At Hot Chips 2026, Arm emerged as the foundational architecture for the next generation of CPUs, with Arm central to four of six first-day CPU papers, including IBM's future dual-architecture Z and Li…
OpenAI presented the architecture of Jalapeno, its general-purpose AI inference accelerator, at the Hot Chips 2026 conference, detailing a design built for end-to-end optimization rather than peak "sh…
XCENA and Samsung Electronics demonstrated MX1, a CXL memory expansion card with 3,072 RISC-V cores and up to 2TB of DDR5, at Hot Chips 2026 on August 25, targeting production in late 2026. The card, …
At Hot Chips 2026, semiconductor vendors presented divergent designs for AI inference, signaling an era of ferment where no dominant architecture has emerged, according to analyst Ben Bajarin. Compani…
Oxmiq Labs presented high-bandwidth flash (HBF) as a capacity tier for AI inference at Hot Chips 2026, claiming it delivers 8 to 16 times the capacity of HBM at the same cost. The company detailed HBF…
Intel Corp. unveiled a three-pronged hardware strategy at the Hot Chips 2026 conference, positioning agentic AI as a system-level workload requiring CPUs, GPUs, memory, interconnects, and advanced pac…
Samsung presented a 16 GB LPDDR5X-PIM memory package at Hot Chips 2026 that delivers 614 GB/s of internal bandwidth to its own compute units, an eightfold increase over the 76.8 GB/s available through…
At Hot Chips 2026, OpenAI presented Jalapeño, an inference ASIC built with Broadcom, which integrates heterogeneous compute, memory, and network resources on a single chip to handle the three phases o…
Arm unveiled its AGI CPU architecture at Hot Chips 2026, targeting AI agent workloads such as tool calling and workflow orchestration. Based on the Neoverse V3 core and manufactured on TSMC's 3nm proc…
Google announced its eighth-generation tensor processor units (TPUs), the TPU 8t for training and TPU 8i for inference, at Hot Chips 2026, marking the first time the company has released two TPU gener…
SambaNova Systems presented technical details of its fifth-generation SN50 reconfigurable dataflow unit (RDU) at Hot Chips 2026, claiming it delivers 5x the FLOPS of the SN40 and scales to 256+ chips.…
Microsoft unveiled the architecture of its second-generation Maia 200 AI accelerator at Hot Chips 2026, a 3nm chip with 140 billion transistors, 750W TDP, and 10,000 TFLOPS of FP4 performance, designe…
Meta Platforms Inc. presented its Meta Training and Inference Accelerators (MTIA) custom AI silicon at Hot Chips 2026, detailing a roadmap of four generations (MTIA 300, 400, 450, 500) to reduce relia…
NVIDIA presented its Spectrum-X Ethernet Multiplane Network Architecture at Hot Chips 2026, claiming 1.9x higher training performance in multi-tenant AI factories and detailing a roadmap to scale AI f…
XCENA, a semiconductor company developing memory-centric computing solutions for AI infrastructure, presented the architecture and measured performance of its MX1 computational CXL memory architecture…
Broadcom unveiled its Thor Ultra Ethernet NIC at Hot Chips 2026, an 800GbE adapter designed for AI and HPC fabrics, featuring a PCIe Gen6 x16 host interface, 8-port 100G MAC, 256 SR-IOV virtual functi…
XCENA presented its MX1 CXL computational memory device at Hot Chips 2026 in a joint session with Samsung, pairing the device with a rack-scale CXL memory plan. The MX1, built on Samsung Foundry's 4nm…
Samsung Electronics presented its LPDDR5X-PIM processing-in-memory solution at Hot Chips 2026, claiming it is the world's first LPDDR-based PIM product for AI inference. The chip delivers 614 GB/s PIM…
AMD presented its Versal Premium Gen2 adaptive SoC platform at Hot Chips 2026, featuring on-package LPDDR5X memory, PCIe Gen7 and CXL 4.0 support, and enhanced security including post-quantum cryptogr…
Intel Corporation unveiled its Crescent Island Xe3P inference accelerator at Hot Chips 2026, featuring up to 480 GB of LPDDR5X memory on a single 350 W air-cooled PCIe card, with Intel positioning the…