
“And these three companies employ thousands of people, tens of thousands collectively. They've all spent ten years building their businesses and Anthropic added their combined businesses in one month.”
Now post“@_sholtodouglas Sholto, thank you for setting the record straight. Larger issue is that multiple very serious people in Silicon Valley have heard some variation of this and believe it to be true. And the reason it is believable to so many is that it is consistent with Dario’s public messaging”
-
consistent anthropic · 85 days apartThen said
“And these three companies employ thousands of people, tens of thousands collectively. They've all spent ten years building their businesses and Anthropic added their combined businesses in one month.”
Now post“$3 trillion for Anthropic in the FT. I feel like I’ve heard that somewhere else. Although public markets generally struggle to accurately value growth when it is this high. https://t.co/2R80q18GUH”
-
consistent nvidia · 0 days apartThen said
“I'm just making this up. But, like, do something. Like, you can because you have money now and credit markets are revolting.”
Now said“I would be going to the buyers of GPUs, Traniums, and whoever and say, I'll participate in the NVIDIA credit wrapper. Now their business is just inherently less stable and predictable,”
-
consistent memory · 0 days apartThen said
“It's the single most important thing you could do to increase token output per unit of compute, and then that obviously, definitionally, actually lowers costs,”
Now said“I just think given that this is, the axis around which kind of everything is revolving, man, like, might blow up your entire business and your franchise by breaking an LTA.”
-
consistent memory · 0 days apartThen said
“Well, if they're breaking their LTAs, it probably means, you know, your oversupply, prices are coming down, and then, you know, capacity naturally contracts. Well, what do you think is gonna happen to Google's allocations?”
Now said“I just think given that this is, the axis around which kind of everything is revolving, man, like, might blow up your entire business and your franchise by breaking an LTA.”
-
consistent memory · 0 days apartThen said
“I'm just making this up. But, like, do something. Like, you can because you have money now and credit markets are revolting.”
Now said“I just think given that this is, the axis around which kind of everything is revolving, man, like, might blow up your entire business and your franchise by breaking an LTA.”
-
consistent spacex · 0 days apartThen said
“What it was is they saw SpaceX have a big installed base of compute and sell some big trading optimized clusters into the market at a truly massive premium to these contracted rates.”
Now said“they put a vast amount of compute into the market overnight, and it wasn't even really a blip. It was like the market just utterly absorbed it.”
-
consistent spacex · 0 days apartThen said
“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
Now said“they put a vast amount of compute into the market overnight, and it wasn't even really a blip. It was like the market just utterly absorbed it.”
-
consistent spacex · 0 days apartThen said
“And they're monetizing at something like 50,000,000,000 a gig and consensus estimates for next year are 73,000,000,000. So forget Starlink v three, forget Starlink direct to sell, Grok 4.5 and Cursor.”
Now said“they put a vast amount of compute into the market overnight, and it wasn't even really a blip. It was like the market just utterly absorbed it.”
-
consistent spacex · 0 days apartThen said
“If they bring on anywhere near that, the consensus estimate is 73,000,000,000. That's eight gigs at 50,000,000,000 a gig. And, obviously, that would not all be lit up at the beginning of twenty seven.”
Now said“they put a vast amount of compute into the market overnight, and it wasn't even really a blip. It was like the market just utterly absorbed it.”
-
consistent memory · 0 days apartThen said
“I'm just making this up. But, like, do something. Like, you can because you have money now and credit markets are revolting.”
Now said“I would be going to the buyers of GPUs, Traniums, and whoever and say, I'll participate in the NVIDIA credit wrapper. Now their business is just inherently less stable and predictable,”
-
consistent ai infrastructure · 707 days apartThen said
“You could choose between faster time to market. If you run a 50% MFU and your competitor's running 40 for an equivalent amount of trading flops, you could be in market 25% faster.”
Now said“It's the single most important thing you could do to increase token output per unit of compute, and then that obviously, definitionally, actually lowers costs,”
-
consistent ai infrastructure · 707 days apartThen said
“MFU is the most important metric because it gives you all of these advantages and ways to differentiate yourself amongst five people, six people who've trained these GPT core class models.”
Now said“It's the single most important thing you could do to increase token output per unit of compute, and then that obviously, definitionally, actually lowers costs,”
-
consistent ai infrastructure · 707 days apartThen said
“You have the same amount of GPUs and the same amount of power presumably. You could choose between faster time to market.”
Now said“It's the single most important thing you could do to increase token output per unit of compute, and then that obviously, definitionally, actually lowers costs,”
-
consistent ai infrastructure · 707 days apartThen said
“MFU, model flops utilization, and that generally runs around 35 to 40%. And that's literally the percentage of compute, theoretical compute flops that you're actually applying to trading.”
Now said“It's the single most important thing you could do to increase token output per unit of compute, and then that obviously, definitionally, actually lowers costs,”
-
consistent power · 237 days apartThen said
“Goes from round numbers, 30 kilowatts, which is 30 American homes to 130 kilowatts, which is 130 American homes.”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent spacex · 237 days apartThen said
“I think the most important thing that's going to happen in the world, in this world, in the next three to four years is data centers in space.”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent power · 237 days apartThen said
“The sun is 30% more intense and this results in six times more irradiance in outer space than on planet earth.”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent power · 237 days apartThen said
“You can have the satellite always kind of catching light. The sun is 30% more intense and this results in six times more irradiance”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent power · 237 days apartThen said
“Going from hopper to Blackwell, first you go from air cooled to liquid cooled. The rack goes from weighing round numbers, a thousand pounds to 3,000 pounds.”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent power · 237 days apartThen said
“The rack goes from weighing round numbers, a thousand pounds to 3,000 pounds. Goes from round numbers, 30 kilowatts, which is 30 American homes to 130 kilowatts, which is 130 American homes.”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent anthropic · 76 days apartThen said
“I think Anthropic probably starts generating cash this year if they are not already generating cash, which I think is probably the case.”
Now said“I think one reason to the podcast two months ago, I got comfortable with that risk was just that you were seeing such incredible things out of anthropic.”
-
consistent anthropic · 76 days apartThen said
“And these three companies employ thousands of people, tens of thousands collectively. They've all spent ten years building their businesses and Anthropic added their combined businesses in one month.”
Now said“I think one reason to the podcast two months ago, I got comfortable with that risk was just that you were seeing such incredible things out of anthropic.”
-
consistent anthropic · 76 days apartThen said
“Anthropic, they added $11,000,000,000 of AR. And what is astonishing to me about this is that the SaaS and cloud revolution it created, we'll call it between 5 and $10,000,000,000,000 of value.”
Now said“I think one reason to the podcast two months ago, I got comfortable with that risk was just that you were seeing such incredible things out of anthropic.”
-
consistent power · 76 days apartThen said
“It's eight feet high. It's four feet deep, three feet wide. It's racks in space and SpaceX has showed you an illustration and it's a rack.”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent power · 76 days apartThen said
“It has these solar wings that are probably 500 feet long on each side. You keep it in a sun synchronous orbit.”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent spacex · 76 days apartThen said
“It's eight feet high. It's four feet deep, three feet wide. It's racks in space and SpaceX has showed you an illustration and it's a rack.”
Now said“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
consistent anthropic · 76 days apartThen said
“Anthropic added their combined businesses in one month. Nothing like that has ever happened in the history of capitalism.”
Now said“I think one reason to the podcast two months ago, I got comfortable with that risk was just that you were seeing such incredible things out of anthropic.”
-
consistent anthropic · 76 days apartThen said
“And I think maybe a true statement is that Infantropic could just wave a magic wand and get all the compute they wanted. They'd probably be doing well north of $100,000,000,000 today, maybe 150.”
Now said“I think one reason to the podcast two months ago, I got comfortable with that risk was just that you were seeing such incredible things out of anthropic.”
-
consistent pricing · 6 days apartThen post
“Market is overreacting to hyperscale credit spreads widening from my perspective. TL;DR Spot pricing for renting GPU compute materially above contracted rates implies hyperscalers are underearning while operating cash flow acceleration is an underestimated source of funds for AI https://t.co/D9FSHXQ8F8”
Now said“I don't think anyone in '24 or '25 thought that the prices of old GPUs would be going vertical. Everybody thought, hey. We're gonna be smart. We're gonna sign these long term contracts.”
