MemryX’s planned MX3+ AI ASIC targets more than 1 POPS and 2.16 GB of on-chip memory across 16 chips, with samples due late 2027 and production targeted for mid-2028.
The FT reports Jensen Huang is weighing a return as NVIDIA CEO after resigning. Allies seek support while Colette Kress remains interim chief and a board investigation continues.
Michigan researchers used PRISM, a polynomial input layer for robot-control networks, to raise simulated humanoid walking survival from 51% to 92.5% and real SO-101 tape-pickup success from 68% to 94%.
Q-CTRL raised $290 million in a Series C led by Salesforce Ventures to expand quantum error-control software and its Ironstone Opal GPS-free navigation system.
Samsung and Google open $395-and-up intelligent-eyewear preorders with Gemini, 12 MP cameras, nine-hour battery claims and new privacy safeguards ahead of December 10 sales.
PNNL has begun construction of M2PC, a 30,800-square-foot microbial science facility planning over 100 automated instruments and open researcher access in 2029.
Epic’s November 2026 Unreal Engine 5.9 MegaJam Preview expands AI editing tools and experimental WebAssembly/WebGPU game exports, but carries serious preview limitations.
KDDI, Ciena’s Blue Planet and Accenture plan broader AI-driven network automation after a six-month lab trial, with mobile expansion targeted for 2028.
Eight tech companies form the SGLang Consortium to coordinate open AI inference. Intel–RadixArk optimizations and AWS Trainium preview show how the ecosystem is expanding.
Former Apple CPU architect Gerard Williams III says AI helped NUVACORE reduce early CPU modeling from 7–12 months to 2–3, while emphasizing efficient CPUs beside GPUs in AI data centers.
MIT researchers report a 39% improvement in a simulated structure’s linear buckling load using an AI-guided, human-in-the-loop topology optimization system, with 15 seconds of overhead.