youtube.com/watch?v=k_bH...
US experts admit losing the #AIrace to the less advanced but radically cheaper China chips and #opensource approach. This is accelerating due to #CoralNPU released by #Google, Meta+Google cooperating on #TorchTPU, #RVA23 and #RVV1 #RISC-V, China's rising #HBM3 capacity.
I hope you understand the world-shaking implications of #RVA23 chips and going off #Microsoft, #Oracle and #Amazon to build sovereign stacks with #PostgresQL, #NoSQL, #Linux, #TorchTPU, #openLLM.
Both #Google and #Apple could be made to drastically expand in Canada as well or offer sovereign stack.
Realizing this, Google & Meta steered clear of the disaster, as had Amazon previously by making it's own chips.
The Google #CoralNPU release and the collaboration on #TorchTPU bypasses all #nVidia / #Palantir #spyware #killswitch to preserve global appeal of #GoogleCloud.
nVidia bought Groq
3/3
Re "localization, openness, repair, and maintenance" this to me suggests at least:
#RVA23 #CoralNPU #TorchTPU
#Linux #PostgreSQL
#ConvergedEthernet #PowerOverEthernet
#GPMI-B as #torus_interconnect
#source_code_escrow
#blind_credential access to #capital_asset
web.archive.org/web/20041220...
NVIDIA’S GROQ DEAL WINS WALL STREET PRAISE FOR “OFFENSE AND DEFENSE”
#NVDA #Groq #NVDAStock #NVDANews #NvidiaStock #NvidiaNews #NvidiaGroq #PyTorch #TorchTPU #AIInference #Semiconductors #Chips #ArtificialIntelligence #TPU #LPUs
Nvidia has signed a non-exclusive licensing agreement for Groq’s inference technology, underscoring a joint push to expand access to high-performance, low-cost AI inference
#NVDA #NVDAStock #NVDANews #NvidiaStock #NvidiaNews #Groq #NvidiaGroq #PyTorch #TorchTPU #AIInference #Semiconductors
$NVDA
#RVA23 single binary for many #RISC-V #SoC riscv-europe.org/summit/2025/...
Expect every board more than #gg1 (about C$200) to be RVA23 in 2026, with most as low as gg0.25 also, everything else #RVA22 and only end sensors #RVA20.
A #datacentre uses these and soon #GPMI-B and #TorchTPU and #CUDA-Q.
Need good explanation of #tile_based and #CUDA-Q and how they relate to #ZLUDA, #ROCm, #TorchTPU and other alt ML APIs.
#TorchTPU is really bad news for #nVidia.
It sets up Google+Meta to challenge both #Huawei and the rising #RVA23 universe.
I expect China to mandate #GPMI-B as a #PoE upgrade & begin standardization of a #GPMI-A to be a universal connector for #IronwoodTPU #torusinterconnect & Ethernet switches.
