AI News: June 6, 2026

1. Google to Pay SpaceX $920 Million a Month for AI Compute

SpaceX. Google agreed to pay SpaceX about $920 million per month from October 2026 through June 2029 for access to roughly 110,000 NVIDIA GPUs plus CPUs and memory, a deal worth around $29.4 billion over 32 months. Google framed it as short-term bridge capacity to meet surging demand for its Gemini Enterprise agent platform, with an exit clause allowing termination on 90 days notice after the end of 2026. The arrangement is roughly half the size of SpaceX’s concurrent compute deal with Anthropic and signals SpaceX emerging as a major compute supplier to frontier labs. Source

2. AirTrunk Commits $30 Billion to Build 5GW of AI Data Centers in India

AirTrunk. Blackstone-backed Australian data center operator AirTrunk said it will invest $30 billion by 2030 to develop 5 gigawatts of new capacity in India, one of the largest commitments to the country’s digital infrastructure. The plan includes a 3GW facility in Maharashtra’s Raigad Pen Growth Center alongside a 600MW pipeline across Mumbai, Chennai, and Hyderabad, announced after AirTrunk CEO Robin Khuda met Prime Minister Narendra Modi. The move underscores how AI demand is driving hyperscale buildouts into emerging markets backed by government tax incentives. Source

3. Florida Sues OpenAI and Sam Altman, Treating ChatGPT as a Defective Product

State of Florida. Florida became the first US state to sue OpenAI and CEO Sam Altman personally, filing an 83-page complaint that frames ChatGPT as both a defective product and a public nuisance. The suit alleges inadequate age verification, harm to minors, data collection before terms of service agreement, and claims that Altman cut short GPT-4o safety testing while allocating only 1 to 2% of compute to safety instead of a promised 20%. Attorney General James Uthmeier is seeking penalties potentially in the billions, marking a notable escalation in state-level AI legal action. Source

4. Flourish Raises $500 Million for Brain-Inspired AI, Backed by Jeff Bezos

Flourish. Neuroscience-focused AI startup Flourish raised $500 million at a $2.5 billion valuation, with roughly a fifth of the capital coming from Jeff Bezos and the rest from Alphabet’s GV, Lux Capital, and Catalio. The company is building Cortex AI, a connectomics-based system that maps real neurons and their connections to develop models that match the human brain’s roughly 20-watt energy budget rather than the far higher draw of current AI chips. Founded by former Amazon executive Rob Williams and neuroscientist Thomas Reardon, the raise reflects investor interest in radically more energy-efficient AI architectures. Source

5. Supabase Raises $500 Million at a $10.5 Billion Valuation

Supabase. Open-source database platform Supabase closed a $500 million Series F at a $10.5 billion post-money valuation, roughly doubling its value in about eight months, in a round led by GIC with Accel, Y Combinator, Coatue, Stripe, and new investor Salesforce Ventures. The company credits explosive growth to AI coding and agentic application development built on its Postgres platform. The raise illustrates how AI-driven app building is pulling fresh capital into backend infrastructure providers. Source

6. Generalist AI Raises $400 Million for Robot Foundation Models

Generalist AI. Robotics startup Generalist AI raised $400 million at a roughly $2 billion valuation in a round led by Radical Ventures, with participation from 8VC, Union Square Ventures, NVIDIA’s NVentures, Bezos Expeditions, and angels including Fei-Fei Li and Naval Ravikant. The company is building foundation models for what it calls physical AGI and says its GEN-1 model demonstrated 99% reliability across dexterous manipulation tasks. The capital will fund expanded robot-learning models, physical data collection, and commercial deployments. Source

7. Airbnb’s Brian Chesky Plans a New AI Lab

Airbnb. Airbnb CEO Brian Chesky is launching a new AI lab focused on user interaction and design, positioning himself among Silicon Valley leaders dissatisfied with current frontier models. Chesky will serve as founding chair rather than lead the lab directly, and the company has not yet named who will run it. The move adds to a wave of consumer-facing companies building their own AI capabilities after finding existing products not yet ready for their needs. Source

8. Microsoft Trained Its MAI Models on Unlicensed Web Data, Report Finds

Microsoft. A technical paper revealed that Microsoft trained its MAI models partly on unlicensed web data from Common Crawl and other public sources, contradicting earlier claims the models used only “enterprise grade, clean and commercially licensed data.” Security researcher Simon Willison flagged the discrepancy between Microsoft’s marketing and its documentation, which describes a pipeline of “publicly available and licensed human-generated data.” The case highlights the persistent gap between AI developers’ public commitments on data sourcing and their actual training practices amid contested fair-use claims. Source

9. The Token Bill Comes Due as Companies Scramble to Control AI Costs

Linux Foundation. A report on runaway AI spending found enterprises blowing through budgets as token consumption explodes, with one company reportedly accumulating a $500 million model bill after failing to set usage limits and heavy users spending 10 times more tokens for roughly double the productivity. The Linux Foundation launched a Tokenomics Foundation to establish common definitions and metrics, while vendors such as Pay-i, Paid, Ramp, and Datadog race to add AI spend management. With Goldman Sachs projecting token usage to grow 24-fold by 2030, the industry’s focus is shifting from speed to cost guardrails. Source

10. Apple Approves Poke as the First AI Agent on Messages for Business

Poke. Apple approved Poke, an AI assistant from Palo Alto startup The Interaction Company of California, as the first AI agent on its Messages for Business platform, letting iMessage users handle tasks like calendar management, smart home control, and photo editing by text. Poke pays Apple on a per-user basis, establishing a new revenue model where the platform profits while AI startups absorb distribution costs. The months-long approval process is likely to create a high barrier for competitors seeking similar access. Source