If you want to do AI, you need an NPU, right?
Wrong. Just a processor that can do matrix math and doesn’t have to be dedicated.
Neuromorphic, the other AI processor
Neuromorphic chips advance toward low-power edge AI.
RISC-V becomes AI hardware’s open foundation
Three integration tracks move RISC-V beyond CPU control.
Mosaic-SoC sits at the intersection of AI and vision
Perception SoC enables real-time spatial intelligence at the edge.
Is a paradigm change needed to realize AGI?
Architectures are evolving as fast as LLMs.
Google TPUs are becoming a market product
This is a fundamental shift.
Bolt brings photorealism to tape
Offers new meaning to the term “tape-out.”
Computer houses may be the pinprick of the bubble
I wouldn’t want one in my backyard.
SambaNova and Intel target agentic inference
Heterogeneous system combines GPUs, RDUs, and Xeon 6 CPUs.
IBM’s latest AI processor is GPU-free
It does have 32 NPUs, however.
Because 138 are not enough
Anthropic is considering building its own AI chips.
AI Vector Vision reframes computer vision
Deterministic, vector-based representation.