April told you the AI race had moved from models to infrastructure. June tells you what infrastructure actually costs — and who pays for it. Not just in dollars. In water. In electricity. In jobs. In a concentration of power so extreme that three companies now run the intelligence layer beneath nearly every major platform on earth.
THE CONSOLIDATION NOBODY NAMED Google's Gemini powers iOS, Android, and Samsung Galaxy. OpenAI underpins Microsoft's Copilot. Anthropic's Claude drives Alexa and AWS. Three models, every major consumer surface on earth. June 8th made it undeniable: Apple unveiled a rebuilt Siri running on a custom Gemini model, paying Google roughly $1 billion a year — the world's most valuable company conceding it couldn't build its own intelligence. The same week, Microsoft unveiled its own MAI family to cut its $13 billion OpenAI dependency, and Google rebuilt Search around AI agents. In one month, the platforms that dominate consumer tech rewired themselves around AI. THE IPO WAVE — AND WHAT IT MEANS FOR YOU OpenAI filed a draft S-1 on June 8th at $852 billion, ChatGPT at ~900 million weekly users, no positive free cash flow before 2029. Anthropic filed June 1st at $965 billion after a $65 billion raise, revenue run-rate crossing $47 billion. SpaceX went public June 12th near $2.1 trillion — three near-trillion-dollar listings in one month, bigger than France and Italy's combined GDP. Here is what the financial coverage consistently fails to say: when these companies go public, the losses become the market's problem — packaged into pension funds, index trackers, and retail portfolios held by people who never decided to fund AI infrastructure. The public does not get a vote on whether it gets built. But increasingly, they pay for it. Communities near data center corridors don't vote on grid allocations or water permits any more than shareholders vote on capex — the resource costs, like the financial ones, land on people who had no say in the buildout.
THE JOBS ARITHMETIC 267 layoff events in 2026 have hit 185,894 workers so far — roughly 1,033 a day — and 56% now cite AI explicitly, up from under 8% in 2025. Oracle cut 30,000 positions despite strong earnings; Block's Jack Dorsey cut 4,000 and said plainly it was "the growing capability of AI tools" — the most candid AI-layoff admission on record. Stanford data shows workers aged 22–25 in AI-exposed roles saw a 13% relative drop in employment: the pipeline producing tomorrow's experienced workers is closing, not just today's jobs. India's stakes are highest — generative voice AI is hollowing out its $40 billion BPO sector (a human agent costs $10–15/hour against pennies-per-minute for AI), and India's top IT firms added a net of just 17 employees in fiscal 2026's first nine months. As the IMF's managing director put it at Davos, AI is hitting labor markets "like a tsunami."
EUROPE BUILDS — UNEVENLY Germany is winning by building AI into machinery, not the consumer interface: Deutsche Telekom and NVIDIA launched the world's first Industrial AI Cloud in Munich on nearly 10,000 Blackwell GPUs, and NEURA Robotics raised up to €1.2 billion for industrial robots. Portugal is repeating India's 1990s playbook, drawing skilled, EU-compliant talent at a fraction of London or Berlin's cost via Microsoft's $10 billion Sines project. France's Mistral is racing for a sovereign full stack, but its €830 million debt raise looks thin beside OpenAI's and Anthropic's near-trillion-dollar valuations — racing to avoid the Apple-Google fate, where the most capable player still couldn't build its own model. Switzerland is the quieter story, and arguably the deeper one. It now directs a higher share of venture capital to deep tech than any country on earth — 63% from 2020–2026, ahead of China and the US — with ETH Zurich alone producing more VC-backed deep-tech spinouts since 2020 than MIT or Stanford. Swiss deep-tech funding hit a record $2.6 billion in 2025; AI-specific funding grew 206% year-over-year. Robotics is scaling even faster, at 3.5x the VC-funded startup density of the US, per capita. The gap: 88% of Switzerland's late-stage capital still comes from foreign investors — the research and founders are Swiss, but the big checks at scale mostly aren't yet. TWO PARALLEL WORLDS China isn't catching up — it's arrived. Chinese open-source models are 30% of global AI downloads against 15.7% for US models; Alibaba's Qwen alone has spawned over 100,000 derivative models. DeepSeek trained V3 for roughly $6 million against an estimated $100 million for GPT-4. The models diverge sharply in framing too — international law and democratic norms versus state sovereignty; on Taiwan, DeepSeek simply declines to answer. The world isn't just choosing a model. It's choosing a worldview. The Gulf is the wild card both sides court and neither controls. Regional electricity runs $0.05–$0.06/kWh against $0.09–$0.15 in the US — cheap energy here is what cheap land once was to agriculture. G42's UAE–US AI Campus in Abu Dhabi, with OpenAI, Oracle, NVIDIA, Cisco, and SoftBank as partners, spans 10 square miles at a $30 billion cost, scaling to 5 gigawatts. Gulf sovereign wealth — PIF, Mubadala, MGX, QIA — now sits inside OpenAI, Anthropic, xAI, and SpaceX simultaneously. One caveat the prospectuses omit: in April 2026, Iran's IRGC claimed to have struck an Oracle data center in Dubai — the Gulf is building the AI future on contested geography.
THE PLANET NOBODY CONSULTED Global data center electricity demand is projected to hit 945 terawatt-hours by 2030 — comparable to Japan's entire annual use. US data centers consumed roughly 17.4 billion gallons of water in 2023, headed toward 38–73 billion by 2028; a UN report this June found AI could use as much water as 1.3 billion people by 2030. Microsoft's carbon footprint has risen 23.4% against its 2020 baseline, Google's roughly 48% since 2019 — both attributed to AI expansion, both leaning on carbon credits over real reductions. Pushback is real: roughly $64 billion in projects delayed or canceled in under a year, Maine's outright ban, and residential electricity prices projected to climb 6–29% on data center demand alone. Not every region absorbs this the same way. Khazna Data Centers, the UAE's largest wholesale provider, built adiabatic and evaporative cooling in from day one — necessity, since local climate ruled out the free-air cooling operators elsewhere use. It targets a PUE below 1.5 against the industry's 1.57 average, holds LEED Gold and Estidama Pearl 4 certification, and its DXB8 site became the first data center globally certified zero-waste in 2026. Part of the gap is simple topography: the UAE's sparsely populated desert interior lets operators build without competing against existing neighborhoods for land or water — a head start dense US corridors never had.
WHAT THE RACE ACTUALLY MEANS Three American companies run the intelligence layer of nearly every consumer platform. Three Chinese companies power the developing world's AI at a fraction of the cost, worldview built in. Three Gulf states fund both sides while building sovereign capability beneath. High-income countries capture 91% of AI venture funding on 17% of global population; Africa and Latin America hold 3% of global AI compute, and only 3% of AI's projected $19.9 trillion benefit by 2030 reaches Latin America. For people in wealthy economies, the consequences are already material: electricity bills funding grid upgrades built for data centers, not residents; pension funds loaded with AI equity before profitability is proven; entry-level jobs vanishing before reskilling exists; public infrastructure consumed by private decisions made without public consent. The AI industry has spent three years asking what intelligence can do. June 2026 forces a harder question: for whom? Infrastructure is not neutral. It embeds choices about who benefits, who pays, whose values get encoded, and whose water gets used. Oil will run out. Compute will not. What we haven't yet decided — and must — is whether the race to control it produces a world that is more intelligent, or merely one where intelligence is more concentrated.