Big news. AMD acquires Taalas.

This is the future: LLM-on-silicon, on every device makes total sense. You have your CPU, GPU and AIU. Local, fast, optimized for one model. If you want more power, or a newer model, use the cloud. Just like we do with CPUs/GPUs.

AMD acquires AI chip startup Taalas to boost inference performance by etching models into silicon

Early tech demos show model-specific integrated circuits churning out up to 17,000 tokens a second

If the title was "Meta developers accessed the internet and hacked another firm", the reactions would have been different. There would be some lawsuits at least.

Why not now?

We are in the mainframe era of AI: A few centralized compute providers and millions of "dumb" clients connected to them.