The chip market will bifurcate beyond simple training-vs-inference into three segments: large-scale training, smaller distillation/reinforcement-learning training chips, and increasingly dedicated, power-constrained inference chips for endpoints that cannot run kilowatt-power GPUs.
“Yes. I and I also think you have a third bucket where training distills down to simpler training chips that you don't need to run a trillion parameter model... and then there's inference that over time will be very dedicated and particularly as you get to uh end points that you can't have a GPU that you know runs at at a kilowatt of power you just it's impossible.”
View on YouTube
Explanation
The chip market has continued bifurcating as predicted: large frontier-training chips (Nvidia's high-end GPUs), smaller distillation/fine-tuning chips, and a growing dedicated low-power inference/endpoint chip segment (edge AI accelerators from Arm, Qualcomm, and others) are all now distinct product categories.
The physical AI / robotics chip market (with robots containing tens to hundreds of chips each) will become a gigantic market, larger than the data center chip market.
“physical AI is going to be a gigantic market I mean today quite candidly bigger than data centers... physical AI, particularly AI that can learn, uh is I think going to be a giant market because the robots themselves will have tens of chips, hundreds of chips.”
View on YouTube
Explanation
No fixed timeframe was given; as of 2026 the data center AI chip market (Nvidia alone reporting roughly $300B/year in data center revenue) remains far larger than the nascent physical AI/robotics chip market, so the claim is not yet resolved either way.
Rene Haas
Right
00:20:57
techpoliticsgovernment
If the US broadly restricts semiconductor exports/licensing, countries with sufficient capability will build an independent parallel computing ecosystem outside the Western architecture, risking that alternative ecosystem becoming the ecosystem of choice in parts of the world.
“If you shut off supply of a computing architecture into other parts of the world, what what will happen? Certain parts of the world that have the capabilities either in terms of people, technology, uh innovation, they will find a way and they will find a way around around the problem. And once that happens, you've now created two parallel universes.”
View on YouTube
Explanation
A parallel Chinese computing ecosystem has emerged as predicted: Huawei's Ascend chips now hold over 50% of China's domestic AI chip market and DeepSeek and other labs have adapted their software stacks to run on non-Nvidia hardware in response to export controls.