-
consistent anthropic · 71 days apartThen said
“And these three companies employ thousands of people, tens of thousands collectively. They've all spent ten years building their businesses and Anthropic added their combined businesses in one month.”
Now post“@jukan05 4 gigawatts this year and another 4 next year per this report. Spot price for compute has gone up since they signed their Anthropic and Google deals.”
-
consistent power · 161 days apartThen said
“The rack goes from weighing round numbers, a thousand pounds to 3,000 pounds. Goes from round numbers, 30 kilowatts, which is 30 American homes to 130 kilowatts, which is 130 American homes.”
Now said“It's eight feet high. It's four feet deep, three feet wide. It's racks in space and SpaceX has showed you an illustration and it's a rack.”
-
consistent power · 161 days apartThen said
“You can have the satellite always kind of catching light. The sun is 30% more intense and this results in six times more irradiance”
Now said“It's eight feet high. It's four feet deep, three feet wide. It's racks in space and SpaceX has showed you an illustration and it's a rack.”
-
consistent spacex · 161 days apartThen said
“I think the most important thing that's going to happen in the world, in this world, in the next three to four years is data centers in space.”
Now said“It's eight feet high. It's four feet deep, three feet wide. It's racks in space and SpaceX has showed you an illustration and it's a rack.”
Minutes 205 quotes
everything on record, newest first
-
Baker notes that a 50% price cut in GPT 5.6 drove 14x volume increase, calling it wild.
X aiconsumption -
Baker says more examples of data center benefits will emerge.
X americadata -
Baker agrees that Jensen is competing with OpenAI and predicts American open-weight models will approach the frontier.
X aicompetition -
Baker notes seven-month GPU rental price increase from mid-$2 to nearly $4 per hour is surprising.
compute costgpu pricingReceipt
“And they're hoping seven months later to pay just under $4. Like, you know, just you hear this today. And that's, like, that's pretty crazy”
-
Baker argues companies shifting to cheaper open source models may reduce spend while increasing GPU compute hours consumed.
compute demandmodel routingReceipt
“a company getting smarter about which model to use for which task, that, you know, that may lead to a a stabilization in their spend or even a decline, but it actually has nothing to do with the amount of, you know, GPU compute hours they are effectively consuming”
-
Baker calculates 20% token spend on knowledge work implies $5 trillion addressable market.
ai capexlaborReceipt
“There's 25,000,000,000,000 in knowledge work. And so let's, you know, let's say that that's, you know, let's take your 20% number. That's 5,000,000,000,000.”
-
Baker cites examples showing blue collar workers benefiting significantly from AI adoption.
economic impactlaborReceipt
“I think one of them gave an example of I forget if it was a plumber or an HVAC contractor, but, like, you know, everybody who's a blue collar worker is doing great because of AI.”
-
Baker argues breaking memory supply agreements could destroy franchises due to allocation loss risk.
memorystrategyReceipt
“I just think given that this is, the axis around which kind of everything is revolving, man, like, might blow up your entire business and your franchise by breaking an LTA.”
-
Baker suggests memory companies should copy NVIDIA's credit wrapper model despite business instability risks.
business modelmemoryReceipt
“I would be going to the buyers of GPUs, Traniums, and whoever and say, I'll participate in the NVIDIA credit wrapper. Now their business is just inherently less stable and predictable,”
-
Baker notes SpaceX added massive compute capacity instantly without any market price impact.
computedemandReceipt
“they put a vast amount of compute into the market overnight, and it wasn't even really a blip. It was like the market just utterly absorbed it.”
-
Baker cites report SpaceX plans eight gigawatts of compute, calling it incredible, with rates up not down.
compute capacitypricingReceipt
“They think that SpaceX is gonna try and bring on eight gigawatts of compute. I will never bet against Elon, but, I mean, that would be a truly incredible feat.”
-
Baker reports AI stocks down 40-60% in a month despite no negative quantitative metrics or deceleration.
ai capexfundamentalsReceipt
“Loads of AI names are down 60% from their highs. We'll call it 40% to 60% in a month in a straight line.”
-
Baker says no one predicted old GPU prices would go vertical, everyone expected gradual declines.
ai capexnvidiaReceipt
“I don't think anyone in '24 or '25 thought that the prices of old GPUs would be going vertical. Everybody thought, hey. We're gonna be smart. We're gonna sign these long term contracts.”
Also said on Invest Like The Best, 4 Aug 2026
-
Baker argues contracted compute trades at massive discount to spot, repricing will accelerate cash flows and answer ROI questions.
ai capexinferenceReceipt
“And so, essentially, you have the contracted base of installed compute trading at a massive discount to the current spot market.”
-
Baker reports hyperscaler operating cash flow accelerated from $28B to $32B despite unusual legal expenses.
ai capexcash flowReceipt
“Operating cash flow from Microsoft, Meta, and Amazon has reported accelerated from '28 to '32. There are some actually pretty big unusual items now like these hyperscalers.”
-
Baker says adjusting for one-time items shows hyperscaler operating cash flow accelerated from $28B to $35B.
ai capexcash flowReceipt
“But there is an unusual amount of onetimers this quarter. And if you adjust for that, we went from 28 to 35. That's that's a material acceleration at this scale.”
-
Baker explains Meta's compute rental was inspired by SpaceX selling clusters at massive premiums to contracted rates.
compute pricingmetaReceipt
“What it was is they saw SpaceX have a big installed base of compute and sell some big trading optimized clusters into the market at a truly massive premium to these contracted rates.”
-
Baker explains Silicon Data token index dip reflects mix shift to cheaper open source, not demand weakness.
market dataopen sourceReceipt
“What the Silicon Data token index captures is mix, and they don't see all the tokens.”
-
Baker says incredible Anthropic results made him comfortable with Blackwell air pocket risk.
ai capexanthropicReceipt
“I think one reason to the podcast two months ago, I got comfortable with that risk was just that you were seeing such incredible things out of anthropic.”
-
Baker reports GPU rental prices doubled from mid-$2 to nearly $4 per hour over seven months for identical clusters.
inferencenvidiaReceipt
“And they had rented a cluster of several thousand black wells, and we'll just call it somewhere in the mid $2 per GPU hour.”
185 more the default view shows 20
-
Baker reports inference cloud expects to pay 100% more for Blackwells when contracts expire, showing hyperscalers are underearning.
gpu pricinginference“They went on a podcast, and they essentially said, we are planning to pay 100% more for Blackwell's when our contract expires. And that just means that essentially all the hyperscalers are under earning.”
-
Baker notes NVIDIA trades at its lowest forward PE in ten years.
nvidiavaluation“NVIDIA is actually, as we record this, at its lowest forward PE of the last ten years.”
-
Baker identifies four companies that matter at scale in long-term supply agreements for AI chips.
competitionmarket structure“Let's just think about the game theory of breaking an LTA. So there's four companies that matter at scale. There's Amazon with their tradiums.”
-
Baker argues only four companies matter at scale for AI chips: Amazon, Google, AMD, and NVIDIA.
ltasmarket structure“So there's four companies that matter at scale. There's Amazon with their tradiums. There's Google with their TPUs. There's AMD, and then there's NVIDIA who's, like, much bigger than everybody else combined.”
-
Baker identifies memory as the single most important factor for increasing token output per compute unit.
ai infrastructurememory“It's the single most important thing you could do to increase token output per unit of compute, and then that obviously, definitionally, actually lowers costs,”
-
Baker explains breaking memory LTAs could destroy companies by losing allocations when market tightens again.
ai capexmemory“Well, if they're breaking their LTAs, it probably means, you know, your oversupply, prices are coming down, and then, you know, capacity naturally contracts. Well, what do you think is gonna happen to Google's allocations?”
-
Baker suggests memory companies should copy NVIDIA's credit wrapper model while they have cash and credit markets are tight.
ai capexmemory“I'm just making this up. But, like, do something. Like, you can because you have money now and credit markets are revolting.”
