Who is on this subjectordered by the date of their first quote here
2 of 4 lanes rest on fewer than 5 quotes and are marked thin. Offsets are days from the middle first-quote date, 21 Jan 2025 — a date, and nothing else. It is not a claim about who reached a view first.
-
Bill Gurley
2
thin
23 Dec 2024 – 23 Dec 2024
-29 d
- Dylan Patel 12 21 Jan 2025 – 9 Jul 2026 +0 d
-
Brad Gerstner
1
thin
28 Aug 2025 – 28 Aug 2025
+219 d
-
Gavin Baker
13
9 Dec 2025 – 4 Aug 2026
+322 d
The chronologysourced and dated, oldest first
-
-
“with Blackwell, not only is it way, way, way faster, anywhere from 10 to 15 times on really large models for inference because they've optimized it for very large language models.”
-
Bill Gurley
Patel says Blackwell delivers five times performance TCO improvement in single year, accelerating LLM cost decline.
“At least that's what Blackwell is, and we'll see what Ruben does. But, you know, five x plus in a single year for performance TCO is an insane pace.”
-
-
-
Dylan Patel
Patel estimates Blackwell delivers 10-15x cost improvement for inference despite NVIDIA claiming 30x at GTC.
“But now, like, Blackwell, NVIDIA's pitching 10 to 15 x improvement in cost. It's like, well, you know, they're massaging the numbers marketing.”
-
Brad Gerstner
Gerstner believes there's over 50% probability NVIDIA can sell Blackwell chips in China, a $50 billion market.
“China is a $50,000,000,000 market for NVIDIA unto itself. So I believe that the probability they're gonna be able to sell chips China is north of 50%.”
-
Dylan Patel
Patel says Blackwell cost is up 50-70% over Hopper but performance exceeds 2x, making it an obvious choice.
“while cost is only up, know, call it 50% or 60% or 70% versus Hopper, the performance is well north of 2x, right?”
-
Dylan Patel
Patel explains Blackwell introduces third memory tier within Tensor Cores requiring complex programming model for full performance.
“Now with Blackwell, there's actually even a third tier of memory, which is memory within the Tensor Core.”
-
Dylan Patel
Patel quantifies Hopper performance improved 30-40% over 2024 with Blackwell showing similar gains in months.
“Even Hopper in January 24 to December 2024, you still had like a 30%, 40% performance improvement. And likewise for Blackwell, you've seen a similar sort of improvement, right?”
-
Dylan Patel
Patel predicts another 50-100% Blackwell performance improvement from software optimization of standard libraries.
“I see another 50% to doubling in the cards, just from the fact that there is so much low hanging fruit or performance left on the table if you are using the general open libraries, right? Whether it's cuDNN, VLM, SGLANG, all these other things,”
-
Dylan Patel
Patel predicts general-purpose Blackwell users could see another performance doubling over the next year from software improvements.
“for the general purpose user, yeah, you're going to have another doubling of performance potentially on the tables over the next year.”
-
Dylan Patel
Patel reveals NVIDIA doubled system-level testing time for Blackwell compared to Hopper generation.
“NVIDIA learning from the issues on Hopper, with Blackwell, they actually doubled the time for system level test.”
-
Dylan Patel
Patel finds 10x performance TCO advantage for Blackwell over H100 at certain operating points.
“But at other points, you can have a 10x performance TCO benefit from going with a more expensive Blackwell server than a cheaper H100 server.”
-
“And the reason for that is it was by far the most complex product transition we've ever gone through in technology. Going from hopper to Blackwell, first you go from air cooled to liquid cooled.”
-
Gavin Baker
Baker details Blackwell's complexity: racks tripled in weight and quadrupled in power consumption versus Hopper.
“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.”
-
Gavin Baker
Baker explains Blackwell racks weigh 3,000 pounds and consume 130 kilowatts, equivalent to 130 American homes.
“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.”
-
Gavin Baker
Baker details Blackwell transition: racks triple in weight and quadruple in power from 30 to 130 kilowatts.
“Goes from round numbers, 30 kilowatts, which is 30 American homes to 130 kilowatts, which is 130 American homes.”
-
Gavin Baker
Baker argues reasoning saved AI by enabling progress during an 18-month gap when Blackwell delays would have halted everything.
“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.”
-
Gavin Baker
Baker argues reasoning models saved AI by bridging an eighteen-month gap while Blackwell was delayed.
“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.”
-
Gavin Baker
Baker explains Blackwell's complexity created deployment challenges that reasoning models had to bridge.
“Had reasoning not come along, there would have been no AI progress from mid twenty twenty four through essentially”
-
Gavin Baker
Baker argues reasoning models saved AI from an eighteen-month progress gap during Blackwell delays.
“Had reasoning not come along, there would have been no AI progress from mid twenty twenty four through essentially Gemini three.”
-
Gavin Baker
Baker predicts xAI will release the first Blackwell-trained model in early 2026 because Elon builds fastest.
“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.”
-
Gavin Baker
Baker predicts xAI will release the first Blackwell model because Elon builds data centers fastest.
“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.”
-
Gavin Baker
Baker predicts Blackwell models will be amazing because pre-training scaling laws remain intact.
“We know that scaling laws for pre training are intact, And this means the Blackwell models are going to be amazing.”
-
-
-
Dylan Patel
Patel says Hopper GPU failure rates improved to 5% while Blackwell remains at 10-15%, expecting higher rates for next generation.
“hopper's now at 5%, but black belt's still 10 to 15%. Wow. Right? Actually started out higher than that. Sure. And when a new generation comes out, it's gonna be higher than 15%.”
-
Dylan Patel
Patel says SemiAnalysis initially doubted Jensen's 25x Blackwell claim, predicting only 15-20x improvement.
“Jensen, when he originally launched Blackwell, had claimed it would be a 25x improvement. And at the time, no one believed him. Right? It's Jensen, right?”
-
Dylan Patel
Patel's benchmarking found Blackwell is 30x faster than Hopper on DeepSeek v3, exceeding Jensen's 25x claim.
“In DeepSeek v three, Blackwell is 30 x faster than Hopper on on somewhere on the continuum.”
-
Dylan Patel
Patel emailed Jensen documenting skepticism about Blackwell's 25x claim from industry observers.
“I emailed him. I like, hey, Jensen. You know, back in 2024, you said or back back when you launched Blackwell, said '24 twenty five x.”
-
Gavin Baker
Baker reports a top startup rented several thousand Blackwells at mid-$2 per GPU hour recently.
“And they had rented a cluster of several thousand black wells, And we'll just call it, you know, somewhere in the mid $2 per GPU hour.”
-
Gavin Baker
Baker notes the startup is renting an identical B200 cluster seven months later at nearly double the price.
“They're renting the exact same cluster, exact same size cluster, essentially identical in every way, b two hundreds, no no differences.”
-