The custom AI ASIC state of play (May 2026) — Broadcom deals, Google TPUs, Meta MTIA & beyond
…Each chip delivers 4,614 FP8 TFLOPS with 192 GB of HBM3E memory at 7.37 TB/s bandwidth. It’s manufactured on TSMC's N3P process in a dual-chiplet design…
Search
All sources
developer.nvidia.com
23
amd.com
8
apple.com
8
semiwiki.com
6
wccftech.com
6
nextplatform.com
5
spectrum.ieee.org
4
storagereview.com
4
xda-developers.com
4
theregister.com
3
press.asus.com
3
newsletter.semianalysis.com
3
Intel Xeon 6 With P-Cores Makes the Case For The Host CPU
…Still, at least for the time being, GPUs and their highly parallel processing capabilities will continue to be the go-to chips for training large language models (LLMs) and for running some…
AMD Ryzen Threadripper™ Processors for Desktop Workstations
…LLMs and diffusion models, powered by AMD Ryzen Threadripper processors. Whether you're designing a product or developing AI, Threadripper processors deliver the compute performance, memory capabilities, and platform flexibility to accelerate…
Optimized Software for Professionals With AMD and ISV Solutions
…AMD Radeon™ PRO Graphics AMD Radeon™ PRO workstation graphics help engineers advance product development workflows to accelerate time-to-market with AMD RDNA™ 3 architecture; the world’s first GPU with chiplet…
Discussions and forums
Inside NVIDIA Groq 3 LPX: The Low-Latency Inference Accelerator for the NVIDIA Vera Rubin Platform | NVIDIA Technical Blog
…Every tray integrates eight LPU accelerators, a host processor, and fabric expansion logic in a cableless design that simplifies rack-scale deployment and tightly couples compute with communication. LPU chip-to-chip…
Amazon’s AI Resurgence: AWS & Anthropic's Multi-Gigawatt Trainium Expansion
…in design decisions bodes well for future volumes. But they’re not giving up on TPUs and Nvidia GPUs either. Our Accelerator Model forecasts Amazon and Google Cloud’s chip purchases broken…
Intel Project Battlematrix With Arc Pro B60 GPUs Delivers 4x Better Perf/$ Versus NVIDIA In MLPerf v5.1 Benchmarks
…AI models, but required capabilities to deploy large language models (LLMs). These new Intel GPU Systems, code-named Project Battlematrix, are designed to meet the needs of modern AI inference and provide…
How to Build In-Vehicle AI Agents with NVIDIA: From Cloud to Car | NVIDIA Technical Blog
…AI LLM acceleration Built on DRIVE AGX, the automotive AI Box provides a modular AI compute solution that augments the limited inference capabilities of traditional infotainment systems on a chip (SoCs), enabling…
Amazon Goes Bigger on the M5 MacBook Pro Discount Than Apple Store (14.2-inch Liquid Retina XDR Display) - Kotaku
…There’s a Neural Accelerator built into each core to deliver blazing fast speeds and a mind-bending performance for all your tasks, especially demanding on-device AI workloads like LLM inference…
How Nations are Forging Their Own Digital Futures with Sovereign AI
…industry-specific model training services built on robust LLM frameworks like our Formosa Foundation Model, optimized for local language, values, and use cases, accelerating the deployment of Gen AI applications across diverse…