When Jensen Huang Meets the ASIC Wave

Disclaimer: The content provided in this newsletter is for informational purposes only and does not constitute investment advice. We are not registered investment advisors, and nothing in this newsletter should be construed as a recommendation to buy or sell any securities. Always do your own research and consult with a licensed financial professional before making any investment decisions.

We previously wrote 5 ASIC articles, starting with a very solid ASIC research series consisting of three articles.

Then we wrote a detailed report introducing CRDO and CLS, which also analyzed in depth why 3D Torus Architecture is Critical to ASIC.

The final article discussed Anthropic's explosive growth and how it might affect AWS. In reality, Anthropic's help to Trainum is much greater than its help to AWS. We heard news in early month about additional orders for Trainum2 2025 Cowos, and in the past two days, we've heard about new orders for Trainum 3 for 2026. We will continue to verify the authenticity of this information over the next two days.

Over the past decade, the race for AI compute has felt like a hardware-turbocharged sequel to Fast & Furious: one moment GPUs are still in front, and the next moment a fleet of ASICs, TPUs, and NPUs roars past like customized supercars. As early as mid-2024 we began discussing ASICs, and at the start of 2025 we analyzed how parameter-server (PD) separation and Mixture-of-Experts (MoE) architectures would reshape Nvidia’s position. In short, model-architecture advances are redefining the demands placed on underlying compute.