The AI War went public in June. Meta attempted the unthinkable: pry away OpenAI’s senior researchers, the minds behind GPT-4, with offers reportedly reaching $100 million per person. When that failed, Mark Zuckerberg aimed to buy Thinking Machines Lab, the startup led by Mira Murati (ex-OpenAI CTO), for $1 billion. She said no.
Next, a commando-style talent raid: a headline proposal of $1.5 billion over six years to Andrew Tulock, a cofounder-level machine-learning prodigy. Another no. That’s the real face of the AI War, not cheerful garage startups, but billion-dollar bids for brains.
This isn’t about using ChatGPT to write emails. It’s about JP Morgan finding its AI does the work of 3,000 analysts, and Klarna cutting 700 support roles because their bot performs faster, 24/7. Companies aren’t dying in a decade; they’re dying now.
Meanwhile, the giants collectively torch $17 million per minute on AI. We’re talking $400B in 2025 alone, data centers drawing city-scale power, and Microsoft literally moving on nuclear power to feed its servers. This isn’t a bubble; it’s a war, where losers stop existing, like Kodak or Nokia. Only this time, it’s faster and more final. The new rule is simple: innovate or die. The giants already get it: AI isn’t the next innovation, it’s the last one, the one that eats the rest.
Trillions in motion: the capex shock of the AI War
We’ve never seen capital mobilized this fast.
- Amazon: >$100B in 2025 via AWS.
- Microsoft: >$80B, plus moves around the Three Mile Island site to power a future AI campus.
- Google: $75B despite trimmed ambitions.
- Meta: $65B, building 2 GW-class data centers, city-scale energy draws.
Add it up: the AI War is larger than Apollo + Manhattan Project combined, the biggest coordinated tech effort in human history. Markets went vertical: OpenAI pegged around $300B, 49 startups raised $100M+ in 2024, and 20 crossed $2B valuations in months. Investors price 17–18× revenue on models that didn’t exist three years ago.
Then there’s Stargate, a joint push tied to OpenAI, SoftBank, and the U.S., aiming to marshal $500B over four years to build global AI infrastructure. Money is only one weapon; the AI War also rages where it hurts most: talent.
Talent is the weapon: how the AI War strips companies of their brains
OpenAI’s median compensation has soared near $865K/year, with top packages >$1.3M. Yet 74% of companies still can’t hire the AI talent they need. Cue the workaround: shadow acquisitions.
- Microsoft–Inflection (Mar 2024): not an “acquisition,” but $650M for the CEO, cofounder, ~70% of the team, plus a license to the tech.
- Amazon–Adept: CEO + core team hired, $330M for licensing.
- Google–Character.AI: billions tied to recruiting Noam Shazeer and key engineers while the shell remains.
Playbook? Don’t buy the company, remove its brain. Regulators watch M&A over $100M; so giants hire the people, license the IP, and leave a husk.
Zuckerberg even contacted a dozen+ employees after Murati’s refusal. Zero departures. In the AI War, loyalty to certain leaders beats money, sometimes.
End-to-end control is the goal: from silicon to UI. Build your own chips, host your own models, lock services into your own ecosystem. If a startup shines too brightly, buy the team, the patents, or the supplier.
Sometimes it’s absurd. Builder raised $450M (Microsoft, Qatar, others) for “self-coding AI” that turned out to be human devs paid $8/hour. In this game, it’s rational to lose $10B on mirages rather than miss the $1T OpenAI-sized wave. Cold math, survival logic.
Geo-politics of the AI War: national survival by silicon
The U.S. lines up roughly $355B when you mix the CHIPS Act, Stargate, and massive private bets on OpenAI/Anthropic, keeping AI leadership onshore.
China targets around $238B with a clear aim: AI dominance by 2030.
Europe wakes up late; Invest AI sketches ~€200B public-private to claw back digital sovereignty.
Export controls on semiconductors tighten, rulebooks proliferate, and a blunt message emerges: whoever dominates AI dominates the 21st century, and perhaps the third millennium, especially once AGI arrives.
AGI, Artificial General Intelligence, would reason, understand, learn, and adapt like (or beyond) humans across domains. It’s not here yet. It doesn’t need to be. Today’s narrow AI already kills companies.
JP Morgan says 60,000 employees now work with one AI assistant. Walmart rolls out Retina AI with AR, and 64% of Sam’s Club members adopt the in-store tools. Siens invests $20B (half AI, half acquiring Alter) to hard-wire AI engineering. Pfizer reports trial success rates rising from 10% → 20% by using AI for molecule selection and study design.
Refuse to equip yourself, and you’ll bleed market share,then relevance.
Cash flow reality: the finance of not adopting
Since 2017, McKinsey has flagged the math: by 2030, non-adopters see competitiveness drop and free cash flow fall ~20% on average. FCF is the real cash left to pay dividends. No cash, no dividends.
Sector-by-sector in the AI War:
- Call centers: 80% automated by 2029.
- Retail: ~65% of roles made obsolete.
- Professional translation: collapsing.
- Finance/Accounting: ~40% of repetitive tasks automated.
For new entrants, timing is cruel. By 2027, you’ll have 18–24 months to build infra, hire talent, and ship processes, or the window shuts. Network effects will hand everything to a handful of winners. The AI War is a winner-takes-all arena.
Concentration & lock-in: the AI War oligopoly
- Nvidia: >90% of AI-class chips. No chips → no training → no AI.
- Amazon, Microsoft, Google: ~65% of the world’s AI hosting capacity.
- OpenAI, Anthropic, Google: ~80% of LLM market share.
Training a frontier model burns >$1B before energy, cooling, red-team, safety, and audits. Barriers are near-impassable. Everything nudges users into closed ecosystems and proprietary interfaces. Once inside, getting out is like switching mobile carriers in 1998. That’s lock-in, and it turns users into economic hostages.
The scorecard of the AI War: winners and the walking dead
Likely winners:
- The pure-play AI giants and hyperscalers.
- Traditional enterprises that mutate fast, buying, reorganizing, and wiring AI into every process.
Likely casualties:
- SMEs with no AI strategy, priced out of models, data, and infra.
- Industries in denial (trucking, old-line manufacturing, B2B services): “we’re different” are famous last words.
- Countries with no national AI plan: digital vassals renting foreign AI to run their economies.
- Workers who refuse adaptation: not replaced for being bad, but for being slow. In many firms, 80% plan to cut headcount with AI within 24 months.
The AI War is turning into the world’s invisible infrastructure, the nervous system of modern civilization, doubling in power every six months. That acceleration is itself a systemic risk: with so few actors holding the stack, a single failure could freeze the global economy in hours. Militaries and sovereign funds see it clearly: technological independence = national independence.
Conclusion: choose your side
We are living through the fastest, deepest economic shift of the modern era, faster than the Industrial Revolution and more structural than the internet because it rewrites how value is created. Three kinds of actors remain:
- Those building AI infrastructure (they’ll write the rules).
- Those mastering the tools (they’ll keep economic relevance).
- Those who disappear, through inaction or denial.
The AI War is underway. Positions are being claimed now.
Which side of history are you choosing?
-
See our full report on AI and jobs to understand the employment shock of the AI War.
-
Read more in our Transhumanism series
