Nvidia (GPUs)

MacBook Neo涨价100美元至699美元，MacBook Air涨价200美元至1,299美元；苹果称AI数据中心需求与消费硬件争夺内存供应，致使成本以前所未有的速度攀升

这款推理处理器旨在降低每次查询成本，松动英伟达的垄断地位

2026财年第三季度6月24日公布，AI存储带动营收创历史新高，数据中心营收增长约150%，全年HBM产能已全部预订

晶圆级芯片挑战者单季营收达1.93亿美元，新增AWS合作，此前刚完成半导体史上最大IPO

英伟达获得LPU知识产权授权并挖走创始人后，Groq重新组建团队，押注新兴云计算容量

美国四座工厂已满负荷预订，亚利桑那第二工厂将3nm设备安装提前至2026年第三季度。AI需求持续超过供给，全年营收成长指引上调至30%以上

英伟达推出Quantum-X Photonics，2026年下半年将共封装光学引入以太网；光取代铜线，以3.5倍效率连接百万GPU集群

AGI CPU（136个Neoverse V3核心，台积电3nm制程）标志着Arm从纯IP授权转向直接出售芯片；开放的芯片小组件系统架构已吸引逾60家合作伙伴

A 560-page spec gives Ethernet InfiniBand-class RDMA and congestion control; the 2026 work item is Programmable Congestion Management

Seoul bilateral yields HBM4E samples and LP40 foundry talks; Tesla A16 to be built at Taylor, Texas; Samsung pins HBM5 comeback on an in-house 2nm base die

A draft registration lands a week after Anthropic's and days before SpaceX trades, the three could become the largest IPOs on record

Nighthawk validated in two independent studies; Quandela cuts QPU latency 5,000ms→30ms over NVQLink; QCi buys a packaging fab to commercialise photonic chips

Jensen Huang confirms Samsung, SK Hynix and Micron passed HBM4 qual on June 5; SK Hynix holds ~60-70% of Rubin volume, 12-layer stacks priced above $600

Helios yields 94 logical qubits from 98 physical; Quantinuum accelerates fault-tolerance to 2030 and signs Mitsubishi; IonQ publishes a 'walking cat' blueprint

Navitas joins Nvidia's MGX 800VDC push; SiC and GaN replace silicon to break the AI 'power wall', borrowing EV power architecture

After a year of bans and reversals, the US approved limited H200 sales, then Beijing told firms to pause and pushed Huawei instead

Up to $25bn with Amazon, a multi-gigawatt Google-TPU/Broadcom deal, and a monthly lease of xAI's Colossus 1 output

Linux Foundation ships OpenMDW-1.1 (Nvidia adopts for four model families); days later the G7 replaces the binary open/closed split with a spectrum

Data-center revenue ~$72.8bn; Vera Rubin ramps to full production by fall; guidance assumes zero China data-center compute, and the stock still slid

Native state-of-the-art computer use, 1M-token context and Codex on Nvidia infra push agentic workflows mainstream

NVLink licensed to third-party CPUs and accelerators (MediaTek, Marvell, Astera, Lightmatter); a $2B Marvell stake reads as a toll booth, not an investment

SoftBank, Nvidia and Amazon lead one of the largest private raises ever; APIs now clear 15bn tokens/minute

Arc Institute's 40B-parameter DNA foundation model predicts disease mutations and designs bacterial-scale genomes, the design-tool side of the biosecurity debate

Jensen Huang says orders are in and manufacturing is 'restarting'; BIS revised the license policy, but as of mid-2026 only a small number of H200s have actually shipped

Amazon ~$200bn, Microsoft ~$190bn, Alphabet $175–185bn, Meta $125–145bn, a combined AI buildout larger than most national budgets, with component inflation now a line item

Nvidia, Cisco, QIA and MGX back the Colossus compute build; xAI then leases Colossus 1 output to Anthropic

Nvidia invests in OpenAI; OpenAI commits to Oracle clouds; Oracle buys Nvidia chips, a self-referential money wheel as Oracle/Microsoft CDS spreads roughly double

Two Fairwater sites stitched by a dedicated AI wide-area network train one model in tandem, hundreds of thousands of Nvidia GB200/GB300 GPUs, 140kW racks, no UPS or gen-sets

First 1GW of Instinct MI450 deploys 2H 2026; the Helios rack targets ~3 AI exaflops with 72 MI455X accelerators, AMD's bid to break Nvidia's data-center grip

The 1.2GW Texas campus runs on Oracle Cloud with first GB200 racks delivered; Crusoe's $1.38bn Series E funds the expansion as SoftBank completes its $41bn into OpenAI