The open source backend supports familiar PyTorch execution on Trainium2 and Trainium3, with compatibility and compilation limits that still matter for migration.
01 · Tech News
Follow the AI industry.
179 reports following 64 companies and partnerships from model release to physical deployment.
All Tech News
Browse every report, including the current lead.
Tag: Chips
12 articles · By source / event date · Page 1 of 1
NVIDIA’s hardware partners will release a 64GB unified-memory tier of the Grace Blackwell-based DGX Spark on October 23, aimed at smaller local AI inference, agent workflows, and multi-unit clustering.
Amazon Web Services has widened regional availability for its latest Arm-based compute families, adding local access points across the Asia-Pacific region, Europe, and AWS GovCloud.
MediaTek will use NVIDIA’s NVLink Fusion platform to integrate customer-designed accelerators into rack-scale data centers, backed by a $3.5 billion convertible bond investment.
Apple introduced the 2 nm `M6` chip in the Mac mini and the multi-die `M5 Ultra` in the Mac Studio, scaling on-device memory bandwidth and capacity for local AI workloads.
Intel detailed three complementary silicon architectures at Hot Chips 2026, pairing upcoming server CPUs and inference GPUs with its already-launched Core Series 3 edge processors.
Amazon disclosed multi-year, multi-gigawatt Trainium capacity commitments from Anthropic and OpenAI alongside triple-digit growth in its custom silicon and cloud AI divisions.
AMD has launched its first factory-integrated AI rack system, combining 72 Instinct `MI455X` GPUs with 18 sixth-generation `EPYC` processors. AMD says MI455X GPUs are in production, while customer rack deployments are planned over late 2026 and 2027.
NVIDIA has started deploying its liquid-cooled Vera Rubin NVL72 racks across five infrastructure partners, within a broader platform spanning seven chips and five rack tray types.
Micron has started volume shipments of its 12-high, 36GB HBM4 memory stacks for Nvidia Vera Rubin platforms, pairing on-package accelerator bandwidth with separate high-capacity system memory and PCIe Gen6 storage tiers.
Meta has outlined plans for four custom accelerator generations developed with Broadcom, shifting internal hardware from recommendation training toward low-precision generative AI inference while sharing rack designs.
Meta has signed a multi-year infrastructure agreement with AMD to support its AI infrastructure with up to 6 gigawatts of AMD Instinct GPUs, aiming to start physical deliveries in the second half of 2026.