A Google DeepMind executive has predicted that global spending on artificial intelligence could hit $1 trillion by 2026. The driver, they said, would be machines that improve themselves through recursive self-improvement — a cycle where AI systems get smarter without direct human input.
The $1 trillion forecast
The prediction came from a senior figure at DeepMind, the Alphabet-owned AI lab behind systems like AlphaGo and Gemini. The executive didn't specify which types of spending would make up that figure, but capital expenditure in AI typically covers data centers, specialized chips, and research infrastructure. A trillion dollars would dwarf current levels. For context, global semiconductor capital spending was roughly $150 billion in 2023. AI-specific capex is a fraction of that today.
Reaching $1 trillion would require an enormous ramp-up in investment from governments, tech giants, and venture capital. The executive's timeline — just three years from now — suggests they believe the industry is on the cusp of a spending explosion.
Recursive self-improvement as the engine
The key mechanism behind the forecast is recursive self-improvement. That's a concept where an AI system can rewrite its own code or design better versions of itself, leading to rapid, compounding gains in capability. Once a machine reaches a certain threshold, it could accelerate progress far beyond what humans can achieve alone.
DeepMind has long studied self-improving algorithms. The executive argued that this process would soon become practical, not just theoretical. If it does, the demand for compute power — and the money to pay for it — could skyrocket. Companies would race to build the most powerful systems, driving capex higher.
What the prediction means for the industry
If the $1 trillion figure proves accurate, it would reshape the tech landscape. Hardware makers like Nvidia and AMD would see orders surge. Cloud providers would expand data centers at an unprecedented pace. Startups building AI infrastructure would attract massive funding.
But the prediction also raises questions. Can the industry actually spend that much in three years? Supply chains for advanced chips are already strained. Energy grids may struggle to power the data centers needed. And there's no guarantee that recursive self-improvement will work as smoothly as the executive expects.
DeepMind itself has not issued an official forecast. The executive spoke in a personal capacity, though their position gives the prediction weight. The company declined to comment further.
For now, the trillion-dollar mark remains a target — one that will test whether the AI industry can deliver on its most ambitious promises.
