TechWatch

AI training vs. inferencing vs. memory bandwidth

Getting closer to very fast memory is the key.

Nvidia makes a dramatic, gigantic acquisition-licensing deal with Groq

As JPR predicted, the biggest private companies will get acquired.

Multiple threads improve GPU performance

How modern GPUs work.

Chinese sovereign VC fund 

Already a big investor and now with the promise of more.

Intel scales up functionality and out physically

The largest multi-chip floor to date.

RISC-V GPUs 

Will a RISC-V GPU be risky?

China’s GPU hopefuls hit the public markets with big multiples, while Nvidia hangs on

MetaX pops ~700% on debut, as China’s GPU trade builds, but Nvidia remains the buyer’s choice.

One more reason why Nvidia is impossible to catch

Nvidia acquires SchedMD, open-source AI workload management.

If AMD and Nvidia can sell chips to China, will it be profitable?

A 25% cost-of-goods hit comes out of the gross margin.

GF’s AI manufacturing push sets the stage for MIPS integration

Siemens and GF team up on AI-driven fab automation—and MIPS is the missing but implied piece.

The tensor contraction bet: Interview with Furiosa.ai CEO June Paik

Furiosa.ai and the architect behind it.

Qualcomm buys Ventana for RISC-V DNA, not a ready-made product line

Qualcomm acquires Ventana—and a lot of know-how.