The AI third world war, and why Apple is waiting for

What happened

The technology industry is witnessing a battle unlike any before. OpenAI, Google, Anthropic, Meta, xAI and Microsoft are spending money on artificial intelligence at a scale that would have seemed absurd even a few years ago. They are not merely competing to make the smartest chatbot. They are fighting to control what could become the next great platform of computing. Many call this clash the “AI Third World War”.

The danger for a loser is not merely having an inferior product. It is being locked outside the ecosystem. That is why the AI Third World War metaphor is useful. The striking thing about Apple is not that it has ignored AI. It plainly has not. Its research and development spending has been rising sharply, with the company explicitly attributing part of the increase to infrastructure and AI investment.

But Apple has not fought the AI war in the way OpenAI, Google, Meta or xAI have. It has not tried to build the world's biggest frontier model at almost any cost. Instead, it has chosen a strikingly different strategy: partner where necessary and integrate where it has an advantage. Apple has a long history of arriving late to technologies and then redefining them.

Why it matters

The phrase is deliberately dramatic. There are no armies or battlefields, but the stakes are comparable to a technological war: who controls the intelligence layer that may sit above operating systems, search, software, commerce and, eventually, much of the digital economy? The numbers explain the intensity. OpenAI’s Stargate project was announced with an ambition to invest $500bn over four years in AI infrastructure.

OpenAI and its partners have subsequently expanded the planned data-centre footprint, with more than $400bn of investment and nearly 7GW of capacity under development by Sept 2027. Google is hardly sitting back. Alphabet has raised its expected 2026 capital expenditure to $195-205bn, driven by the need for servers, data centres and other technical infrastructure. Meta expects to spend $130-145bn on capital expenditure this year, with AI infrastructure a principal driver.

Amazon expects to spend roughly $220bn, much of it on AI-related infrastructure, while Microsoft expects around $190bn in capital expenditure in calendar 2026. Even the younger AI companies are thinking in wartime numbers. Anthropic has reportedly signed a $45bn, six-year agreement to rent computing capacity from Nscale. It has also committed billions more to computing through deals involving Amazon, Google and SpaceX.

xAI, meanwhile, raised $20bn in Jan and says its Colossus supercomputers had crossed the equivalent of 1mn H100 GPUs by the end of 2025. This is therefore not simply a software race. It is a race for chips, electricity, data centres, engineers, capital and customers. Whoever possesses enough computing power can train bigger models; better models attract more users; more users generate revenue and data; revenue finances more computing.

What to watch

The iPod was not the first MP3 player. The iPhone was not the first smartphone. Yet Apple understood that winning a platform is about much more than inventing a category. It is about integrating hardware, software, services, design, distribution and a compelling user experience. Why spend hundreds of billions discovering which architecture, model and business model will ultimately win if competitors are willing to spend that money for you?

OpenAI, Google, Anthropic, Meta and xAI are educating consumers, developing the infrastructure, recruiting the talent, and testing the economics. Apple can watch. And watching is not necessarily weakness. Apple has something the AI labs desperately need: distribution. 5bn active Apple devices put the company in an extraordinary position if AI eventually becomes a mainstream interface to computing. That may explain why Apple’s current strategy looks less like surrender than hedging. Platforms become powerful precisely because they attract developers and users early.