Reports
AI-generated structured vendor updates
Moonshot AI Launches Kimi K3: 2.8T Parameter Open-Source MoE Model at $3/$15
Moonshot AI unveils Kimi K3, a 2.8T parameter open-source MoE model with 896 experts (16 active), native vision, and 1M context. API pricing at $3/$15 input/output per million tokens undercuts rivals. Open weights release July 27. GPU capacity exhausted within 2 days. Arena score 1679 tops Fable 5.
HPE Expands Private Cloud AI with NVIDIA Vera Rubin, Enabling Agent-Native AI Factory
HPE expands its Private Cloud AI line with NVIDIA Vera Rubin NVL72 and HGX Rubin NVL8, introducing Compute XD700, Cray GX240 blade with Vera CPU, and Quantum-X800 InfiniBand. New software includes Agent Toolkit and NemoClaw for agent-native AI, with Alletra Storage MP X10000 in Q4 2026.
NVIDIA Agent Toolkit Shifts AI Agent Control from Cloud to Local DGX Station
NVIDIA launches Agent Toolkit for DGX Station, comprising NemoClaw, Nemotron 3 Ultra, Omniverse Libraries, and OpenShell. It enables local AI agent deployment in 30 minutes, locking developers into NVIDIA's hardware-software stack and shifting control from cloud services to on-premises hardware.
NVIDIA Vera Rubin Platform and Dynamo 1.0 Disaggregate Inference, Shift Focus to Intelligence per Dollar
NVIDIA unveils Vera Rubin platform with a 7-chip stack (Vera CPU, Rubin GPU, NVLink 6, etc.) and Dynamo 1.0 inference disaggregation. A single NVL72 rack packs 72 GPUs/36 CPUs with 1.6 PB/s bandwidth, achieving up to 7x inference performance. The new 'intelligence per dollar' metric signals a shift from training to inference cost competition.
AMD and HPE Launch Helios Open AI Infrastructure to Rival NVIDIA Ecosystem
AMD and HPE expand partnership to launch Helios, an open-stack AI infrastructure platform integrating EPYC CPUs, Instinct MI455X GPUs, Pensando networking, and ROCm software. Each rack delivers up to 2.9 exaFLOPS FP4, built on OCP principles with Juniper switches, targeting simplified deployment and energy efficiency.
TSMC Pledges $100B More for 6 US Fabs, Localizing 3nm for AI Chip Supply Chain
TSMC announces an additional $100B investment in Arizona, bringing total US commitment to $265B, with plans for 6 fabs focused on 3nm and beyond. This move localizes advanced process for AI chip demand from NVIDIA, Apple, AMD, reshaping global semiconductor supply chain. Q2 net profit surged 77% YoY, FY capex raised to $60-64B.
Qualcomm Snapdragon X Control Panel Centralizes GPU and NPU Driver Updates, Tightening Ecosystem Grip
Qualcomm updates the Snapdragon X Control Panel to version 2026.2.0, unifying GPU and NPU driver updates and improving game library navigation. While enhancing user experience for Arm PCs, it strengthens Qualcomm's control over driver distribution and system optimization, paving the way for deeper ecosystem lock-in.
AMD与OpenAI达成6GW算力供应历史性协议 1.6亿认股权证可获10%股权 股价盘前涨35%
...
NVIDIA Debuts T3000/T2000 Modules and Cosmos 3 Edge, Builds Sovereign AI Ecosystem in Japan
NVIDIA unveils T3000/T2000 compute modules (Thor architecture) and Cosmos 3 Edge world model, signs Japan Noetra alliance for 13,750 Vera CPUs + 27,500 Rubin GPUs (140MW). Sovereign AI revenue triples to $30B+ in FY2026, accelerating the physical AI ecosystem.
NVIDIA-Nokia Alliance Redefines RAN Ecosystem with GPU-Based AI Acceleration
NVIDIA and Nokia are jointly developing AI-powered RAN technology, using NVIDIA GPUs to accelerate baseband processing and AI algorithms for beamforming and spectrum optimization. Targeting commercial deployment by 2027 and 2x spectral efficiency by 2028, this partnership marks a fundamental shift from dedicated RAN hardware to GPU-based, software-defined AI networks.