-
Baker cites Dwarkesh noting compute margins rising alongside volume and inference margins, driving lab acceleration.
compute marginseconomics“Borgesch, he's a very smart guy. He's very plugged in. Then he pointed out that margins on compute are going up.”
-
Baker cites analysis showing compute margins, quantity, and inference margins all rising simultaneously, driving lab acceleration.
ai capexinference“The amount of compute is going up and inference margins going up. And if you multiply those three, that's how you're getting this crazy acceleration into some of the labs plus open source,”
-
Baker cites report that SpaceX plans to bring on eight gigawatts of compute in eighteen months, calling it incredible.
ai capexpower“A Substack writer will fund the AI. They think that SpaceX is gonna try and bring on eight gigawatts of compute over the next eighteen months.”
-
Baker estimates SpaceX monetizes compute at $50B per gigawatt versus $73B consensus, with Grok and Cursor hitting $10B ARR quickly.
inferencerevenue“And they're monetizing at something like 50,000,000,000 a gig and consensus estimates for next year are 73,000,000,000. So forget Starlink v three, forget Starlink direct to sell, Grok 4.5 and Cursor.”
-
Baker calculates eight gigawatts at $50B per gigawatt implies $400B revenue versus $73B consensus for SpaceX.
ai capexrevenue“If they bring on anywhere near that, the consensus estimate is 73,000,000,000. That's eight gigs at 50,000,000,000 a gig. And, obviously, that would not all be lit up at the beginning of twenty seven.”
-
Generating Alpha Podcast, 8 July 2026
bubbleearly career“I published a note, I think in, February 2000 after six weeks of looking at these stocks.”
-
Baker explains Hock Tan's model is opposite of Dell's: find drained profit pools with no competition, enter, and raise prices.
broadcombusiness model“Hock's model has always been to find a profit pool that has been drained and there's no competition left, enter that and jack prices up.”
-
Baker says Jensen Huang was obviously exceptional when they first met in early 2000s.
jensen huangnvidia“I was a pretty young man. It was obvious to me he was exceptional, like for sure one of the most exceptional people I'd ever met.”
-
Baker argues that whether AI buildout is funded by debt or cash flows is critical for bubble risk.
ai bubblecapex“And a very important distinction we will come to is whether or not that build out is funded out of debt or cash flows. It's a critical distinction for AI.”
-
Baker says Internet bubble scars with 80-85% drawdowns keep a lid on tech valuations today.
drawdownsinternet bubble“In the bubble, you went down 80 or 85. Dramatically different. And so the scars are so deep from that bubble that just that puts a lid on tech valuations.”
-
Baker argues TSMC remains a bottleneck even if orbital data centers solve power shortages.
orbital computepower“But even if we solve the shortage of Watts with orbital data centers, the TSM is still a bottleneck. And semi's and semiconductor manufacturing, you have a big advantage if you were ahead.”
-
Baker argues silicon will eat the world because AI is far more computationally intensive than deterministic software.
aicompute“And the nature of AI just as AI eats the world, silicon is going to eat the world because it's so much more computationally intensive than deterministic software written by humans.”
-
Baker explains AI recomputes answers probabilistically each time, enabling superhuman capabilities but requiring extreme computational expense.
aicompute“AI, even if you put a harness on it, even if you do the chain of thought, even if you have multiple agents, it's probabilistic that it is recomputing the answer each time.”
-
Baker calculates rendering Monopoly Go with VEO3 would cost over 100x the game's revenue.
aigaming“Monopoly Go is a game where we have the revenue and the hours played and it is to render it using list prices for something like VEO3 is more than two orders of magnitude greater than its revenue. So what is the role for human creativity, man? I don't know.”
-
Baker says rendering Monopoly Go with AI would cost over 100x its revenue at list prices.
ai inferenceeconomics“So what is the role for human creativity, man? I don't know. I don't know. Like what I hope, one of the reasons Neuralink is such an exciting company”
-
Baker quotes mentor Jennifer Yurig: investors either panic early or double down late, not both.
decision-makingdrawdowns“Ultimately, as an investor, you either have to panic early or double down late. And essentially no one does both. And know thyself.”
-
Baker's consistent advice to young people is to be both kind and scrappy, a rare combination.
careerlife advice“And I always give the same advice, be kind and be scrappy. A lot of people who are kind are not scrappy and a lot of people are scrappy are not kind.”
-
Baker recounts Schwartzman's story of being blown off by a Boston endowment during Blackstone's early fundraising struggles.
blackstoneearly struggles“I was on a hot Friday afternoon and everybody at the endowment had left. They got completely blown off and then it started pouring rain.”
-
Baker says Anthropic added $11B ARR, comparing it to the entire SaaS revolution's $5-10T value creation.
ai capexanthropic“Anthropic, they added $11,000,000,000 of AR. And what is astonishing to me about this is that the SaaS and cloud revolution it created, we'll call it between 5 and $10,000,000,000,000 of value.”
-
Baker says Anthropic added $11B ARR in one month, matching what Palantir, Snowflake, and Databricks built in ten years combined.
ai capexanthropic“And these three companies employ thousands of people, tens of thousands collectively. They've all spent ten years building their businesses and Anthropic added their combined businesses in one month.”
-
Baker says Anthropic adding $11B ARR in one month is unprecedented in capitalism's history.
ai capexanthropic“Anthropic added their combined businesses in one month. Nothing like that has ever happened in the history of capitalism.”
-
Baker estimates Anthropic would be doing $100-150B ARR if not compute-constrained, versus current $50B.
ai capexanthropic“And I think maybe a true statement is that Infantropic could just wave a magic wand and get all the compute they wanted. They'd probably be doing well north of $100,000,000,000 today, maybe 150.”
-
Baker estimates Anthropic could be doing $100-200 billion in revenue if unconstrained by compute availability.
anthropiccompute“So I think they would be doing materially more a hundred, one hundred and fifty, maybe 200,000,000,000. So you might be buying it at more like five times unconstrained.”
-
Baker estimates Anthropic could do $100-200B in unconstrained revenue if compute weren't limited.
anthropiccompute“So you might be buying it at more like five times unconstrained. I'm gonna make up a new number. URR, unconstrained revenue.”
-
Baker believes Anthropic is likely already generating cash or will start this year.
anthropicinference“I think Anthropic probably starts generating cash this year if they are not already generating cash, which I think is probably the case.”
-
Baker describes orbital compute as Blackwell-sized racks in space with 500-foot solar wings, not giant data centers.
orbital computepower“It's eight feet high. It's four feet deep, three feet wide. It's racks in space and SpaceX has showed you an illustration and it's a rack.”
-
Baker describes orbital compute as Blackwell-sized racks in space with 500-foot solar wings in sun-synchronous orbit.
orbital computepower“It has these solar wings that are probably 500 feet long on each side. You keep it in a sun synchronous orbit.”
-
Baker argues America will consume all available compute, reducing edge AI bear case concerns.
ai capexinference“And I just think the same is true of compute. It's why I'm probably less worried about like an edge AI bear case than I was.”
-
Baker argues America will consume all available compute, making him less worried about edge AI bear cases.
ai capexedge ai“It's why I'm probably less worried about like an edge AI bear case than I was. We're going to consume as much compute as we can.”
-
Baker reveals Jensen Huang has never had a contract with TSMC, operating entirely on handshakes and fairness.
nvidiatsmc“I do think it's wild that Jensen has never had a contract with Taiwan Semi. They do business on what seems fair in handshakes.”
-
Baker argues NVIDIA could sell $2-3 trillion in GPUs if TSMC expanded capacity, but TSMC's restraint prevents a bubble.
ai capexnvidia“If Taiwan Semi did what Jensen wanted, I think Nvidia could sell $2,000,000,000,000 of GPUs in '26 or '27, maybe 2,500,000,000,000, maybe 3,000,000,000,000. But there is a limit where consumers would consume so much.”
-
Baker argues Taiwan Semi's capacity constraint prevents AI bubble; NVIDIA could sell $2-3T GPUs otherwise.
ai capexnvidia“If Taiwan Semi did what Jensen wanted, I think Nvidia could sell $2,000,000,000,000 of GPUs in '26 or '27, maybe 2,500,000,000,000, maybe 3,000,000,000,000.”
