AirTrunk committed $30B+ to build 5GW of AI data center capacity in India by 2030 — among the largest data center commitments in Asia, positioning India as a major AI compute destination.
Snapshot — June 5, 2026
39 stories
- Anthropic experienced elevated error rates across multiple Claude models (Opus 4.6/4.7/4.8, Sonnet 4.6) for approximately 3 hours on June 5, attributed to infrastructure issues.
- In a cascading impact, Notion temporarily disabled Anthropic model access after performance disruptions.
- The incidents highlight that enterprise SaaS companies building on third-party AI models face availability risks traditional stacks don't — a single upstream model issue can cascade across an entire product's AI features.
- Reuters reported that Anthropic—which had been blacklisted by the White House—has quietly eased tensions with the administration ahead of its IPO.
- The rapprochement is significant: Anthropic needs regulatory goodwill for a smooth public listing, while the White House needs cooperation from frontier labs on its voluntary safety framework.
Are you in or are you out?
Berkeley RDI *(No new Berkeley RDI emails found for 2026-06-05)*
Business Insider *(No new Business Insider emails found for 2026-06-05)*
CIO Dive - [2026-06-05] [EXTERNAL] June 5 - Agent-led software development | How to scale AI
*Coverage from newsletter subscriptions for 2026-06-05*
DealBook (Andrew Ross Sorkin / NYT) - [2026-06-05] [EXTERNAL] DealBook: A.I.s nonproliferation moment?
DeepSeek confirmed V4 was trained on Huawei AI chips, after earlier inference success on the same hardware. The milestone weakens the assumption that U.S. export controls will durably constrain Chinese AI development.
Google released quantization-aware-training (QAT) versions of Gemma 4 across five sizes (E2B, E4B, 12B, 26B A4B, 31B), preserving quality while sharply reducing memory needs. Together with this week's Gemma 4 12B launch, they push capable multimodal models onto phones, laptops, and consumer GPUs — extending local inference and lowering cloud dependency.
Google will pay SpaceX $920M/month for AI compute at xAI data centers, totaling over $30B. The deal validates xAI's buildout as commercially viable and entangles Google's compute with Musk's ecosystem ahead of SpaceX's record IPO.
Japan's digital minister warned the country risks becoming an "AI colony" if it falls behind in domestic AI. The statement signals Japan may accelerate public investment in AI infrastructure — echoing EU concerns and the global trend toward AI sovereignty.
Meta is considering raising tens of billions through equity to fund ~$145B in AI investments. Follows Alphabet's $85B raise — the AI arms race is forcing public equity dilution to finance compute buildouts.
- Mustafa Suleyman told VentureBeat a contractual change ~6 months ago gave Microsoft formal authority to pursue superintelligence independently.
- Seven MAI models operate without OpenAI.
- He called Anthropic "too expensive," positioning Microsoft's models as cost-competitive.
- The timing — weeks ahead of OpenAI's expected IPO — is deliberate.
- Microsoft's AI chief stated the company has been "set free" from OpenAI to pursue its own frontier AI ambitions, including building superintelligence independently.
- In separate remarks, he called Anthropic's AI services "too expensive," positioning Microsoft's in-house MAI models as a cost-competitive alternative.
Microsoft published research identifying seven new ways AI agents can be hacked — novel prompt injection variants, tool-use exploits, and multi-agent coordination attacks. A concrete threat taxonomy for enterprises deploying agents alongside Anthropic's 31.5% browser-agent hijack rate data.
MIT's third annual SERC symposium convened MIT, Google DeepMind, and OpenAI researchers. Panelists argued the central alignment question is increasingly governance — who sets the values AI systems encode — rather than capability alone.
MIT used a Battleship-style task to show that improving question-planning lets a small model jump from rarely beating humans to winning most games — at ~1% of cost. Better agent design, not just bigger models, is a path to capability. ________________________________ RESEARCH
- York passed a one-year moratorium on large data center construction — the first such ban in the U.S.
- It reflects growing tension between AI infrastructure demand and community concerns about energy, water, and displacement.
- Permitting risk is now a material factor in site selection. ________________________________ SECURITYHOT
- Anthropic has deployed ~6 engineers to the NSA to help the agency use Mythos, its frontier cybersecurity model.
