Musk Sees AI’s Next Bottleneck in the Power Grid, Warns of 15-Gigawatt Shortfall

• Elon Musk estimates AI could expand the global economy by 20% to 30%, adding as much as $30 trillion a year in output.

• He says AI could reach “Stockfish-level” performance in software development within 12 to 18 months.

• Musk expects at least one billion humanoid robots within a decade, with each producing roughly five times the output of a human.

• Behind those forecasts, he sees a more immediate problem: at least 15 gigawatts of additional power could be needed for AI chips in 2027.

September 02, () — Elon Musk has spent years predicting that artificial intelligence will transform the economy.

At the G20 Innovation Ministerial in Chapel Hill, North Carolina, on September 1, he put a number on the transformation and a limit on how quickly it could happen.

Speaking virtually to ministers and technology leaders, the Tesla and SpaceX chief estimated that AI could increase the size of the global economy by 20% to 30%, equivalent to roughly $20 trillion to $30 trillion in additional annual output.

“I think, just to give you some sense of scale here, AI will probably increase the global economy by 20 to 30%,” Musk said. “That is my rough estimate. Meaning, on the order of $20 to $30 trillion per year.”

His timeline for reaching that potential is equally aggressive. Musk said AI could become capable of handling essentially any digital task that does not involve physically manipulating matter by the end of 2027.

Software development, he suggested, could reach a particularly dramatic turning point.

Musk compared future AI coding ability with Stockfish, the chess engine capable of operating at a level far beyond human players.

“My prediction is that sometime next year, AI software will become so good that it reaches the Stockfish level,” Musk said, arguing that humans would then be unable to compete with AI in writing software.

He expects AI to become extremely capable across engineering and other digital work within roughly 12 to 18 months.

Power Becomes the Constraint

For all the optimism surrounding AI, Musk spent much of his remarks describing a problem that has little to do with algorithms.

Electricity.

Musk warned that the rapid expansion of AI chips is running ahead of the growth in available power outside China.

The Tesla boss stated that analysts following the sector expect a power shortfall of at least 15 gigawatts in 2027 for AI chips.

AI chip production, according to Musk, is increasing by roughly 40% to 50% a year, while electricity availability outside China is growing at only about 10% to 20%.

If those rates continue to diverge, data centers could have increasingly powerful chips sitting idle because there is not enough electricity to run them.

“There’s actually quite a crisis of power,” Musk said. He argued that the shortage is already influencing where companies place computing capacity.

Musk pointed to Google, Anthropic and other companies leasing compute from SpaceX, saying his company was able to bring AI capacity online quickly because it built its own power plants.

He argued that countries seeking to attract AI investment should therefore focus not only on data centers but also on building the electricity infrastructure needed to operate them.

China, he noted, has substantial electricity capacity, but U.S. restrictions on advanced GPU exports limit the ability to build cutting-edge AI data centers there.

That leaves electricity availability outside China as a particularly important constraint for the global AI buildout.

Musk’s longer-term vision extends well beyond software. He expects humanoid robots to multiply rapidly as improvements in AI software, chips and mechanical systems reinforce one another.

Within 10 years, he said, there could be at least one billion humanoid robots, with each producing roughly five times the output of a human. Collectively, he argued, they could become more productive than the entire human population.

“I’d be willing to put serious money betting on” that outcome, Musk said, describing the estimate as conservative.

Regulation was the other major part of his message. Musk urged governments to create an environment where emerging technologies are presumed legal rather than restricted by default, taking particular aim at the European Union’s regulatory approach.

“You have to have an environment that’s relatively free of regulation,” Musk said, arguing that excessive regulation “inhibits progress of new technologies” and slows development.

His comments came as the U.S. pushed the G20 toward a lighter-touch approach to AI governance through the proposed Carolina Principles, while European regulators continued implementing the EU AI Act.