-
Baker argues Taiwan Semi's capacity discipline is single-handedly preventing an AI bubble.
ai capexbubbles“So Taiwan Semi, if we don't get a bubble, we need to throw a party for them because they will have single handedly prevented a bubble.”
-
Baker identifies Taiwan Semi's capacity decisions as the single most important indicator of AI bubble risk.
ai capexbubble risk“The pace at which they expand capacity. If I were to watch one thing to understand where there's a bubble, it's Taiwan Semi's capacity decisions.”
-
Baker explains TSMC caught Intel because semiconductor equipment suppliers wanted to avoid Intel's monopsony power.
intelsemiconductors“They wanted them to catch up. They don't like having a monopsony. The a teams were in Taiwan working. Intel made some mistakes and presto.”
-
Baker predicts Musk will build Taiwan town, Japan town, and Koreatown near TerraFab to recruit top engineers.
manufacturingspacex“And then we'll have Japan town, same thing. We're going to have Koreatown. We're going to have all these things exactly dialed to recruit the best engineers.”
-
Baker explains harness engineering matters significantly and harnesses are increasingly co-developed with models.
ai developmentinference“And it turns out that harness engineering is not as important as the model, but it really matters. And these harnesses in these models are increasingly being co developed.”
-
Baker says understanding frontier AI now requires enterprise usage-based plans, not $250 monthly subscriptions which are rate-limited.
ai capexanthropic“To understand what Frontier AI is capable of today, even for a non coding use case, need to have Cloud Code or Codex five point Codex.”
-
Baker says understanding frontier AI now requires enterprise usage-based plans, not consumer subscriptions, due to rate limiting.
ai modelsinference“To understand what Frontier AI is capable of today, even for a non coding use case, need to have Cloud Code or Codex five point Codex. And you need to be on an enterprise plan.”
-
Baker says AI shifting from flat pricing to usage-based is extremely bullish as people consume more AI.
ai capexinference“AI is just shifting from all you can eat to pay by the drink. Then it turns out people really like to talk to their friends long distance.”
-
Baker predicts OpenAI and Anthropic will exceed $200B ARR this year due to shift to usage-based pricing.
anthropicinference“So I think the shift to usage based pricing is probably why you will see OpenAI and Anthropic exceed well over $200,000,000,000 in ARR this year.”
-
Baker predicts OpenAI and Anthropic will exceed $200B ARR this year due to usage-based pricing shift.
anthropicinference“I think the shift to usage based pricing is probably why you will see OpenAI and Anthropic exceed well over $200,000,000,000 in ARR this year.”
-
Baker explains continual learning as models dynamically updating weights in real-time, unlike current reinforcement learning approaches.
aicontinual learning“Continual learning is a model that dynamically adjusts its weights or adjusts in some way in real time.”
-
Baker argues GPU useful lives will extend to 10-15 years due to inference disaggregation, contradicting AI skeptics.
ai capexinference“The disaggregation of inference means that I think these GPUs are going to have ten or fifteen year lives. The AI skeptics are like, oh, these companies are all cooking their books.”
-
Baker predicts GPUs will have 10-15 year useful lives due to prefill-inference disaggregation, extending older chips' value.
inferencememory“The disaggregation of inference means that I think these GPUs are going to have ten or fifteen year lives.”
-
Baker argues GPU useful lives will extend to 10-15 years due to prefill/decode disaggregation, not 1-2 years.
gpu lifecycleinference“The useful life of GPU is only a year or two. The useful life of CPU is only four years because the rapid technological change.”
-
Sohn Conference Foundation, 15 May 2026
contrarianmemory“I would say based on every memory cycle we have had for the last twenty five years, this is the time to be selling memory 100%.”
-
Baker takes the over on every AI capex and memory forecast number given by other speakers at the conference.
ai capexforecast“And I would just say, I take the over on every number that they gave. Every single number.”
-
Baker argues this may be the first true capacity cycle and that fundamental shortages help avoid a destructive bubble.
ai capexbubbles“This may be the first true capacity cycle. And I and and I and I do think that these fundamental shortages are good for us as investors.”
-
Baker predicts orbital compute will solve power shortages in five to seven years but wafer shortages will persist longer.
nvidiapower“We're gonna address the watt shortages with orbital compute for sure in the next five to seven years. But the wafer shortage, I think, is going to persist for a long time.”
-
Baker believes NVIDIA could sell $11.52 trillion in chips next year if TSMC tripled capacity but bubble risks prevent it.
ai capexmemory“And if they doubled or tripled capacity, like, NVIDIA could probably sell $11,520,000,000,000 worth of chips next year. I really believe that. But the other side of that might be very painful for everyone.”
-
Baker believes NVIDIA could sell $11.5 trillion worth of chips next year if TSMC expanded capacity sufficiently.
ai capexnvidia“And if they doubled or tripled capacity, like, NVIDIA could probably sell $11,520,000,000,000 worth of chips next year. I really believe that.”
-
Baker says AI models shifting to usage-based pricing with overage reveals no ceiling on spending yet.
ai capexinference“We're just moving from these all you can eat plans to usage based plans with overage, where those usage tokens cost a lot more, and we're finding out that there's we're nowhere near the amount of, you know, people ceiling price for how much they'll spend.”
-
Baker emphasizes only 0.1% of the world uses AI models properly yet there's massive shortage despite trillions spent.
ai capexinference“And we're in an insane shortage despite spending cumulatively trillions of dollars. What happens when 5% of the world's population is using these models the way the cutting edge 10 basis points are?”
-
Sohn Conference Foundation, 15 May 2026
ai capexinference“What happens when 5% of the world's population is using these models the way the cutting edge 10 basis points are? Like, it's just it's unimaginable. This is why orbital compute is a necessity.”
-
Sohn Conference Foundation, 15 May 2026
amazoninference“Tranium, by far. Tranium is going to be to 2026, especially in the second half of this year when Tranium three really ramps, as TPUs were to twenty twenty five.”
-
Baker predicts Trainium will dominate 2026 like TPUs did in 2025, with Trainium 3 ramping in second half.
ai capexinference“Tranium is going to be to 2026, especially in the second half of this year when Tranium three really ramps, as TPUs were to twenty twenty five.”
-
Baker argues Trainium is most underestimated because frontier mixture-of-expert models require switched scale-up networks for inference.
inferencenvidia“And so Trainium is for sure the most underestimated, not only because of those design choices, but because the all of these frontier models are what are called mixture of expert models.”
-
Baker states only NVIDIA and Amazon Trainium have functioning switched scale-up networks for inference today.
inferencenvidia“And the only two functioning switched scale up networks in the world today are the ones that power NVIDIA GPUs and Amazon's Trainiums.”
-
Baker reveals Atreides could have invested over $50 million in CoreWeave at $1.1 billion valuation but was conflicted out.
ai capexinference“I could've Atreides could've invested over $50,000,000 in the round at 1,100,000,000, And I was conflicted out by Crusoe,”
-
Baker predicts within seven years we may stop building terrestrial data centers, causing pain for infrastructure companies beforehand.
data centerspower“But I can't imagine a day when in the next seven years, when we never build another terrestrial data center. Mhmm.”
-
Baker predicts terrestrial data center buildout will stop within seven years, causing pain for power and cooling companies.
ai capexpower“And the years leading up to that are gonna be very painful for a lot of the companies and the power cooling spaces, you know, these industrial names,”
-
Baker says no one on earth knows how or why scaling laws for pre-training work, only that they do.
aigemini“They stated that unequivocally. And that's important because no one on planet earth knows how or why scaling laws for pre training work.”
-
Baker says no one on earth knows how or why scaling laws for pre-training work, it's an empirical observation.
aiscaling laws“And that's important because no one on planet earth knows how or why scaling laws for pre training work. It's actually not a law.”
-
Baker says ARC AGI scores jumped from 8% to 95% in three months with reasoning models.
ai progressmetrics“The reason we've had all this progress, maybe we could show like the ARC AGI slide where you had zero to eight over four years, zero to 8% intelligence, and then you went from 8% to 95% in three months when the first reasoning model came out from OpenAI.”
