AI's promise to improve productivity remains largely theoretical. Until that happens, it's helping the economy in another way, with massive spending.

Companies are pouring billions of dollars into artificial-intelligence development. Their promise: AI will allow fewer employees to get more done, streamline operations, reduce waste, and ultimately bolster corporate profits and the broader economy. While it might be question of when that happens -- rather than if -- the "when" is still up for debate.

In the meantime, the investment stage of the AI buildout has been good for the economy. ING Chief International Economist James Knightley estimates that AI will account for a third of U.S. economic growth in 2026, including an array of inputs, from computing, to software and data centers.

That roughly dovetails with estimates from St. Louis Fed economists, who in January calculated that AI-related categories (information processing equipment, software, R&D, and data centers) made up 39% of total gross domestic product growth in the third quarter of 2025.

Major tech players have already committed eye-watering amounts of capital to AI, but the government is ponying up as well. This spring, the Defense Department proposed spending nearly $30 billion on its new AI Arsenal initiative in fiscal 2027, which would go toward building government-owned AI infrastructure.

But AI's contribution to the economy and stocks isn't entirely positive. All that spending has pushed up inflation too, and not just for the tech sector. In August, electricity prices surged 3.8% year over year, according to consumer price index data.

Moreover, for all the money that has been spent, productivity gains haven't shown up yet.

Apollo Chief Economist Torsten Slok looks at total factor productivity, or TFP, as the closest gauge of technological progress. According to the Bureau of Labor Statistics, "TFP tells us how much can be produced without adding more inputs" -- such as more hours worked or more machines added. By that metric, AI hasn't really made much of a dent, Slok says.

Essentially, AI is often getting more credit than it deserves, he says. That's because productivity gains are coming from companies adding more inputs to the equation, in the form of more machinery or worker hours.

For instance, the San Francisco Fed sees "no sign of acceleration" in utilization-adjusted TFP since the AI capital expenditure cycle began, he says. That metric is just below zero.

"Output per hour, by contrast, is running near 2.5%, comfortably above its post-2005 average, and that strength is exactly what gets cited as evidence that AI is already working," Slok writes. "But strong output per hour alongside flat TFP is the signature of capital deepening, not of a technology shock."

But previous technological innovation also took time to show up in the numbers.

"The bottom line is that the AI boom is clearly visible in investment data and in equity valuations, but it is not yet visible in the productivity statistics, which means the productivity payoff from AI remains a forecast rather than an observation," Slok concludes.

All of that may lead investors to wonder about what will keep the economy afloat after the billions of buildout money has been spent.

But that sky-high spending is beneficial in its own right, writes Capital Economics' Paul Dales, even when accounting for the drag of higher prices. Increasing demand for AI could help avoid a recession in economies around the world. Without a downturn, the eventual economic boost from AI productivity gains-when they do finally come on line-could come sooner.

Of course, stocks are reaping the rewards now, with both the S&P 500 and Nasdaq Composite reaching new all time highs this week-something each index has already done more than 20 times in 2026, largely fueled by the AI trade.

If AI's buildout continues to keep economies growing, that could mean a more seamless transition to the point when its productivity does start to make a meaningful difference.