- Whether Mythos is in active offensive operations remains unclear.
- The reporting is notable given both a DoD designation and a prior federal ban on the agency's use of Anthropic technology.
- Nvidia authorized an $80 billion share repurchase — its largest ever — and raised its dividend, finishing the week as the only "Magnificent 7" name to close higher.
- The move follows Q1 revenue of $81.6B (up 85% YoY), with data-center revenue alone at $75.2B.
- The program signals management views the stock as undervalued relative to its earnings trajectory.
Nvidia's Nemotron 3 Ultra — a 550B-parameter MoE (~55B active) with a 1M-token context window — reached general availability on Hugging Face, OpenRouter, and NVIDIA NIM with open checkpoints and published training recipes. It posts the highest Artificial Analysis Intelligence Index for a U.S. open-weights model and runs 3–6× faster than comparable Chinese open models, though Moonshot's Kimi K2.6 still leads overall.
OpenAI and Anthropic share approximately 90 investors, creating an unusual dynamic where the same capital base funds both sides of the frontier AI race. The overlap raises questions about conflicts, information walls, and portfolio construction as both approach public listings.
- Pfizer signed a license agreement with Chai Discovery to accelerate drug discovery using AI.
- The deal pairs Pfizer's pharmaceutical development pipeline with Chai's AI-driven molecular design capabilities.
- It's the latest in a series of frontier-pharma AI partnerships (alongside OpenAI's GPT-Rosalind) that signal big pharma is moving from AI experimentation to committed commercial integration.
- SpaceX's $75B offering — 555.55M shares at $135 — is oversubscribed.
- Morgan Stanley projects $3.4T in revenue by 2040.
- The deal folds in xAI from February's merger and would be the largest IPO on record, a bellwether for the AI-driven listing pipeline.
Songs generated by Google's Gemini app (via DeepMind's Lyria model) included spoken watermarks from stock-music preview tracks — suggesting unlicensed audio may have entered training data. The finding intensifies IP and provenance scrutiny of generative-audio systems.
- Supabase raised $500M at ~$10.5B, doubling for the third time in two years.
- CEO credits Claude Code and Codex for driving database launches up 600%+ YoY — 60%+ now initiated by an AI tool.
- The company serves ~10M developers as the default backend for Bolt, Figma, Lovable, and Replit.
Data-center operator Switch is in talks with Brookfield, KKR, Goldman, and JPMorgan at $50B+. Among the year's largest private AI-infrastructure financings — capital rotating from AI apps toward the physical backbone.
Tencent appointed Yao Shunyu — a former OpenAI researcher — as AI chief with a mandate to build AGI using "OpenAI's playbook." The hire signals Tencent's shift from cautious integration to aggressive frontier research, part of China's broader talent acquisition pattern.
The Information *(No new The Information emails found for 2026-06-05)*
The Tactical Allocation Letter *(No new The Tactical Allocation Letter emails found for 2026-06-05)*
TechCrunch synthesized Uber's budget blowout, Altman's "huge issue" admission, and Copilot billing backlash into a systemic narrative: usage-based AI pricing is creating cost management challenges the industry hasn't solved.
This file catalogs email subjects received. Full article extraction requires individual email processing via merge_publications.py.*
- President Trump called for the military to accelerate its adoption of AI technology, with the Washington Post framing it as a directive to fast-track AI integration across defense operations.
- The announcement follows the Pentagon's ongoing effort to rewrite its AI procurement processes and aligns with the administration's broader push to position the U.S. as the global leader in AI-driven defense capabilities.
- Trump plans to meet AI companies next week on government profit-sharing or equity stakes.
- Senior officials are eyeing shares in OpenAI and Anthropic.
- Forbes framed it as "Universal Basic Capital" — bipartisan convergence around public ownership in companies built on publicly funded research.
Wall Street Journal / WSJ *(No new Wall Street Journal / WSJ emails found for 2026-06-05)*
WSJ Pro CyberSecurity *(No new WSJ Pro CyberSecurity emails found for 2026-06-05)*
WSJ Wealth Advisor *(No new WSJ Wealth Advisor emails found for 2026-06-05)*