-
Baker details Blackwell's complexity: racks tripled in weight and quadrupled in power consumption versus Hopper.
blackwellnvidia“Going from hopper to Blackwell, first you go from air cooled to liquid cooled. The rack goes from weighing round numbers, a thousand pounds to 3,000 pounds.”
-
Baker explains Blackwell racks weigh 3,000 pounds and consume 130 kilowatts, equivalent to 130 American homes.
blackwellnvidia“The rack goes from weighing round numbers, a thousand pounds to 3,000 pounds. Goes from round numbers, 30 kilowatts, which is 30 American homes to 130 kilowatts, which is 130 American homes.”
-
Baker details Blackwell transition: racks triple in weight and quadruple in power from 30 to 130 kilowatts.
blackwellnvidia“Goes from round numbers, 30 kilowatts, which is 30 American homes to 130 kilowatts, which is 130 American homes.”
-
Baker argues reasoning saved AI by enabling progress during an 18-month gap when Blackwell delays would have halted everything.
ai progressblackwell“Had reasoning not come along, there would have been no AI progress from mid twenty twenty four through essentially Gemini three. There would have been none. Everything would have stalled.”
-
Baker explains Blackwell's complexity created deployment challenges that reasoning models had to bridge.
blackwellnvidia“Had reasoning not come along, there would have been no AI progress from mid twenty twenty four through essentially”
-
Baker argues reasoning models saved AI from an eighteen-month progress gap during Blackwell delays.
ai progressblackwell“Had reasoning not come along, there would have been no AI progress from mid twenty twenty four through essentially Gemini three.”
-
Baker argues reasoning models saved AI by bridging an eighteen-month gap while Blackwell was delayed.
ai progressblackwell“Had reasoning not come along, there would have been no AI progress from mid twenty twenty four through essentially Gemini three. There would have been none.”
-
Baker says AI is the first time in his career that being the low-cost producer matters in tech.
ai economicscost“And this is really important because AI is the first time in my career as a tech investor that being the low cost producers ever matter.”
-
Baker says AI is the first time in tech where being the low-cost producer matters economically.
ai economicscost“AI is the first time in my career as a tech investor that being the low cost producers ever matter. Apple is not worth trillions because they're a low cost producer of phones.”
-
Baker predicts xAI will release the first Blackwell-trained model in early 2026 because Elon builds fastest.
blackwellnvidia“One, we will see the first models trained on Blackwell in early twenty twenty six. I think the first Blackwell model will come from xAI.”
-
Baker predicts xAI will release the first Blackwell model because Elon builds data centers fastest.
blackwellprediction“I think the first Blackwell model will come from xAI. And the reason for that is just according to Jensen, no one builds data centers faster than Elon.”
-
Baker predicts Blackwell models will be amazing because pre-training scaling laws remain intact.
blackwellnvidia“We know that scaling laws for pre training are intact, And this means the Blackwell models are going to be amazing.”
-
Baker explains Google pays Broadcom 50-55% gross margins for back-end TPU design and Taiwan Semi management.
broadcomgoogle“I think part of that is, so Google, let's say the TPU, so there's front end and back end of semiconductor design, and then there's dealing with Taiwan Semi.”
-
Baker predicts data centers in space will be the most important development in three to four years.
data centersprediction“I think the most important thing that's going to happen in the world, in this world, in the next three to four years is data centers in space.”
-
Baker explains space offers six times more solar irradiance than Earth due to constant sunlight and intensity.
data centerspower“You can have the satellite always kind of catching light. The sun is 30% more intense and this results in six times more irradiance”
-
Baker explains space data centers get six times more solar irradiance than earth-based facilities.
powersolar“The sun is 30% more intense and this results in six times more irradiance in outer space than on planet earth. So you're getting a lot of solar energy.”
-
Baker explains space offers six times more solar irradiance than Earth for data centers.
powersolar“The sun is 30% more intense and this results in six times more irradiance in outer space than on planet earth.”
-
Baker says there hasn't been a true semiconductor capacity cycle since the late nineties due to TSMC smoothing supply.
capacitysemiconductors“We haven't seen a true capacity cycle in semis arguably since the late nineties. And that's because Taiwan Semi has been so good at aggregating and smoothing supply.”
-
Baker warns a true DRAM capacity cycle could see prices increase 10X, not the typical 30-50%.
drammemory“If the price of DRAM, if like a DRAM wafer is valued at like a five carat diamond.”
-
Baker argues SaaS companies are repeating brick-and-mortar retailers' mistake by rejecting AI due to lower margins.
ai adoptionmargins“Well, I think that application SaaS companies are making the exact same mistake that brick and mortar retailers did with e commerce.”
-
Baker argues SaaS companies are repeating brick-and-mortar retailers' mistake by rejecting lower-margin AI business.
aimargins“I think that application SaaS companies are making the exact same mistake that brick and mortar retailers did with e commerce.”
-
Baker notes AI companies generate cash earlier than SaaS despite lower margins due to fewer employees.
ai economicsmargins“The crazy thing is because of those efficiency gains, they're generating cash way earlier than SaaS companies did historically, but they're generating cash earlier, not because they have high gross margins, but because they have very few human employees.”
-
Baker says AI companies generate cash earlier than SaaS despite lower margins because they have very few human employees.
ai economicsmargins“And it's just tragic to watch all of these companies. Like you want to have an agent, It's never going to succeed.”
-
All-In Podcast, 4 January 2025
2025 predictiongpu shortage“inference compute is gonna be the the kind of derivative winner of that. We're gonna run out of GPUs, accelerators, compute in 2025 the same way we did in '23.”
-
Baker argues the ultimate AI winner must be the lowest cost compute provider, which requires owning rather than renting compute.
ai economicscompute ownership“the ultimate AI winner will be the one with the lowest cost of infrastructure cost, the lowest cost of compute.”
-
All-In Podcast, 4 January 2025
ai productivityderegulation“My most contrary belief, so I think that America over the next at some point over the next four years will print at least one year of greater than 5% GDP growth, real GDP growth.”
-
Baker predicts frontier AI labs will stop releasing leading-edge models to prevent IP theft via knowledge distillation, citing DeepSeek.
ai labsdeepseek“I think you will see the Frontier Labs stop releasing their leading edge models to prevent knowledge distillation and their IP effectively being stolen.”
-
Baker reveals high bandwidth memory is a bigger cost component for NVIDIA than Taiwan Semi, with only two capable manufacturers.
hbmmemory“High bandwidth memory is a bigger part of NVIDIA's cogs on GPUs than Taiwan Semi is. And today, there's two companies that can make it, Hynix and Micron.”
-
Baker identifies three axes of AI scaling: pretraining, inference time compute, and now reasoning as the third multiplicative dimension.
ai scalinginference“And then we started scaling around inference time compute. And it's very clear that we have now added a third axis of scaling performance, and that is reasoning.”
-
Baker says tech CEOs aren't thinking about ROI because they believe they're racing to create a digital god.
ai capexroi“Mark Zuckerberg, Satya, and Sundar just told you in different ways, we are not even thinking about ROI.”
-
Baker claims tech founders believe creating AGI first is worth tens or hundreds of trillions.
ai valuationscaling laws“And if you create that first digital god, we can debate whether it's tens of trillions or hundreds of trillions in value, and we can debate whether or not that's ridiculous.”
-
Baker reports Larry Page has said internally he is willing to go bankrupt rather than lose the AI race.
ai racegoogle“Larry Page has evidently said internally at Google many times, I am willing to go bankrupt rather than lose this race.”
-
Baker says xAI built a 100,000 GPU cluster using Elon's physical engineering approach from SpaceX and Tesla.
elon muskgpu cluster“x AI decided they were gonna build a 100,000 GPU cluster. And I would say with Elon's unique physical engineering mind, which we've seen play out at SpaceX and Tesla, and now at Neuralink, where he figured out how to miniaturize everything along with really capable teams.”
