The Minutes of Dylan Patel on economics https://minutesof.com/dylan-patel/on/economics/ Everything Dylan Patel has said on economics: 22 verbatim quotes between February 2023 and August 2026, each with a timestamp and a link to the recording… en Sun, 30 Aug 2026 16:22:17 +0000 Patel reports Anthropic now generates up to $50 million per megawatt, a 5x return on compute costs. https://minutesof.com/q/e06c3440-97f6-494c-88a7-77a755b4e54a/ https://minutesof.com/q/e06c3440-97f6-494c-88a7-77a755b4e54a/ “And and what that now enables them to do is, hey, if I spend $10 on inference capacity, actually generate $50 of revenue,” — Dwarkesh Patel Tue, 25 Aug 2026 15:57:53 +0000 economics Patel argues hyperscalers would pay 8% interest rates versus current 5-6% given AI compute returns. https://minutesof.com/q/470deb1c-05eb-4eaf-9d8c-4112dbe35a93/ https://minutesof.com/q/470deb1c-05eb-4eaf-9d8c-4112dbe35a93/ “Meta's raised at, like, 5% to 6%. I don't see why they wouldn't pay 8% Because they would happily pay 8% because the return from the compute that they're going to build is humongous.” — Dwarkesh Patel Tue, 25 Aug 2026 15:57:53 +0000 economics Patel reports that tokens now represent 30% of his 90-person company's costs versus 70% for employees. https://minutesof.com/q/289f5368-a657-42fd-9826-8e45417ec2ab/ https://minutesof.com/q/289f5368-a657-42fd-9826-8e45417ec2ab/ “My my own company of 90 people, 30% of my cost now is tokens versus 70% employee costs.” — RAISE Summit Wed, 22 Jul 2026 16:43:49 +0000 economics Patel says OpenAI's total gross margin rose from 30% to 55% over the past year. https://minutesof.com/q/780c360e-dead-4817-88e5-9b6077a27b0e/ https://minutesof.com/q/780c360e-dead-4817-88e5-9b6077a27b0e/ “OpenAI late last year, their margins had were roughly 30% gross margin, but if you stripped away the free users, they were at 50%. Now, total company gross margin is closer to 55%,” — RAISE Summit Thu, 16 Jul 2026 16:44:57 +0000 economics Patel explains KV cache storage and reuse enables massive cost decreases in inference. https://minutesof.com/q/3bea5703-d869-4e93-bec6-70e3ecd20e14/ https://minutesof.com/q/3bea5703-d869-4e93-bec6-70e3ecd20e14/ “you calculate that once, you store it off in memory, whether it be system memory or storage, and then you pull it back in when you run the turn.” — RAISE Summit Thu, 16 Jul 2026 16:44:57 +0000 economics Patel states GPU rental rates were $12-13 billion per gigawatt before recent increases. https://minutesof.com/q/c1abee78-d176-457f-8a5b-90e353f41471/ https://minutesof.com/q/c1abee78-d176-457f-8a5b-90e353f41471/ “GPUs, at least before the craziness of the last six months, usually went around 12 to $13,000,000,000 per gigawatt.” — Sequoia Capital Tue, 30 Jun 2026 12:00:25 +0000 economics Patel cites estimates of 20% economic decline if Taiwan is invaded, comparable only to world wars. https://minutesof.com/q/3d3aa12b-148b-4e36-8a9c-ef47962c54bc/ https://minutesof.com/q/3d3aa12b-148b-4e36-8a9c-ef47962c54bc/ “And then subsequently, that that sends the world into a depression. Right? The world has not experienced a 20 free fall since maybe, like, a one of the great wars.” — Swole as a Service Wed, 13 May 2026 20:00:36 +0000 economics Patel observes perception shift where 90th percentile people consider themselves middle class while 50th perce https://minutesof.com/q/bd34934b-1867-41db-90e5-d28702f49580/ https://minutesof.com/q/bd34934b-1867-41db-90e5-d28702f49580/ “Nowadays, people in the ninetieth percentile think they're middle class, and people in the fiftieth percentile don't think they're middle class.” — Matthew Berman Mon, 09 Mar 2026 17:07:47 +0000 economics Patel says he now supports UBI despite being very capitalist, a shift from his prior views. https://minutesof.com/q/bec4f50b-ae61-4935-b577-c45c8527112c/ https://minutesof.com/q/bec4f50b-ae61-4935-b577-c45c8527112c/ “Over the last couple years I've realized, wait, I actually think UBI is perfectly fine. Which I think is crazy because again, I'm very capitalist.” — Matthew Berman Mon, 09 Mar 2026 17:07:47 +0000 economics Patel calculates $100B AI revenue requires $250B infrastructure spend at five-year depreciation, double curren https://minutesof.com/q/b206ab39-c11b-4186-80e2-7930fb9768d1/ https://minutesof.com/q/b206ab39-c11b-4186-80e2-7930fb9768d1/ “That $50,000,000,000 of COGS needs to burn on infra, which cost roughly with if a five if you're talking about five year depreciation, call it $250,000,000,000, right, of infra Yeah. For a $100,000,000,000 of revenue.” — The MAD Podcast with Matt Turck Thu, 05 Feb 2026 13:30:16 +0000 economics Patel explains users will pay 10x more for 10x faster inference completion, justifying Cerberus economics. https://minutesof.com/q/d63342d7-794d-43b2-8529-800ca943aa5a/ https://minutesof.com/q/d63342d7-794d-43b2-8529-800ca943aa5a/ “for a lot of people, I'm fine to spend 10x the price on something that completes 10x faster. So Cerberus sort