Part II: The long, windy trail of Intel GPUs
The evolution of Intel discrete graphics: A strategic pivot.
Is a paradigm change needed to realize AGI?
Architectures are evolving as fast as LLMs.
Bolt brings photorealism to tape
Offers new meaning to the term “tape-out.”
Taalas bets on hard‑wired models to beat GPUs at inference
Toronto start-up Taalas fixes trained LLMs directly in silicon.
SiFive raises again as CPUs return to the center of the AI data center story
$400 million gives SiFive space ahead of a probable IPO.
Fujitsu joins the AIP crowd with an NPU
Targeted at inference servers using advanced Japanese semi production.
AI processor architectures are evolving almost as fast as LLMs are
AGI will not be realized with just more clever software.
Arm stops hinting and starts shipping
Launches first production silicon CPU for AI data centers.
Supply, the growth limiter, not circular investment
Supply chain management is as important today as design.
Nvidia is (re)-proving the Jevons paradox
The more you buy, the more you save.
AMD and Samsung partnership on AI memory
Can’t be too rich, too thin, or have too much memory.
Nvidia’s Groq tie-in
Why Nvidia keeps making smart decisions.