Top Stories
Cerebras Day 2: Stock Falls 10% After Biggest Tech IPO Since Uber. After raising $5.5B and surging 68% on its Nasdaq debut Wednesday (closing at $331, ~$95B market cap), CBRS slid 10% on its second trading day. The IPO was 20x oversubscribed with shares priced at $185 — well above the raised $150-160 range. It's the largest tech IPO since Uber in 2019. The bull case: 750MW OpenAI compute deal, $510M revenue, 47% net margin. The bear case: customer concentration, UAE investor scrutiny, and now the classic post-IPO pullback. Developing — priced May 14, now in volatile early trading. Watch whether it holds $300. (CNBC — "Cerebras stock falls after blockbuster IPO debut"; Motley Fool — "Cerebras Just Pulled Off the Biggest IPO Of 2026")
Trump-Xi Summit Ends: H200 Chips Approved but Stalled, AI Was the "Elephant in the Room." The two-day Beijing summit concluded with an uneasy truce. The concrete AI outcome: US Commerce Dept approved sales of Nvidia H200 chips to ~10 Chinese firms (Alibaba, Tencent, ByteDance, JD.com) at up to 75K chips each. The catch: no deliveries yet — Beijing is actively discouraging purchases to push domestic chip self-sufficiency. Trump extracted a novel 25% revenue-sharing deal (chips must transit US soil). No bilateral AI governance agreement was signed. Jensen Huang joined the delegation last-minute at Trump's invitation. Time's post-summit analysis: AI was conspicuously absent from the formal agenda despite being the obvious subtext. Developing — from Day 1 arrivals (May 13) through closing today. H200 delivery stall is the story to watch. (Time — "How A.I. Was the Elephant in the Room"; CNBC — "U.S. clears H200 chip sales to 10 China firms")
Musk v. Altman: Closing Arguments Done, Jury Deliberations Monday. The trial's courtroom phase is over. Closing arguments on May 14 saw Musk's attorneys frame OpenAI as a stolen charity; Altman's team argued Musk abandoned a nonprofit and can't claim it back. Advisory jury of nine convenes May 18. Judge Gonzalez Rogers delivers the binding final ruling — likely by end of May. Musk seeks $130B+ in damages, Altman/Brockman removal, and unwinding of the for-profit conversion. Stakes are existential for OpenAI's planned IPO. Expert consensus: partial liability finding likely, full unwinding unlikely. Developing — Week 3 complete. Deliberations begin Monday. (CNBC — "Closing arguments conclude in Musk v. Altman"; TechTimes — "OpenAI Jury To Begin Deliberations Monday")
Meta: 8,000 Layoffs Begin May 20, $135B AI Restructuring. Meta confirmed 10% workforce cuts starting next week — 8,000 employees plus 6,000 canceled reqs (14,000 total headcount impact). Reality Labs, Facebook, recruiting, and sales all hit. The driver: $115-135B in AI infrastructure spending this year, with teams reorganized into AI-focused "pods" under new Chief AI Officer Alexandr Wang's "Superintelligence Labs." Zuckerberg has now cut ~25,000 since 2022. Severance: 16 weeks base + 2 weeks/year of service + 18 months health coverage. Developing — extends the AI layoffs pattern from May 14 (Cloudflare 1,100, Cisco 4,000). Tech layoffs have topped 100K in 2026. (TheNextWeb — "Meta to cut 8,000 jobs on 20 May"; Variety — "Meta Laying Off 8,000 Employees")
Google I/O Monday: Gemini 4 and Leaked "Omni" Video Model. The keynote (May 19, 10am PT) is expected to headline Gemini 4 with 10M+ token context, native multimodal capabilities, and a major reasoning jump. Leaked ARC-AGI-2 scores show 84.6% — vs 68.8% for Opus 4.6. The wildcard: "Gemini Omni," spotted May 2 in production UI, a unified text/image/video generation model that would be the first top-tier omni-model with native video output. Also confirmed: Android XR glasses preview, Android 17 "Adaptive Everywhere" merging Android, Chrome OS, and XR. Developing — from pre-I/O leaks. Monday is the confirmation. (Android Authority — "What to Expect from Google I/O 2026"; WaveSpeed — "Google's Mysterious 'Omni' Video Model")
Model Updates
- GPT-5.6 in internal testing. "Ember Alpha" and "Beacon Alpha" checkpoints reportedly active. Mid-June release window. No new public info since yesterday. (WorldofAI — "GPT-5.6, Gemini 3.2, Robots")
- Gemini 4 reveal Monday. If leaked 84.6% ARC-AGI-2 score holds, it's a significant reasoning jump over current frontier. (BusinessToday — "Everything to expect at Google I/O 2026")
- Current frontier standings unchanged: Claude Opus 4.7 (coding/finance), GPT-5.5 (agentic/default), DeepSeek V4 Pro (open-source leader), Gemini 3.1 Ultra (multimodal).
Open Source
- Five frontier-class open models shipped in the last 30 days: Llama 4 (Scout/Maverick), Qwen 3.5, DeepSeek V4 (Pro/Flash), Gemma 4, Mistral Medium 3.5. MoE is now the default architecture at scale.
- DeepSeek V4 Pro posts 80.6% SWE-Bench Verified, 90.1% GPQA Diamond with 1M context — sitting alongside GPT-5.5 and Opus 4.7 on agentic benchmarks at a fraction of the cost.
- The open-weight frontier is now genuinely competitive with closed models for the first time. (CodersEra — "Best Open-Source LLM in May 2026")
Industry Moves
- Anthropic acquiring Stainless for $300M+. Would give Anthropic control of the SDK generation layer used by OpenAI, Google, Cloudflare, and Cerebras. Every
pip install openaiwould install a package built by an Anthropic subsidiary. Advanced talks, no term sheet yet. (The Information — "Anthropic in Talks to Buy Developer Tools Startup") - Cerebras $5.5B IPO — largest tech IPO since Uber. $95B peak market cap on day one.
- Novo Nordisk–OpenAI drug discovery partnership deploying across R&D, manufacturing, and commercial ops by EOY 2026. (CNBC — "Novo Nordisk partners with OpenAI")
- Tech layoffs top 100K in 2026. AI cited in 27,600 cuts (13% of all). Cloudflare (1,100), Cisco (4,000), Meta (8,000), Amazon (~30,000 in 5 months).
Safety & Policy
- No bilateral AI governance from Trump-Xi summit. The biggest missed opportunity of the week. AI compute access was negotiated, but governance was not.
- EU AI Act omnibus amendments from the May 7 trilogue moving toward formal adoption. Key: nudifier/CSAM ban, high-risk deadlines pushed to Dec 2027/Aug 2028, watermarking delayed to Dec 2026.
- US state AI laws proliferating: NY RAISE Act signed (effective Jan 2027), Colorado AI Act active, California AI Safety Act in effect since Jan 2026. Georgia chatbot safety bill signed, Hawaii deepfake bill advancing.
Worth Watching
- "How A.I. Was the Elephant in the Room at the Trump-Xi Summit" — Time's post-summit analysis of why the most consequential technology of our era was barely discussed formally. https://time.com/article/2026/05/15/trump-xi-us-china-summit-ai-semiconductor-chips/
- "Nobody Understands How Big This Is (Cerebras IPO)" — Pre-IPO deep dive on why this matters for the AI chip market.
- "Why some of the biggest names in tech are at an historic US-China summit" — CNN segment on the CEO delegation and what they're actually negotiating.