The numbers from Q1 2026 tell a story that should concern anyone who cares about where this technology goes next. OpenAI, Anthropic, xAI, and Waymo absorbed $188 billion in a single quarter โ€” roughly 80% of all global AI funding, concentrated in four companies that already had enterprise traction. That's not a competitive market. That's a land grab with a quarterly body count.

The Enterprise Playbook Working Exactly as Designed

Anthropic's growth trajectory is the most telling data point in this entire discussion. From $87 million ARR in early 2024 to $47 billion by May 2026 โ€” the fastest B2B revenue ramp ever recorded, according to every tracker covering the period. Eight of the Fortune 10 are Claude customers. Over a thousand companies now spend more than $1 million annually with them. The company charges roughly $211 per monthly user versus OpenAI's $25. That's an 8x premium that SaaStr described as 'the enterprise playbook working exactly as designed.' And they're getting it, which tells you everything about who this technology is actually for. The token pricing structure makes the gate explicit. Claude Opus 4.7 runs at $5 per million input tokens and $25 per million output tokens. Average SMB spend on Anthropic hits $24,000 annually; enterprise customers pay $85,000. Those numbers describe organizations with procurement departments and IT budgets. They don't describe a teacher in Karachi trying to grade papers faster, a clinician in Nairobi who needs diagnostic support, or a first-generation student without access to an institution willing to sign a volume contract.

GitHub Copilot's Token Shock

GitHub Copilot's billing switch on June 1 made this concrete for individual developers. Flat-rate subscriptions got replaced with token-based pricing, and engineers reported costs jumping from $29 to $750 per month as agentic workflows consumed credits at scale. TechCrunch called it the end of 'the golden age' of GitHub Copilot โ€” a phrase that should land differently when you realize what they're actually describing is developers discovering that automation multiplies consumption faster than unit costs fall. Microsoft learned this lesson the hard way: they cancelled Claude Code licenses after engineers burned through their entire 2026 AI budget in four months. Uber's CTO reported the same dynamic. The models are getting cheaper per token. But agentic workflows โ€” extended reasoning chains, long-horizon tasks, frontier-level capabilities โ€” consume at a rate that makes those unit economics irrelevant for anyone without enterprise-scale procurement.

The Global Pricing Gap

Claude Pro runs at $20 per month globally with no purchasing power parity adjustment. In Vietnam, that's five to seven percent of average urban monthly income. OpenAI introduced a locally-priced ChatGPT Go tier at the equivalent of $5 per month for Vietnam and fifteen other Asian markets. Anthropic has not. One analysis tracking AI access across Asia described Claude Pro as 'the conspicuous outlier' โ€” technically available but priced at a level that functionally excludes entire markets from frontier capabilities.

Export Controls Achieve What Pricing Does

The Trump administration's Commerce Department export directive arrived at the same destination from the opposite direction, pulling Fable 5 and Mythos 5 from every non-American user. Helen Toner of Georgetown put it plainly: barring foreign nationals from frontier models is 'essentially equivalent to preventing any company affected from doing any further AI R&D work.' State authority achieves through security controls what enterprise pricing achieves through market mechanics. The people excluded are identical in both cases โ€” and that's not an accident.

Open Source Closes the Gap, But Not for Everyone

Here's where the argument gets complicated. Meta, Alibaba, DeepSeek, and MiniMax have closed the performance gap with proprietary APIs faster than most predicted. Epoch AI research shows frontier inference prices falling by 9x to 900x annually depending on benchmark. The MMLU gap between open and proprietary models narrowed from 17.5 percentage points to 0.3 in a single year. Anyone with a consumer GPU can now run models approaching frontier capability at near-zero marginal cost. But Gartner's counterpoint lands: 'Chief Product Officers should not confuse the deflation of commodity tokens with the democratisation of frontier reasoning.' Cheaper tokens don't equal cheaper outcomes when consumption scales with capability. And open-source access still requires hardware โ€” running a 70B parameter model locally needs GPU clusters costing thousands of dollars. Self-hosting at production scale requires infrastructure that a university in Lagos or a clinic in Dhaka cannot provision. The accessibility argument for open source holds for developers in high-income economies who already have the infrastructure. For the Global South, it substitutes one barrier for another.

Key Takeaways

  • Four companies captured 80% of global AI funding in Q1 2026 โ€” $188 billion concentrated among OpenAI, Anthropic, xAI, and Waymo
  • Enterprise pricing ($85K/year average) and export controls (Fable 5/Mythos 5 blocked internationally) achieve the same exclusion through different mechanisms
  • Open-source models have closed performance gaps dramatically, but hardware requirements still gate Global South access
  • Agentic workflows multiply token consumption faster than unit costs fall โ€” making 'cheaper AI' a misleading headline

The Bottom Line

A capable model openly accessible could compress decades of medical research for hospitals without research divisions, give first-generation students the analytical support that elite tutoring provides, let small businesses in Accra compete with multinationals. That capability exists right now. The question isn't whether it can reach those people โ€” it's whether the people controlling it will allow it to. From both the labs and Washington, the answer is not yet.