测试情报-NVIDIA AI chip news
...
NVIDIA Jetson Thor T3000/T2000: Blackwell GPU Crashes Edge AI Cost Barrier
NVIDIA unveils Jetson Thor T3000 and T2000 modules. The T3000 packs a Blackwell GPU and 8-core Neoverse CPU, delivering 865 FP4 TFLOPS at half the power of the T5000. New Jetson Agent Skills automate memory optimization, aiming to scale deployment of humanoid robots and edge AI.
NVIDIA CUDA 13.3 Introduces clmad for Hardware-Accelerated Carryless Multiplication on GPUs
NVIDIA CUDA 13.3 adds the clmad hardware instruction for carryless multiply-accumulate on Ampere+ GPUs. GHASH throughput reaches 6.3 TB/s on B200, up to 18.8x faster than bitsliced. Sum-check protocol accelerates 3-13x. The instruction also benefits CRC, Reed-Solomon, and post-quantum cryptography.
Intel 18A Yield Hits 85%, Secures Orders from NVIDIA, OpenAI, Reshaping Foundry Landscape
Intel reports 18A process yield improvement to 85%, from 65% last quarter, nearing TSMC N2's 90%. Secured foundry deals with NVIDIA, AMD, OpenAI, etc. EMIB advanced packaging yield reaches 98%, used in NVIDIA Feynman, Google TPU. This marks a strategic inflection in AI chip manufacturing.
Cisco, G42, AMD Deploy 1GW AI Cluster in UAE, Pushing GPU Diversification and Full-Stack Integration
Cisco, G42, and AMD partner to deploy a large-scale AI cluster in the UAE based on AMD MI350X GPUs, integrating Cisco's full-stack secure AI infrastructure (UCS servers, Nexus 9K switches, Firepower firewalls). This marks Cisco's transformation into a full-stack AI infrastructure integrator and positions AMD as a second GPU supplier for US-allied nations, locking out Chinese vendors in the UAE market.
AMD Confirms Zen 6 EPYC Venice: First 2nm Server CPU Launching July 2026
AMD confirms Zen 6 EPYC Venice launch at Advancing AI 2026 (July 22-23). As the first 2nm server CPU, it features triple-core hybrid architecture, up to 192 cores, ~29% single-thread and ~22% multi-thread gains, targeting AI inference and tight CPU-GPU synergy via Infinity Fabric.
Meta Iris Chip to Mass Produce in September: 6-Month Cadence Threatens NVIDIA GPU Hegemony
Reuters confirms Meta's Iris AI chip mass production in September, targeting 2.5GW by end-2026 and 14GW by 2027. Meta's 6-month MTIA generation cadence directly challenges NVIDIA's annual GPU cycle, signaling a hyperscaler shift from GPU dependency to custom ASIC sovereignty.
TSMC CoWoS Capacity to Reach 200k Wafers by 2027, Diversifying from GPU to CPU and ASIC
TSMC targets 200k wpm CoWoS capacity by 2027, narrowing supply-demand gap from 20% to 10%. Customer base diversifies from NVIDIA GPU to include AI server CPUs (MediaTek, AMD) and ASICs (Broadcom). CoPoS panel-level packaging enters pilot production in 2027.
WhiteFiber and DriveNets Achieve 111.2 Tbps Cross-DC AI Fabric, Breaking Power Constraints
WhiteFiber announces Project Redwood, partnering with DriveNets Ethernet AI fabric (FSE, VOQ, deep buffers), WEKA storage, and NVIDIA H200 GPUs, achieving 111.2 Tbps bandwidth and 0.9ms latency over 83km dark fiber, treating two geographically separated GPU clusters as a single logical supercluster. Commercialization planned for Q3 2026.
Meta Invests $9.17B in Canada AI Data Center, Iris AI Chip Mass Production Begins MTIA Roadmap
Meta announced a $9.17B AI data center in Canada with 1GW capacity, and its first in-house AI chip Iris will mass produce in September, kicking off the MTIA four-generation roadmap. Meta targets 14GW compute by 2027, using 6-month chip iterations to challenge NVIDIA's annual cadence and reduce GPU dependency.