of just makes a ton of sense there.” — TBPN Tue, 03 Feb 2026 23:08:03 +0000 economics Patel states AI infrastructure represents 25% to 75% of current US quarterly economic growth. https://minutesof.com/q/70800ca7-aaca-45bb-b128-a5925c5404e1/ https://minutesof.com/q/70800ca7-aaca-45bb-b128-a5925c5404e1/ “It's anywhere from 25% to 75 of the current quarter's economic growth. That's insane for The US.” — Chris Best Substack Fri, 31 Oct 2025 18:23:40 +0000 economics Patel says as the largest technology Substack, the 10% platform fee is worth paying for the growth. https://minutesof.com/q/46f7f4fc-5c60-4b8f-8293-dd4f74e0b5f6/ https://minutesof.com/q/46f7f4fc-5c60-4b8f-8293-dd4f74e0b5f6/ “Even though a 10% cut sounds like a lot, I'm the largest technology substack and I'm telling you that it's worth it.” — Chris Best Substack Fri, 31 Oct 2025 18:23:40 +0000 economics Patel states 80% of GPU data center cost is capital equipment, only 20% is land, power, and cooling. https://minutesof.com/q/f14c478d-6c68-4537-b652-a709482a769a/ https://minutesof.com/q/f14c478d-6c68-4537-b652-a709482a769a/ “It's the it's the physical data center conversion power conversion equipment. All of this stuff is, like, 80% of the cost. And then 20% is gonna be your land and your power and your cooling” — a16z Mon, 18 Aug 2025 18:26:43 +0000 economics Patel describes Chinese manufacturers as locked in constant cost reduction cycle rather than margin expansion. https://minutesof.com/q/017a5839-c71d-42bb-951b-8594f045c861/ https://minutesof.com/q/017a5839-c71d-42bb-951b-8594f045c861/ “there's this constant pressure cooker of continual cost reductions and engineering. It's not like, Oh, well, we make 80% margins. Our cost is low.” — Going Direct Sun, 22 Jun 2025 21:30:00 +0000 economics Patel argues that efficiency gains alone without capability increases cannot justify massive AI infrastructure https://minutesof.com/q/675ee285-ed22-4889-b6cf-5f6b6892d5c7/ https://minutesof.com/q/675ee285-ed22-4889-b6cf-5f6b6892d5c7/ “That would not pay for all of these build outs. Right? AI is useful today, but it's not capable of doing a lot of things.” — Alex Kantrowitz Wed, 23 Apr 2025 16:30:06 +0000 economics Patel breaks down full H100 GPU cost at $40,000 to $45,000 including networking and infrastructure versus $24, https://minutesof.com/q/837f0a3c-d3fb-454e-ba7f-3504bfd5a136/ https://minutesof.com/q/837f0a3c-d3fb-454e-ba7f-3504bfd5a136/ “each GPU, right, once you include networking, building, all this sort of stuff maybe is you know, the GPU itself of h 100 is, like, 24,000, but once you add everything else up, it's, like, forty, forty five thousand dollars per GPU all in of everything.” — Unsupervised Learning: With Jacob Effron Tue, 21 Jan 2025 14:00:12 +0000 economics Patel states OpenAI is paying $12.5 billion over five years for one-fifth of Oracle's data center site. https://minutesof.com/q/58cb4df9-acea-480b-8af4-0b91ab6d00bd/ https://minutesof.com/q/58cb4df9-acea-480b-8af4-0b91ab6d00bd/ “So for one fifth of the site, they're paying something like $12,500,000,000 over the next five years.” — Scaling Intelligence Tue, 12 Nov 2024 04:17:00 +0000 economics Patel argues AI demand is what makes continued semiconductor node advancement economically viable today. https://minutesof.com/q/c17dd34d-6330-45ef-ac94-0e2d4084112b/ https://minutesof.com/q/c17dd34d-6330-45ef-ac94-0e2d4084112b/ “funding the next node would not be economically viable anymore if it weren't for AI taking off. Right? And then generating all this humongous demand for the most leading edge chip.” — Dwarkesh Patel Wed, 02 Oct 2024 14:59:36 +0000 economics Patel argues training costs are irrelevant, noting GPT-4 used 20,000 A100s. https://minutesof.com/q/7037941a-ee80-4057-a5c4-ce393d0ea4d6/ https://minutesof.com/q/7037941a-ee80-4057-a5c4-ce393d0ea4d6/ “training costs are irrelevant. Right? Like, GPT four, right, like, 20,000 a one hundreds, that's that's like, I know it sounds like a lot of money.” — Latent Space Tue, 05 Dec 2023 18:16:28 +0000 economics Patel calculates doubling lithography cost only increases total chip cost 10 percent, enabling five nanometer. https://minutesof.com/q/a81c789d-69a2-44e4-83c3-7fc48f4f956d/ https://minutesof.com/q/a81c789d-69a2-44e4-83c3-7fc48f4f956d/ “Because again, the manufacturing cost of lithography is only one fourth to one fifth. So if you double lithography costs, you're only increasing the total cost of a chip by 10%, right?” — Hidden Forces Wed, 04 Oct 2023 14:00:36 +0000 economics Patel says memory now represents 30-50% of GPU manufacturing cost and NVIDIA applies 4x markup. https://minutesof.com/q/e22458f3-983f-450f-85ae-780001a92a79/ https://minutesof.com/q/e22458f3-983f-450f-85ae-780001a92a79/ “And and now today, if you look at it, the cost of memory is 30 to 50% of the GPU cost. Great cost. And then obviously, NVIDIA does a four x markup on their” — Gradient Flow Thu, 02 Feb 2023 12:00:39 +0000 economics