ai accelerators
30 articles about ai accelerators in AI news
UALink 2.0 Spec Finalized, Aims to Challenge NVLink for AI Clusters
The UALink 2.0 interconnect specification has been finalized, providing a standardized way to link AI accelerators from AMD, Intel, and others. However, it lags behind NVIDIA's established NVLink technology in real-world deployment.
AMD Backs UALink Open Interconnect to Challenge NVIDIA NVLink in AI
AMD is supporting the newly formed UALink Consortium, which aims to create an open standard for connecting AI accelerators. This move challenges NVIDIA's control over the critical NVLink technology that underpins its AI data center systems.
Intel & Google Announce Multiyear AI & Cloud Infrastructure Partnership
Intel and Google have announced a multiyear strategic collaboration to advance AI and cloud infrastructure, focusing on optimizing Google Cloud for Intel's Xeon processors, Gaudi AI accelerators, and future chips.
Silicon Photonics Breakthrough Enters Mass Production, Paving Way for Next-Generation AI Infrastructure
STMicroelectronics has begun mass production of its PIC100 silicon photonics platform, enabling 800G and 1.6T data rates critical for AI data centers. This breakthrough technology replaces copper with light for faster, more efficient data transmission between AI accelerators.
China PLP System Passes Validation at 510×515mm Panel Size
CFMEE validated a 510×515mm PLP system for AI chip packaging, challenging TSMC and Intel. The system targets HBM and AI accelerators, with potential 30% cost savings.
Nvidia Invests $2B in Marvell to Expand NVLink Fusion Chip Partnership
Nvidia is investing $2 billion in Marvell Technology to deepen their partnership on NVLink Fusion, a chip-to-chip interconnect crucial for scaling AI training clusters. This strategic move aims to secure supply and accelerate development of high-bandwidth links between GPUs and custom AI accelerators.
Google TPU Humufish Drops TSMC CoWoS for Intel EMIB-T
Google's next TPU Humufish uses Intel EMIB-T packaging instead of TSMC CoWoS, breaking the industry default for AI accelerators.
Qualcomm Taps TSMC for 3nm/2nm Dragonfly C100 CPUs, AI300 Accelerators
TSMC to fab Qualcomm's Dragonfly C100 and AI300 chips on 3nm/2nm nodes. The move challenges NVIDIA in data center AI, but timelines and performance remain undisclosed.
China's Nuclear Revolution: How Particle Accelerators Could Power Civilization for a Millennium
Chinese scientists are developing an accelerator-driven subcritical reactor that burns nuclear waste as fuel, potentially providing clean energy for 1,000 years while solving radioactive waste problems. The megawatt-scale prototype aims for 2027 operation.
Musk Pitches Moon as AI Compute Site via Electromagnetic Launchers
Musk proposes Moon-based electromagnetic accelerators to build solar panels for AI compute, leveraging lunar materials and low gravity.
Google Virgo Fabric: 100K-Accelerator AI Network Cuts Latency
Google unveiled Virgo, a data center fabric for AI clusters of 100,000+ accelerators, using a flatter two-layer topology to reduce latency and improve bisection bandwidth for synchronized training workloads.
AI Coding Agents Get Smarter: How Documentation Files Cut Costs by 28%
New research reveals that adding AGENTS.md documentation files to repositories can reduce AI coding agent runtime by 28.64% and token usage by 16.58%. The files act as guardrails against inefficient processing rather than universal accelerators.
Orbital AI Data Centers: Compute's Next Frontier?
DCD floats orbital AI data centers as next compute frontier. Physics and economics hurdles—power, cooling, latency, launch costs—remain unsolved. No concrete plans announced.
OpenAI's Ultrafast Mode Hits 750 Tokens/s on GPT-5.6 Sol
OpenAI launched Ultrafast mode for GPT-5.6 Sol at 14x speed and 750 tokens/s, powered by Cerebras. Preview limited to select customers, targeting latency-sensitive enterprise workflows.
Musk: xAI to Hit 10GW Compute, $300-500B Revenue by 2027
Musk told SpaceX staff xAI will hit 10GW by 2027, projecting $300-500B revenue. The 7x expansion faces unprecedented supply chain and monetization challenges.
DeepSeek Builds Gigawatt-Scale AI Data Center in Inner Mongolia
DeepSeek is building a gigawatt-scale AI data center in Inner Mongolia, per Bloomberg. The project marks a strategic pivot from efficiency to raw compute scale.
TSMC's 1.4nm Fab Slips to Mid-2028 as AI Chip Demand Tightens
TSMC's 1.4nm fab is ahead of schedule, with mass production by mid-2028, driven by AI chip demand from Google and Nvidia.
Nvidia, SK Group Announce $500B AI Infrastructure Partnership
Nvidia and SK Group announced a $500B partnership for HBM4 memory supply and a 2 GW AI data center in South Korea, locking in SK Hynix as Nvidia's primary memory supplier through 2030.
KV Cache Offload Makes Storage the New AI Bottleneck
Storage, driven by KV cache offload and rising SSD costs, is now the primary AI bottleneck per Supermicro and SemiAnalysis.
China's AI ecosystem standardizes on MoE with wide expert parallelism
China's AI ecosystem standardizes on MoE with wide expert parallelism to survive on weaker NPUs. Hardware makers now design 'supernode' systems.
Moonshot AI's Kimi K3: 2.8T params, 1M token window, $3/M input
Moonshot AI released Kimi K3, a 2.8T-parameter mixture-of-experts model with 1M token context window and $3/M input pricing, claiming autonomous chip design and research capabilities.
Silicon Photonics Hits 300-mm Wafer Scale for AI Interconnects
Silicon photonics moves to 300-mm wafers for AI interconnects, cutting cost per Gbps by ~30% and addressing bandwidth bottlenecks in 100,000+ GPU clusters.
US Allows ZTE to Buy Nvidia H200 AI Chips, Joining Alibaba, Tencent
US authorized ZTE to buy Nvidia H200 AI chips, joining Alibaba, Tencent, ByteDance in accessing Hopper architecture under targeted export controls.
Dongfang Suanxin Claims 14nm HBM-Free Chip Beats H200 Bandwidth
China's Dongfang Suanxin claims a 14nm HBM-free AI chip beats Nvidia H200 memory bandwidth, challenging US export controls.
China's 14nm AI Chip Hits 520 TFLOPS Via Architecture, Not Shrink
China's 14nm AI chip claims 520 TFLOPS and 6.4TB/s bandwidth via software-defined and 3D near-memory architecture, bypassing advanced node restrictions.
Meta Iris AI Chip Production May Start September – Report
Meta could produce Iris AI chip in September 2026 for data center inference, per report. Reduces NVIDIA reliance.
PKU Chip Hits 2.12ms Brain Latency, 478x A100 Speedup
PKU chip achieves 2.12ms step latency with 478x speedup over Nvidia A100 for brain modeling using phase-change memristors.
Scale-Across: Cloud Giants Link Datacenters for Million-Accelerator AI Clusters
Cloud providers are linking multiple datacenters for million-accelerator AI clusters, a new 'scale-across' paradigm.
Amazon Designs Custom AI Silicon for Future Devices, Panay Says
Amazon hardware chief Panos Panay confirmed Amazon is designing its own end-to-end silicon for some devices, signaling a strategic push into custom AI hardware.
Upscale AI Raises $500M for AI-Native Networking Silicon
Upscale AI raised $500M for AI networking silicon, with Google Cloud as a strategic partner. The deal targets GPU cluster interconnect bottlenecks.