Now Playing
OpenAI's Jalapeño vs Nvidia: HBM4 Wars & MATCH Act | Ep. 1
Audio is available on Spreaker — see link below.
OpenAI's Jalapeño inference chip claims 2x efficiency over Nvidia's GB200 — but HBM4 memory allocation may decide who wins. From TSMC foundry dynamics to the MATCH Act threatening ASML-China ties, today's briefing covers the sharpest fault lines in AI hardware.
Audio is available on Spreaker — see link below.
OpenAI just put a chip on the table. Literally.
Here's the constraint that ties almost every story in today's briefing together. Samsung is allocating over fifty percent of its Pyeongtaek foundry capacity to HBM4 base die production.
AMD looks interesting in this environment. Raymond James upgraded the stock to Strong Buy, raising its price target to six hundred forty-one dollars from five sixty-five, and projects a forty-four percent five-year compound annual growth rate in the server CPU market, reaching two hundred and one billion dollars by twenty-thirty.
A less obvious story that deserves attention. IBM demonstrated native dual-execution of z/Architecture and ARM at its Hot Chips session, enabled by a two-nanometre mainframe design.
On the policy front, the MATCH Act is gaining real traction. The bipartisan bill is on track for inclusion in the US defense budget and would effectively ban ASML DUV sales and servicing to China.
One more development worth logging. Japan launched Shunkai, a neutral-atom quantum computer built by IMS, Hitachi, and Infleqtion, targeting ten thousand qubits and fault-tolerance by March twenty-thirty-one.
To close, here's where to keep your attention. Watch for independent validation of Jalapeño's benchmarks against Vera Rubin.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.