-
Baker says xAI built a 100,000 GPU cluster using Elon's physical engineering approach from SpaceX and Tesla.
gpu clusterspacex“And I would say with Elon's unique physical engineering mind, which we've seen play out at SpaceX and Tesla, and now at Neuralink, where he figured out how to miniaturize everything”
-
Baker says xAI built a 100,000 GPU cluster with Elon rearchitecting the Memphis data center from first principles.
gpu clusterspacex“And I would say with Elon's unique physical engineering mind, which we've seen play out at SpaceX and Tesla, and now at Neuralink, where he figured out how to miniaturize everything along with really capable teams. I think he rearchitected the x AI data center from first principles in Memphis.”
-
Baker explains xAI achieved coherence in a 100,000 GPU cluster with Hopper without next-generation networking technologies.
gpu clusternvidia“And because of that, they're able to get enough density that they could effectively make a 100,000 GPU cluster coherent even with Hopper without next generation networking technologies from NVIDIA, Broadcom, and others.”
-
Baker cites Chase Coleman's stat that less than 1% of Internet market cap was founded in two years after Netscape.
ai applicationsinternet history“Chase Coleman, who runs Tiger, who's an exceptional investor, had a really interesting statistic, which was ChatGPT came out in 2022 and was to AI as NetScape Navigator, was to the Internet in 1994.”
-
Baker cites Chase Coleman's stat that less than 1% of Internet market cap was founded in two years after Netscape.
ai applicationsinternet history“Only less than 1% of current global Internet market cap was founded. Founded in the two years after Netscape Navigator came out.”
-
Baker explains MFU runs at 35-40%, measuring the percentage of theoretical compute actually used for training.
ai infrastructureefficiency“MFU, model flops utilization, and that generally runs around 35 to 40%. And that's literally the percentage of compute, theoretical compute flops that you're actually applying to trading.”
-
Baker explains higher MFU allows 25% faster time to market with same GPU and power spending.
ai infrastructureefficiency“You have the same amount of GPUs and the same amount of power presumably. You could choose between faster time to market.”
-
Baker explains that 50% MFU versus 40% MFU enables 25% faster time to market for AI models.
ai infrastructureefficiency“You could choose between faster time to market. If you run a 50% MFU and your competitor's running 40 for an equivalent amount of trading flops, you could be in market 25% faster.”
-
Baker predicts GPT-7 or GPT-8 will cost $500 billion to train, ending commodity status of models.
ai capexmodel costs“By the way, these models are commodities today, but I am suspicious once we get to scaling laws continue, GPT seven or eight literally cost $500,000,000,000 to trade.”
-
Baker proposes a new metric decomposing MFU that he developed the morning of the interview.
ai infrastructuremetrics“MFU is the most important metric because it gives you all of these advantages and ways to differentiate yourself amongst five people, six people who've trained these GPT core class models.”
-
Baker explains GPUs got 50x faster while data center infrastructure only got 4-5x faster, causing low utilization.
bottleneckdata center“GPUs have gotten 50 times faster, and the rest of the data center has only gotten four to five times faster. And that is why MFU is so low.”
-
Baker says GPUs got 50x faster while data centers only 4-5x faster, creating investment opportunity in infrastructure.
bottleneckgpu“GPUs have gotten 50 times faster, and the rest of the data center has only gotten four to five times faster.”
-
Baker says only three US locations can provide gigawatt nuclear power, charging 10x average electricity rates.
data centersnuclear“There's only basically three places in The United States today where you can get a gigawatt of power to a single data center that's reliable enough, I. E. That's nuclear.”
-
Baker argues global warming is solved because solar efficiency compounds at 10% yearly and batteries at 8%.
climateenergy“And it is solved because photovoltaic cell efficiency is compounding just under 10% a year, and battery efficiency is compounding 200 bps less than that.”
-
Baker argues return on invested capital has risen for AI spenders, dismissing ROI skepticism.
ai capexroi“These companies are all public, and there is something called return on invested capital. And ROIC has gone up for all of these companies since they ramped CapEx.”
-
Baker says Tesla FSD 12.5 is at GPT-3 level and can consistently drive him most places without interventions.
autonomous drivingfsd“I think 12 dot five is like GPT three, and it can consistently drive me most places with no interventions. I'm a seasonal driver. I really only drive my Tesla in the summer.”
-
Baker says FSD 12.3 contained all the progress of the previous ten years in one release.
fsdprogress“This year, it's like when we turned on 12 dot three, it's like all of the progress over the last ten years was in that one release from the first time I had that Tesla.”
-
Baker predicts Tesla FSD will achieve 100x improvement quickly as compute scales to GPT-4.5 levels.
fsdinference“I think they're going to go really fast to GPT-4.5 compute, which means you're going to get, using these orders of magnitude, you're going get a 100x improvement really fast.”
-
Baker predicts FSD skeptics will be humiliated and says Tesla's data is 100-10,000x bigger than Waymo's.
teslatraining data“We could argue whether it's a 100 x, a thousand x, 10,000 x, bigger than the second biggest trading dataset, which is Weibo.”
-
Baker quotes Elon saying only synthetic data and real world video scale infinitely, identifying synthetic video as key risk.
synthetic datatesla“The only two data sources that will scale infinitely are synthetic data and real world video. And I thought that was interesting.”
-
Baker estimates average tech growth stocks under $100B market cap are down 40-65% from all-time highs.
market correctiontech stocks“I would say probably the average tech or consumer oriented growth stock that's below a $100,000,000,000 in market cap is probably somewhere between 40 and 65 percent off its all time high.”
-
Baker says Nasdaq returns in 2021 were dominated by Google, Microsoft, Nvidia, and Tesla, requiring 60% concentration.
concentrationnasdaq“The market return, particularly for the Nasdaq, was really dominated by a few stocks, Google, Microsoft, Nvidia, Tesla.”
-
Baker notes Amazon spent more on CapEx in two years than in the preceding twenty years.
amazoncapex“It's not a comment on Amazon. It's just a comment on the supplier response. They have spent more money on CapEx in the last two years than they did in the preceding twenty years.”
-
Baker provides specific numbers showing Amazon's CapEx jumped from $62B over 20 years to $87B in just two years.
amazoncapex“From 1999 to 2019, they spent $62,000,000,000 on CapEx. They're going spend $87,000,000,000 in 2020 and 2021. That's crazy.”
-
Baker notes Taiwan Semi will spend more on CapEx in two years than in the preceding five years.
capexsemiconductors“Taiwan Semi, their 2022 CapEx is going to be many multiples of 2019. 2021 and 2022, they'll spend more than they did in the preceding five years. So there is a massive supply response coming.”
-
Baker says Taiwan Semi will spend more on CapEx in 2021-2022 than in the preceding five years.
capexsemiconductors“Taiwan Semi, their 2022 CapEx is going to be many multiples of 2019. 2021 and 2022, they'll spend more than they did in the preceding five years.”
-
Baker says goods spending is $500B above trend while services spending is $500B below trend.
consumer spendingcovid“If you trend out and look at real personal consumption expenditures, goods spending is roughly $500,000,000,000 above the pre COVID trend line. Services spending is 500,000,000,000 below the trend line.”
-
Baker argues semiconductor industry has consolidated into monopolies or duopolies in every subsector over fifteen years.
consolidationmarket structure“Let's step back and look at the last fifteen years of semis. The industry has completely consolidated to where you almost have these monopolies and almost monopolies or duopolies.”
-
Baker cites research showing AI algorithm quality doubles with every 10x increase in training data.
aidata“Microsoft wrote about this in a research paper ten years ago, or maybe not twelve years ago, the quality of a given AI algorithm doubles with every 10x increase in the amount of data you use to train that algorithm.”
-
Baker argues AI is now eating software, extending Andreessen's famous thesis.
aidisruption“Marc Andreessen wrote this op ed, whatever it was, ten years ago about how software is eating the world. Now AI is eating software.”
-
Baker identifies extreme consolidation in semiconductors with only three companies making leading edge logic.
manufacturingmonopoly“Today, Intel, Samsung, Taiwan, Semi, they the only companies in the world that can make leading edge logic. There's only three companies that can make leading edge DRAM, maybe four for NAND.”
-
Baker says Google was first cloud provider to stop competing with independent infrastructure software providers.
cloudgoogle“But they try and force the open source infrastructure build themselves. And I think Google was the first one to realize, like, Hey, this is kind of silly.”
