Anthropic Becomes the World's Most Valuable AI Company, GitHub Copilot Flips to Usage Billing, Colorado's AI Act Arrives

June opened with a seismic shift in AI's power structure. Anthropic closed a $65 billion round that valued it at $965 billion — leapfrogging OpenAI to become the world's most valuable private AI company. GitHub's Copilot billing model flipped overnight from flat-rate to pay-per-token. And Colorado's rewritten AI Act now takes effect June 30, giving businesses 29 days to get their compliance houses in order.

Let's cut through the noise.

Signal #1: Anthropic's $965B Valuation — The Frontier Just Went Bipolar

What happened: On May 28, Anthropic announced a $65 billion Series H at a $965 billion post-money valuation, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with Coatue and ICONIQ co-leading. That's a dramatic jump from its roughly $380 billion valuation just a few months ago. It's also the largest single funding round ever raised by a private AI company. Anthropic now surpasses OpenAI in private-market value.

Alongside the funding, Anthropic shipped Claude Opus 4.8 — an upgraded flagship model with better coding, stronger reasoning, and more capable multi-step task performance. The company also announced that Mythos-class models will reach customers in the coming weeks, once updated safety controls are ready.

Why it matters: This isn't just a vanity number. A near-trillion-dollar valuation gives Anthropic the capital to fund aggressive model training, subsidize enterprise expansion, secure infrastructure contracts (including a massive SpaceX Colossus compute deal worth ~$1.25B/month through 2029), and stay in the frontier race even as costs escalate. For the broader market, it means the frontier model space is now genuinely competitive at the top. A multi-polar market gives buyers leverage, keeps pricing contested, and prevents any single provider from dictating terms. That's good for everyone building on AI.

What doesn't matter: Whether Anthropic is "worth" $965 billion on a revenue multiple basis. Private valuations at this scale are strategic bets on compute, talent, and market positioning — not DCF analysis. The signal is that serious capital believes Anthropic can be a defining infrastructure company, not just a model shop.

What to do: If you've been building on a single provider, now is the time to adopt a multi-model architecture. With two credible frontier players trading the lead, vendor lock-in is a strategic risk. Platforms like SIM2Real can help you benchmark and route across providers, ensuring you're always on the best model for each task — not just the one you happened to integrate first. And if you're selling into enterprises, Anthropic's Big Four partnerships (KPMG, Deloitte, PwC) mean your procurement contacts are about to start hearing a lot more about Claude.

Signal #2: GitHub Copilot Flips to Usage-Based Billing — The Free Lunch Is Over

What happened: As of June 1, 2026, GitHub Copilot has officially transitioned from flat-rate pricing to usage-based billing. The new structure gives organizations pooled AI Credits that map 1:1 to US dollars, then charges per token once credits are consumed. Copilot Business ($19/user/month) includes $19 in monthly AI Credits per user; Copilot Enterprise ($39/user/month) includes $39 per user. Credits are pooled at the organization level, so a team of 10 Business users gets $190/month in shared credits. After that, pricing depends on the model you're using.

GitHub is framing this as "more flexibility and model choice" — and technically they're right, since the new system opens access to multiple models beyond just the default. But the practical reality is that heavy Copilot users are about to see their bills climb.

Why it matters: This is a structural shift in how developer-tool pricing works. For two years, Copilot was the flat-rate bargain of AI tooling — $10–19/month for essentially unlimited AI coding assistance. That era is over. The shift to usage-based billing mirrors what's happening across SaaS: as AI inference costs become more transparent, vendors are passing those costs downstream. For founders and engineering leaders, this means your dev-tool budget is now variable. A developer having a productive day with Copilot could easily burn through $50+ in credits, and your org will pay the overage.

What doesn't matter: The specific per-token pricing for individual models. What matters is the structural shift from predictable flat costs to variable usage costs, and whether your team has the budget monitoring in place to handle it.

What to do: Audit your Copilot usage now. Understand how many AI Credits your team burns through in a typical sprint, and model the cost under the new system. If you're running large-scale agent workflows that hammer Copilot, consider routing heavy-lift tasks to lower-cost providers like DeepSeek (whose permanent 75% price cut we covered last week). Tools like SIM2Real can help you set up dynamic routing so the right model handles the right task at the right cost. And if you're building developer tools, take note: the era of flat-rate AI pricing is ending. Plan your own billing accordingly.

Signal #3: Colorado's Rewritten AI Act Takes Effect June 30

What happened: On May 14, Colorado Governor Polis signed SB 189, which repealed and replaced the state's original 2024 AI Act with a more targeted framework. The new law takes effect June 30, 2026. Key changes: it narrows the definition of "high-risk" AI systems, shifts enforcement to the Colorado Attorney General (no private right of action), requires risk management programs and impact assessments for covered systems, and mandates transparency disclosures for deployers of AI in consequential decisions. The original law's broad duty-of-care obligations are gone, replaced with more specific compliance requirements.

Why it matters: This is the first significant state-level AI regulation to take effect in the US, and it sets a template. If you're an AI company serving Colorado consumers — which, given Colorado's population, most national companies are — you have 29 days to assess whether your systems fall under the new requirements. The law applies to deployers and developers of "high-risk" AI systems that make or substantially assist in consequential decisions about employment, education, financial services, housing, and similar domains.

What doesn't matter: The philosophical debate about whether state-level AI regulation creates a "patchwork problem." That ship has sailed — multiple states are regulating AI regardless. What matters for builders is: does my product trigger compliance obligations, and can I meet them?

What to do: If you're deploying AI in hiring, lending, insurance, housing, or education decisions, run a gap analysis now. Platforms like ProvenanceOS are built for exactly this — providing audit trails and traceability for AI-assisted decisions so you can demonstrate compliance. Eco-Auditor can help assess whether your AI systems meet emerging regulatory standards around bias, transparency, and impact documentation. Don't wait until June 29.

Noise: SpaceX's $60B Cursor Acquisition Option

SpaceX secured an option to acquire the coding startup Cursor for $60 billion — or alternatively pay $10 billion for a collaborative work arrangement. The deal highlights how compute and tooling are consolidating under a small number of infrastructure players. But for builders, this is noise today. The option has no closing date set, Cursor continues operating independently, and the practical impact on your tooling choices is zero. You don't need to migrate away from Cursor or reconsider your IDE. If the acquisition closes and things change, we'll cover it then.

Our Take

Three stories, one theme: AI's center of gravity is shifting from model capability to infrastructure and governance. Anthropic's megaround proves that capital is flowing toward companies that can control the full stack — compute, models, and deployment. GitHub's billing change proves that the economics of AI usage are becoming more transparent and more variable, whether you like it or not. And Colorado's AI Act proves that regulation is no longer theoretical — it has a compliance deadline 29 days from now.

The founders who win in this phase won't be the ones who picked the best model in 2024. They'll be the ones who built architectures flexible enough to route between models, track their AI decision-making for compliance, and manage costs in a world where the best model changes every month. Build for flexibility, audit for compliance, optimize for cost. That's the June 2026 playbook.