AI news4
Google agrees to pay SpaceX $920 million a month for AI compute
Google will pay SpaceX around $920 million a month from October 2026 through June 2029 to rent roughly 110,000 Nvidia GPUs, drawing on the Colossus data centre that xAI, now part of SpaceX, originally built. The deal mirrors one Anthropic struck with SpaceX in late May worth $1.25 billion a month. SpaceX announced both just a week before its shares are due to start trading on the Nasdaq.
techcrunch.com ↗Why it matters
SpaceX is quietly turning surplus AI compute into a major revenue line, and rivals are now renting capacity from each other to keep up.
AirTrunk commits $30 billion to build 5GW of AI data centres in India
Blackstone-backed operator AirTrunk will invest $30 billion by 2030 to develop 5 gigawatts of data centre capacity in India, one of the largest digital-infrastructure commitments the country has seen. The state of Maharashtra has signed a letter of intent for land for a 3GW site worth around $21 billion. CEO Robin Khuda announced the plan after meeting Prime Minister Narendra Modi.
techcrunch.com ↗Why it matters
The AI build-out is moving into the Global South at scale, with India positioning itself as a cloud and AI hub.
US lawmakers float a national AI law that would override state rules
Representatives Jay Obernolte and Lori Trahan released a 269-page bipartisan discussion draft, the Great American AI Act, that would preempt state laws regulating AI development for three years. It would require frontier developers above $500 million in revenue to publish safety frameworks, report serious incidents and submit to twice-yearly third-party audits. The preemption has drawn opposition from labour and consumer groups, and it lands weeks before Colorado's first-in-the-nation AI law takes effect on 30 June.
fedscoop.com ↗Why it matters
A federal framework could wipe out the patchwork of state AI rules just as the first big one goes live, reshaping how AI is governed in the US.
Nvidia unveils the RTX Spark superchip to put AI agents on Windows PCs
At Computex Nvidia showed the RTX Spark, a Windows-on-Arm superchip pairing a 20-core Arm CPU with a Blackwell GPU, delivering 1 petaflop of AI compute and up to 128GB of unified memory. Nvidia is working with Adobe to rebuild Photoshop and Premiere as GPU-accelerated apps for the chip. Systems are due in the second half of 2026.
nvidianews.nvidia.com ↗Why it matters
Nvidia is pushing frontier-scale AI onto personal devices, setting up local agents as the next battleground for the PC.
AI in the nonprofit sector1
Canada launches "AI for All" with a national push on AI literacy
Prime Minister Mark Carney launched Canada's national AI strategy, "AI for All," on 4 June, anchored by a National AI Literacy Initiative offering free, practical AI training. The program aims to reach 1 million entry-level post-secondary students and train more than 3,000 educators with ready-made AI learning kits. The wider strategy targets $200 billion in economic growth over five years and lifting AI adoption from 12% to 60% by 2034, with Carney stressing that the benefits "must be shared widely."
pm.gc.ca ↗Why it matters
A G7 government is treating AI literacy as public infrastructure, a model nonprofits and educators can point to when arguing for accessible AI training.