-
Baker argues Google was first cloud provider to embrace rather than compete with third-party infrastructure software.
cloud computinggoogle“B, there's a lot of money to be made just in providing these primitives. Let's stop trying to compete with these independent infrastructure software providers right on top of the cloud.”
-
Baker argues AI has changed cybersecurity economics, enabling companies to exceed $20B valuations for the first time.
aicybersecurity“That's why for the first time ever, for a long time, I forget the exact numbers, but you never had a cybersecurity outcome over 20,000,000,000.”
-
Baker describes adopting a multistage venture strategy from mentor Antonio Gracias.
multistagestrategy“But his whole strategy a lot of his strategy for venture, which I ruthlessly copied, is around the idea of their profound advantages towards being multistage.”
-
Baker says the key trait for public equity investors is the ability to be rational when wrong.
decision-makinginvesting“If I could know one thing about a person to know whether or not they would be a good public equity investor, it would simply be, are you capable of being rational when you're wrong?”
-
Baker says great public equity investors are often wrong more than right, with best batting averages around 55%.
investingperformance“They're wrong more than they're right, but because they make more money when they're right, they still have a great track record.”
-
Baker says valuation spreads within sectors reached 4.5 standard deviations, now at 3.5 standard deviations.
market structurevaluation“Valuation spreads within a sector between individual stocks are at all time highs. It was a 4.5 standard deviation relative to the norm only a few days ago.”
-
Baker argues the most obvious trades work at the beginning of every bear market.
airlinesbear market“At the beginning of almost every bear market, doing the most obvious things works. So who crushed it for the first three to four weeks of this bear market?”
-
Baker says investors are running a new metric: days of solvency with zero revenue.
analysisliquidity“Days of liquidity with zero revenue has been the acid liquidity test that I think a lot of investors have been working through.”
-
Baker says investors are running a new metric: days of solvency with zero revenue.
analysiscoronavirus“Days of liquidity with zero revenue has been the acid liquidity test that I think a lot of investors have been working through. And then from a broader perspective, I think a couple of insights.”
-
Baker notes that highly leveraged names like Domino's were among the best performers after 2009.
2009leverage“Go back to what were the best stocks coming out of June 2009? Netflix, booking.com, Ulta Beauty, Domino's Pizza. Domino's was a name that was incredibly levered.”
-
Baker challenges the notion of truly recurring revenue, noting software was only 7-10% of IT spend in 2008-09.
recurring revenuerisk“A lot of people are baselining off what happened in 'eight, 'nine. And I think that is dangerous because in 'eight, 'nine software was seven to 10 of IT spend.”
-
Baker argues software at 30% of IT spend today versus 7-10% in 2008-09 means it will get cut.
it spendrevenue“And I think that is dangerous because in 'eight, 'nine software was seven to 10 of IT spend. And today it's roughly 30% excluding spending on the big cloud hyperscale infrastructures”
-
Baker notes software is now 30% of IT spend, or 50% including cloud, versus 7-10% in 2008-09.
cloudit spending“And today it's roughly 30% excluding spending on the big cloud hyperscale infrastructures and service companies like Azure, Amazon web services, and Google GCP.”
-
Baker references Robert Smith's claim that software is better than first lien debt.
debtsoftware“Robert Smith is the CEO of Vista Equity Partners, brilliant man. And he famously said software is better than first lien debt.”
-
Baker recounts how a dentist cut her software bill in half in one hour by threatening to switch vendors.
pricingsmall business“He said, tell Gavin exactly what you just told me. And his sister is a dentist and obviously she's grappling with how to run her practice.”
-
Baker reports a dentist cut her software bill in half in one hour with zero pushback.
pricing powerrecession“In one hour, she cut her software bill in half. One hour, she said there was zero pushback.”
-
Baker recounts a dentist cutting her software bill in half in one hour with zero pushback.
pricing powersmall business“In one hour, she cut her software bill in half. One hour, she said there was zero pushback. And I've been telling people this story.”
-
Baker reports small businesses are cutting software bills 30-50% in an hour by simply asking vendors.
pricing powersmall business“In the last two days, lots of friends have come to me and said, Hey, I told my brother, my sister, my mom, my dad, my uncle, my aunt, my friend who has a small business, that story, they did the same thing and they cut their software bill 30 to 50% in an hour.”
-
Baker observes Zoom became woven into social fabric in ways unimaginable two weeks earlier.
adoptionsocial“It is being woven into the social fabric of the country in a way that I don't think anyone could have imagined even two weeks ago.”
-
Baker says Apple, Google, and Facebook missed massive opportunity by not matching Zoom's UI and UX quality.
competitionui/ux“And I can only imagine what it feels like to be at whether it's Apple with FaceTime, Google with Hangouts, Facebook with whatever their video chat is.”
-
Baker argues that 50 basis points of UI/UX superiority can create hundreds of billions in market value.
competitive advantagetechnology“Would argue personally they're at parody, but 50 bps is huge. If you're 50 bps better, that can be the foundation of hundreds of billions of dollars in market value.”
-
Baker says Google found two milliseconds of latency reduction had huge impact on search volumes.
competitive advantagelatency“Google found that milliseconds of latency mattered to search. That if you could take two milliseconds out of search latency, that it had a huge impact on search volumes.”
-
Baker argues milliseconds of latency and small UI differences like Zoom's create hundreds of billions in value.
productui/ux“That if you could take two milliseconds out of search latency, that it had a huge impact on search volumes. And Zoom's UI and UX for all sorts of reasons is just better”
-
Baker argues Zoom's superior UI/UX versus Google, Apple, and Facebook products explains its dominance during coronavirus.
competitive advantageui/ux“And Zoom's UI and UX for all sorts of reasons is just better than Hangouts, FaceTime or whatever the comparable Facebook product is.”
-
Baker predicts a majority of people will spend most waking hours in the metaverse within his lifetime.
metaverseprediction“The metaverse is simply a series of connected virtual worlds that I firmly believe a majority of people will spend a majority of their waking hours in within my lifetime.”
-
Baker cites Telecom Italia data showing video game traffic up 75%, social media traffic flat.
social mediatraffic data“I think the last data I saw, this was a Telecom Italia stat, so we should not mix and match, but Telecom Italia saw video game traffic up 75% and social media traffic up 0%.”
-
Baker argues we have thirty to forty years of rapid disruption ahead, comparing the microchip to the steam engine.
carlota perezdisruption“I would argue her work and kind of history suggests that we have another thirty to forty years of rapid and powerful disruption ahead of us.”
-
Baker argues we may be entering peak disruption, seventy to one hundred years after the microchip's invention in 1959.
disruptionhistory“I would argue peak economic, political, social change as a result of that invention was seventy to one hundred years later during the peak of the industrial revolution eighteen twenty to eighteen fifty.”
-
Baker says growth investors must assume multiple compression and be paid on business outcomes, not valuation.
growth investingphilosophy“It's critical in growth investing to assume multiple compression. It doesn't always happen, but you want to be paid based on business outcomes and assume valuation is not part of the outcome.”
-
Baker recounts how Sunu Kalra's simple iPod vs Walkman comparison gave him conviction to hold Apple long-term.
appleinvesting“And as soon as I heard that, it's like, wow. It really doesn't matter. Apple went on to be seven x, 10 x, 15 x from there.”
-
Baker argues used iPhone resale value is critically important to Apple's business model and pricing power.
applecompetitive advantage“One is, I think the price to used iPhones is critically important to Apple's business model. Why is that? Used iPhones retain their value much better than used Android phones.”
-
Baker says used iPhone resale value enables Apple to charge consumers less effectively than Android despite higher sticker prices.
applecompetitive advantage“Used iPhones retain their value much better than used Android phones. When you trade in your phone, you capture that.”
-
Baker argues used iPhone resale value enables Apple to charge premium prices while maintaining lower effective consumer cost.
appleconsumer“When you trade in your phone, you capture that. That lets the used iPhones retain their value better, supports a dramatic price premium for Apple.”
-
Baker explains Apple's effective price is $600 versus Android's $850 due to trade-in value differences.
androidapple“That comparable Android phone is being traded in for probably a $150, which means you're paying $850.”
