It was one of the most infrastructure-heavy weeks in AI history. Microsoft shipped its own coding model to replace OpenAI’s. Google raised $80 billion in a week. Anthropic signed a nine-figure monthly compute deal with SpaceX. DeepSeek emerged from stealth to raise capital at a $59 billion valuation. Here’s the AI news this week — everything that happened and why it matters.
Microsoft Cut the OpenAI Cord at Build 2026
The biggest story out of Microsoft Build this week wasn’t a feature update — it was a declaration of independence. Microsoft announced Project Polaris, its own in-house AI coding model that will replace GPT-4 Turbo as the default engine for all GitHub Copilot subscribers starting in August 2026.
Project Polaris uses a mixture-of-experts architecture with specialized sub-modules for individual programming languages, and it outperforms GPT-4 Turbo on HumanEval and MBPP benchmarks — with particular strength in low-resource languages like Rust and Haskell. It runs on Microsoft’s custom Maia 200 AI accelerators inside Azure, which reduces per-inference latency and cost compared to routing through OpenAI’s API.
Alongside Polaris, Microsoft shipped multi-agent VS Code support (a planner-and-specialist structure where an orchestrator decomposes tasks to subagents), Windows Agent Framework 1.0 as open-source, and Copilot Workspace going generally available. The timing is pointed: the Polaris announcement lands just months after Microsoft and OpenAI ended their seven-year exclusive partnership in April.
Why it matters: Microsoft is competing on model quality now, not just distribution. If Polaris delivers at scale, it changes the economics of enterprise coding AI significantly — and tightens the squeeze on competitors like GitHub Copilot alternatives and Cursor.
Alphabet Raised $80 Billion in a Week
Google’s parent company announced a proposed $80 billion equity raise on June 1 to fund AI compute expansion — then upsized it to $84.75 billion including at-the-market offerings. The deal includes a $10 billion strategic investment from Berkshire Hathaway, Warren Buffett’s first major tech commitment in years. Read the full breakdown at our Alphabet $80 billion analysis.
Alphabet projects capital expenditures as high as $190 billion in 2026, primarily targeting data centers, custom AI chips, and energy systems. The scale is almost impossible to contextualize: $190 billion is more than twice what the entire global cloud market spent on infrastructure five years ago.
Why it matters: This cements Google as the largest single AI infrastructure spender on the planet for 2026. It also means Gemini has more compute runway than any competitor — relevant context as Google and Anthropic race to match Claude Code’s developer mindshare.
Anthropic Is Paying SpaceX $1.25 Billion a Month
The financial details of Anthropic’s compute deal with SpaceX became public this week when SpaceX’s S-1 IPO filing hit the SEC. Anthropic is paying $1.25 billion per month through May 2029 for the entire output of the Colossus 1 data center in Memphis — 220,000 NVIDIA GPUs across 300 megawatts. The deal has already delivered: Claude Code’s rate limits are doubled and peak-hours throttling is gone for Pro and Max users.
The wrinkle: Elon Musk told followers on X the arrangement is a 180-day lease with a 90-day mutual cancellation provision. “If compute gets super tight I said we might need it back at some point,” he posted on June 1. Full analysis in our Anthropic SpaceX compute deal deep-dive.
Why it matters: The deal confirms Claude’s compute bottleneck was real and is now, at least partially, resolved. But Musk’s public exit clause is a reminder that Anthropic’s larger capacity builds — with Amazon and Google — don’t fully come online until 2027.
DeepSeek Raises Capital for the First Time — at a $59B Valuation
DeepSeek, the Chinese AI lab that blindsided the industry in January with its low-cost V3 and R1 models, is preparing to raise $7.4 billion in its first-ever external funding round. The deal would value the company at up to $59 billion, with Tencent and CATL reportedly among investors. Founder Liang Wenfeng is contributing roughly 40% of the round himself.
The round could close within weeks, according to multiple reports. It marks a significant shift for a company that built global prominence without any outside venture capital — and signals that DeepSeek is moving from research novelty to sustained competitor.
Why it matters: A $59 billion valuation for DeepSeek isn’t just a China story — it’s a direct signal that open-weight AI models have a viable long-term business behind them. Every Western AI lab will be watching the terms closely.
OpenAI Opened ChatGPT Ads to Everyone — No Minimum Spend
OpenAI’s self-serve advertising platform launched in beta on June 5, removing the previous $50,000 minimum spend requirement. U.S. advertisers can now register, set budgets, launch campaigns, and access cost-per-click bidding directly through an Ads Manager portal. Conversion-optimized campaigns began rolling out the same day.
OpenAI is targeting $2.5 billion in ad revenue for 2026 and $100 billion by 2030. Agency partners Dentsu, Omnicom, Publicis, and WPP are already in the ecosystem, alongside ad tech platforms including Adobe, Criteo, and StackAdapt.
Why it matters: This is the moment AI search advertising becomes a real category. ChatGPT’s self-serve launch puts it on the same access tier as Google Ads — meaning small and mid-size businesses can now run ads inside the product that’s increasingly replacing their Google searches.
Anthropic’s Mythos Model Is Now Protecting Critical Infrastructure in 15 Countries
Anthropic expanded Project Glasswing — its AI-powered vulnerability discovery program — to approximately 150 new organizations in more than 15 countries, including sectors like power, water, healthcare, and communications. Claude Mythos, Anthropic’s most powerful model yet, is powering the scans and can identify thousands of zero-day vulnerabilities over several weeks.
New partners include Okta, Samsung, SK Hynix, and SK Telecom. NATO and the EU’s ENISA cybersecurity agency are also participating, per Financial Times reporting. The expansion came one day after Anthropic’s IPO filing last week.
Why it matters: AI as infrastructure-security tooling is a genuinely new category. The question is whether Anthropic can demonstrate zero-day discovery at scale without introducing new risks — and whether rival models like OpenAI’s GPT-5.5-Cyber compete on this territory.
What to Watch Next Week
Apple WWDC starts June 9. The keynote will be the first to show the new Siri architecture running on Google Gemini under the hood — and Apple’s answer to the question of whether it can compete in the AI era without a frontier model of its own. Separately, watch for SpaceX’s IPO pricing timeline: the S-1 filing is now public, and a roadshow could begin within weeks. If it prices above $200B, it will be the largest tech IPO since Meta.
Last Updated: June 2026
