BREAKINGVIEWS-Sam Altman inflates one-man AI bubble

(The author is a Reuters Breakingviews columnist. The opinions expressed are his own.)

By Robert Cyran

NEW YORK, Feb 9 (Reuters Breakingviews) - No one will ever accuse Sam Altman of thinking small. But the boss of artificial intelligence group OpenAI seems determined to inflate a one-man investment bubble. He is talking to investors about building a slew of semiconductor plants which could accelerate the arrival of self-teaching computers, the Wall Street Journal reported on Thursday, citing people familiar with the matter. That would apparently require as much as $7 trillion. Raising such a sum seems fanciful. Putting it to work would single-handedly eclipse previous booms.

The ambition is staggering. The world spent more than $150 billion on capital expenditure to increase chip production in 2023, according to consulting firm Semiconductor Intelligence. That sum has grown about 8% a year over the past 20 years, as the number of semiconductors increase, and production facilities grow more costly. The mooted $7 trillion figure is equivalent to global chip investment over nearly the next 20 years, assuming spending keeps growing at the same rate. Even if a sizeable chunk is earmarked for other projects like power plants needed to feed AI’s voracious demand for electricity, it’s still equivalent to about 7% of the world’s gross domestic product.

Notwithstanding the current mania for all things AI, it’s hard to imagine any project raising such amounts. Altman has had talks with the government of the United Arab Emirates, the Journal said. But its flagship sovereign wealth fund, the Abu Dhabi Investment Authority, has total assets under management of about $1 trillion, according to the Sovereign Wealth Fund Institute. Most other available pools of capital are smaller.

Even if Altman could raise the cash, putting it to work efficiently would be difficult. Taiwan Semiconductor Manufacturing (2330.TW) , the world’s largest chipmaker, spent over $30 billion on capital expenditure in 2023. Yet the start of its new Arizona plant has been delayed in part because of labor issues. Just training the workers required for several decades’ worth of new plants would represent a bottleneck. So would sourcing the specialized equipment required to make high-end chips.

Perhaps the industry is massively underinvesting, and the opportunities created by AI justify a step-change in chip production. Or maybe not. Consider the telecom bubble of the late 1990s. Existing operators and startups borrowed around $1 trillion to build fiberoptic networks capable of carrying the explosion in data unleashed by the internet, according to the U.S. Federal Communications Commission chairman at the time. That was about 3% of world GDP at the time, and far too much. Though much of the capacity was eventually put to work over the next decade, those who funded the bonanza lost their shirts. Investors must hope Altman’s one-man bubble never gets off the drawing board.

Follow @rob_cyran on X

CONTEXT NEWS

OpenAI CEO Sam Altman is in talks with investors to raise funds for a project that would increase the world’s chip-building capacity and remove constraints on the growth of artificial intelligence, the Wall Street Journal reported on Feb. 8, citing people familiar with the matter. The project could require raising as much as $5 trillion to $7 trillion, one of the people told the Journal.

Altman is offering a partnership between OpenAI, investors, chip makers and power providers which would provide money to build semiconductor plants, according to the Journal. The manufacturing facilities would be run by existing chip makers, and much of the venture would be financed by debt.

“OpenAI has had productive discussions about increasing global infrastructure and supply chains for chips, energy and data centers—which are crucial for AI and other industries that rely on them,” an OpenAI spokeswoman told the Journal. “We will continue to keep the U.S. government informed given the importance to national priorities, and look forward to sharing more details at a later date.”

(Editing by Peter Thal Larsen and Sharon Lam)