Governments moved from the sidelines to the center of the AI business this week: Washington lifted its export ban on Anthropic’s most powerful models, and OpenAI reportedly offered the federal government an ownership stake. Underneath the politics, the build-out kept accelerating — Meta opened a cloud business to sell its spare compute, global startup funding hit a fresh record, and the fight over the energy and data that fuel AI spilled into utility boardrooms, Canadian streets, and the courts.
1. OpenAI offers Washington a 5% stake as the US lifts its AI export ban
The US Commerce Department lifted its export ban on Anthropic’s Fable 5 and Mythos 5 models — with Anthropic tying the return to a new safety classifier — while OpenAI reportedly floated giving the federal government a 5% ownership stake in the company. For business leaders, access to America’s most capable AI is now entangled with Washington, which may end up part-owner of the very companies they buy from.
2. Anthropic launches Claude Sonnet 5, a cheaper model built to run agents
Anthropic released Claude Sonnet 5, its most agentic mid-tier model yet, with stronger coding, lower pricing, and improved safety aimed at running AI agents at scale. For companies automating professional work, it lowers the cost of putting capable AI to work on everyday tasks.
3. Meta launches Meta Compute to sell its spare AI cloud capacity
Meta launched Meta Compute, a cloud business that rents out its massive spare AI data-center capacity — echoing a similar move by SpaceX — and the news pushed chip and AI-infrastructure stocks lower. It signals that even social-media giants now see selling compute as a core business, giving enterprises another major cloud option to weigh.
4. Global startup funding hits a record $510 billion, powered by AI
Global startups raised a record $510 billion in the first half of 2026 — already surpassing all of 2025 — with AI, energy, and biotech leading the way, according to Crunchbase. The figures show capital is still flooding into AI, keeping valuations and competition high for any business buying or building with it.
5. Microsoft starts a $2.5 billion firm to help companies adopt AI
Microsoft formed Microsoft Frontier, a $2.5 billion professional-services business to help organizations build and run their own AI applications. The move acknowledges that most companies still struggle to turn AI tools into working systems — and that the help itself is now a large market.
6. The AI data-center boom is fueling a $200 billion utility M&A wave
The scramble for AI data-center power is driving a roughly $200 billion wave of mergers and acquisitions among utilities, as tech firms race to lock up electricity for their build-outs. For business leaders, it is a reminder that energy — not just chips — is becoming the binding constraint on AI growth.
7. OpenAI acquires cloud startup Ona to give its Codex agents a home
OpenAI acquired Ona, a startup that builds secure cloud environments for AI agents, to let its Codex coding tool handle longer, more complex tasks. The deal shows the race has moved past chatbots to the infrastructure that lets AI agents work reliably for hours at a time.
8. Anthropic and Meta turn to Samsung for custom chips to cut AI costs
Anthropic and Meta are both reportedly in talks with Samsung to design custom AI chips — with Meta said to be near a $6.5 billion deal — as the industry works to cut costs and reduce its dependence on Nvidia. The push toward in-house silicon could eventually lower prices and widen supply for the companies that buy AI services.
9. Hundreds rally across Canada against the AI data-center buildout
Hundreds of people rallied in cities across Canada to oppose the rapid buildout of AI data centers, citing energy and water demands and local impact. The protests preview a growing tension every organization planning AI infrastructure will face: public resistance to where and how that compute gets built.
10. Microsoft signs publisher AI deals as US local newspapers sue
Microsoft signed a Copilot content deal with Australian publishers even as a group of US local newspapers filed suit over AI training, laying bare the split between media companies signing and those suing. For any business using generative AI, it underscores that the rules — and costs — around training data are still being fought out in court.