-
Baker provides specific numbers showing Apple's effective price is $600 versus Android's $850 after trade-ins.
applenumbers“So Apple can charge the consumer more than its competitors, but actually have a lower effective price to the consumer.”
-
Baker explains Intel's historical advantage of reaching semiconductor nodes first gave them 15-20% embedded advantage.
intelpower“Intel for the last, I don't know, fifty years, sixty years always got to the next node in semiconductor manufacturing first.”
-
Baker quantifies Intel's manufacturing lead as 15-20% advantage that Taiwan Semi overtook in 2018.
intelnvidia“And in 2018, and it's important that the Taiwan Semi seven nanometer node is equivalent to the Intel 10 nanometer node.”
-
Baker says Taiwan Semi reached the node before Intel in 2018, creating a competitive shift.
intelsemiconductors“Taiwan Semi got to the node before Intel, which meant that Taiwan Semi, who supplies most semiconductor companies that you can think of Yeah.”
-
Baker argues AI revolution stems from cloud computing power and mobile-generated data, not algorithmic advances.
ai capexinference“The only thing that has enabled the AI revolution that we're living through, which I think we're at the bottom of the first inning in, is one, we had the ability to do cloud computing, so just apply significantly more computational power to old algorithms, and then b, we had dramatically more data.”
-
Baker predicts autonomous vehicles will reverse urbanization trends by making long commutes tolerable through sleep and phone use.
autonomous vehiclesreal estate“So you've had this huge trend towards people moving back into the cities because they wanna be able to walk around. People don't like long commutes.”
-
Baker predicts most people will spend most waking hours in virtual worlds within current lifetimes.
metaverseprediction“I think it is inevitable that a majority of people will spend a majority of their waking hours in virtual worlds within our lifetimes.”
-
Baker argues measurable things like search may be overvalued compared to unmeasurable things like physical slotting fees.
measurementsearch“You could just measure search in a way you can't measure slotting fees. Anything you can measure generally gets, in some ways, may be overvalued in today's Internet world.”
-
Baker argues customer acquisition cost is replacing rent as the primary cost for many businesses.
business modelse-commerce“CAC is the new rent. Customer acquisition cost, online ad spend. It's not just replacing other forms of ad spend. For a lot of businesses, it is replacing rent. Storefronts.”
-
Baker provides specific monetization rates: mobile games earn 8-10 cents per hour versus 50 cents for console games.
gamingmonetization“From a monetization perspective, your average mobile phone game monetizes, let's call it, 8 to 10¢ an hour. If it's not average, that's a reasonably well monetized game.”
-
Baker states data quantity is the single most predictive element of AI quality, not algorithms or infrastructure.
ai capexinference“The single most predictive element of knowledge about AI quality is the quantity of data used to train the algorithm.”
-
Baker cites research showing AI quality doubles with every 10x increase in training data quantity.
aidata“All that matters is the quantity of data, and it's been very well established in multiple papers from both Google and Microsoft research that for every order of magnitude increase in the data you use to train an algorithm, the quality of the AI doubles.”
-
Baker cites research showing every 10x increase in training data doubles AI quality.
ai capexinference“and it's been very well established in multiple papers from both Google and Microsoft research that for every order of magnitude increase in the data you use to train an algorithm, the quality of the AI doubles.”
-
Baker cites research showing every 10x increase in training data doubles AI quality.
aidata“it's been very well established in multiple papers from both Google and Microsoft research that for every order of magnitude increase in the data you use to train an algorithm, the quality of the AI doubles.”
Posts 10 posts
top 10 by rank, newest first
-
post Baker notes that a 50% price cut in GPT 5.6 drove 14x volume increase, calling it wild.
https://x.com/GavinSBaker/status/2093759017638084628
-
post Baker says future is multi-model with specialized open-source models on customer data working with frontier models behind custom router
https://x.com/GavinSBaker/status/2091887341786833180
-
post Baker predicts every accelerator company will do some variant of 3D DRAM or zHBM discussed at HotChips
https://x.com/GavinSBaker/status/2091618684603105717
-
post Baker says Atreides internal AI spend will be 100x higher in August 2026 versus March 2026 and still doubling monthly
https://x.com/GavinSBaker/status/2091590204486451277
-
https://x.com/GavinSBaker/status/2091542026072338623
-
post Baker estimates Citadel shed 80% of 45 billion SALP portfolio levered 4x so unwound 180 billion gross with market impact
https://x.com/GavinSBaker/status/2090817261476983199
-
https://x.com/GavinSBaker/status/2090114970595864991
-
post Baker would not take under on 250 billion for Anthropic unless they stumble or fail to secure compute
https://x.com/GavinSBaker/status/2089415607162634501
-
post Baker estimates his personal AI usage is up 100x and calls Grok Bot another Claude Code moment
https://x.com/GavinSBaker/status/2089379355692527813
-
post Baker says Grok 4.6 matches Fable 5 Max at 85% discount, 80% cheaper input and 88% cheaper output, Pareto dominant
https://x.com/GavinSBaker/status/2087567239423676519
Reported 8 quotes
in print, highest ranked first
-
reported
Baker predicts Anthropic will price its IPO around two trillion dollars then trade upward to three to four trillion.
“ Baker said. His working prediction: price around $2 trillion, then let the stock trade upward to roughly $3 trillion, possibly toward $3–4 trillion. Polymarket traders, for what it's worth, were already pricing an 80 percent probability of an Anthropic IPO this year and 67 percent odds that the company holds the top model at year-end. The strategic reason this pricing matters extends far beyond one company. Sacks argues that Anthropic's quarterly disclosures will become the industry's single most important demand signal — the pace car for NVIDIA, TSMC, Micron, SK Hynix, and everyone else in the chain. ”
-
reported
Baker reports that Ilya Sutskever's Safe Superintelligence plans to launch its first model in August 2026.
“SITUATION DETECTED: Ilya Sutskever's Safe Superintelligence is reportedly planning to launch its highly anticipated first model in August 2026, per Gavin Baker in an interview with Patrick O'Shaughnessy.”
-
reported
Baker reports hearing from multiple sources that Dario Amodei believes Anthropic could become the only private company in the world, with only governments remaining.
“Internally, Anthropic is very confident. I have been told by multiple people I trust that Dario has said that Anthropic might be the only private company in the world at some point. Think about that. In this vision, an Anthropic maximalist vision, there's Anthropic, and then there are governments, and that's it. So there's a lot of confidence. I'm sure they have more advanced checkpoints than Fable up their sleeve. They've executed really, really well. I would probably take the under on them being the only private company in the world. Yeah, I think it's gonna be a long time before they land a rocket.”
-
reported
Open-source AI models have more than doubled their market share at Vercel from 28% to 62% over two months, even as OpenAI and Anthropic usage accelerated.
“More data than open-source AI is taking share from OpenAI and Anthropic. Open source has gone from 28% token share to 62% token share @vercel over the last 2 months. Chart from @rauchg Super impressive given that the sum of OpenAI and Anthropic accelerated in July. So net”
-
reported
Baker's firm will spend one hundred times more on AI in August 2026 than in March 2026, with spending doubling monthly.
“Suspect we will start to hear about a “Pareto optimal” balance of computationally efficient humans, cheaper open-source tokens and frontier tokens. Our internal AI spend @Atreidesmgmt will be roughly 100x higher in August 2026 vs. March 2026. Still roughly doubling every month.”
-
reported
Private jet resale values are elevated because buyers are harvesting engines from old aircraft to use as turbines for data center power generation.
“Residuals for private jets are extremely strong these days. And a big part of the reason is people are taking jet engines off old planes and repurposing them as turbines to power data centers. ”
-
reported
Private jet engine demand is being driven by data center operators repurposing them as power turbines.
“Residuals for private jets are extremely strong these days. And a big part of the reason is people are taking jet engines off old planes and repurposing them as turbines to power data centers. ”
-
reported
Open-source AI models have more than doubled their market share at Vercel from 28% to 62% over two months while OpenAI and Anthropic usage also grew.
“More data than open-source AI is taking share from OpenAI and Anthropic. Open source has gone from 28% token share to 62% token share @vercel over the last 2 months. Chart from @rauchg Super impressive given that the sum of OpenAI and Anthropic accelerated in July. So net”