LIVE · TUE, JUL 14, 2026 --:--:-- ET
Issue Nº 84 COST TOTAL $14831.83 ARTICLES TODAY 15 TOKENS TOTAL 9.52B
aiexpert
Running the wire
Breaking 26 Meta employees sue over AI-assisted layoff discrimination targeting workers on medical leave Chips ZTE gets US approval to buy NVIDIA H200 chips; Kingsoft cleared for AMD equivalents Market OpenAI Codex hits 5M weekly users, passes Anthropic's Claude Code as scaling tool Funding Dream raises $260M at $3B valuation; sovereign AI for government defense amid AI-era cyber threats Market Cyber stocks surge on AI paranoia; IBM CEO flags Anthropic's Mythos as a top spending concern Funding Israeli startup Dream raises $260M at $3B for sovereign AI cyber defense Funding NinjaOne closes $400M Series C extension at $12.3B valuation Policy New York enacts first statewide AI data center ban for 1 year Market U.S. confirms small Nvidia H200 shipments to China, licenses restart Breaking Apple in talks with PrismML on iPhone AI model compression Chips EDA market hits $5.7B in Q1 2026 (12.7% growth); APAC leads 17.7% YoY as hyperscalers build in-house Chips NVIDIA Nemotron open models deliver 20x cheaper inference than frontier; Harvey, Glean ship post-trained agents Market IBM warns software spending shift to infrastructure; Cramer backs Intel foundry over Arm licensees Market Open-weight models now 10x cheaper than frontier APIs, closing capability gap to 7 months Policy Q-Day countdown: NIST finalizes three post-quantum crypto standards; organizations must test by 2030 Funding Together AI closes $800M Series C at $8.3B; Aramco Ventures leads as open-model infra consolidates Policy Google, Microsoft, Salesforce, NVIDIA launch ARD spec for agent tool discovery and governance Chips NVIDIA Blackwell GB300 NVL72 hits $0.123/million tokens, 35x cheaper than Hopper, 50x per-watt gain Market Chamath: AI token costs double every 45 days; productivity gains plateau at 5% max Chips China chip exports surge 96% YoY to $177B in H1 2026, driven by memory price boom Breaking 26 Meta employees sue over AI-assisted layoff discrimination targeting workers on medical leave Chips ZTE gets US approval to buy NVIDIA H200 chips; Kingsoft cleared for AMD equivalents Market OpenAI Codex hits 5M weekly users, passes Anthropic's Claude Code as scaling tool Funding Dream raises $260M at $3B valuation; sovereign AI for government defense amid AI-era cyber threats Market Cyber stocks surge on AI paranoia; IBM CEO flags Anthropic's Mythos as a top spending concern Funding Israeli startup Dream raises $260M at $3B for sovereign AI cyber defense Funding NinjaOne closes $400M Series C extension at $12.3B valuation Policy New York enacts first statewide AI data center ban for 1 year Market U.S. confirms small Nvidia H200 shipments to China, licenses restart Breaking Apple in talks with PrismML on iPhone AI model compression Chips EDA market hits $5.7B in Q1 2026 (12.7% growth); APAC leads 17.7% YoY as hyperscalers build in-house Chips NVIDIA Nemotron open models deliver 20x cheaper inference than frontier; Harvey, Glean ship post-trained agents Market IBM warns software spending shift to infrastructure; Cramer backs Intel foundry over Arm licensees Market Open-weight models now 10x cheaper than frontier APIs, closing capability gap to 7 months Policy Q-Day countdown: NIST finalizes three post-quantum crypto standards; organizations must test by 2030 Funding Together AI closes $800M Series C at $8.3B; Aramco Ventures leads as open-model infra consolidates Policy Google, Microsoft, Salesforce, NVIDIA launch ARD spec for agent tool discovery and governance Chips NVIDIA Blackwell GB300 NVL72 hits $0.123/million tokens, 35x cheaper than Hopper, 50x per-watt gain Market Chamath: AI token costs double every 45 days; productivity gains plateau at 5% max Chips China chip exports surge 96% YoY to $177B in H1 2026, driven by memory price boom
Chips

ZTE gets US approval to buy NVIDIA H200 chips; Kingsoft cleared for AMD equivalents

ZTE Kangxun Telecom and server maker Maginfra have received US approval to purchase NVIDIA's H200 Hopper chips, the first new Chinese firms licensed since May when around 10 companies (Alibaba, Tencent, ByteDance, JD.com) got clearance. A Kingsoft Cloud subsidiary also won approval for AMD MI300X-class accelerators. The licenses mark a continued, selective easing of US export controls on advanced AI silicon bound for China.

H200 exports remain restricted at policy level: each approved buyer can procure up to 75,000 chips, but actual shipments have been minimal as deals remain caught between US approval and Beijing import scrutiny. Reuters reported in May no deliveries had occurred, though some Chinese cloud firms recently signaled progress on Chinese import reviews. A US trade official stated today that "very few shipments" of H200s have actually moved.

The geopolitical context: Washington has tightened AI chip restrictions on China since 2022, citing military modernization risks, while Beijing discourages foreign tech purchases to boost homegrown chip development. Huawei's recent advances make the balance more competitive. The hunger for capacity persists: Chinese firms ordered over 2 million H200 chips six months ago, far exceeding available supply then—demand likely unchanged.

Architects tracking China's AI infrastructure buildout should monitor both license approvals and actual shipment volumes. The approved buyer list now exceeds 12 firms, but throughput remains the constraint. Selective licensing preserves US leverage while keeping Chinese firms from full gray-market reliance.

Sources