Goldman Sachs is betting $7.6 trillion that artificial intelligence will reshape the global economy over the next five years. That figure, pulled from the investment bank’s own analysis, covers the total expected spending on AI through 2029. Most of it will go to computing power and data centers. Those data centers need energy grids and water supply. The money has to come from somewhere.
Here is the hard question nobody can answer yet: Will that trillion-dollar wave drive inflation up or pull it down? The answer depends entirely on whether AI boosts productivity faster than it adds costs.
Right now the numbers are moving in only one direction. U.S. investment in AI is expected to hit $750 billion this year. That is double the $375 billion spent last year. Next year the total could top $1 trillion. The companies building AI models and applications are pouring cash into training, infrastructure, and hardware at a pace that has no recent precedent.
This is not a slow ramp. It is a surge.
The economic logic is simple enough. If AI makes workers and businesses significantly more productive, they can produce more goods and services without proportional increases in labor or materials. That extra output, relative to input, tends to cool inflation. Lower inflation gives central banks room to cut interest rates. That is the optimistic scenario.
The pessimistic scenario is equally straightforward. Building and running AI systems is expensive. It requires vast amounts of electricity, specialized chips, cooling systems, and construction labor. Those costs get passed through the economy. If the productivity gains from AI turn out to be modest or slow to materialize, the spending will simply add to demand without a matching supply response. That is a recipe for higher inflation and higher interest rates.
We are in the early phase of deployment. That is the key phrase in the Goldman analysis. Early phase means there is almost no hard data yet on whether AI is actually making the economy more efficient at scale. There are plenty of claims. There are plenty of projections. There is not enough real-world evidence to settle the argument.
The next few years will decide it. That is not a hedge. It is a statement of timing. The investment is happening now. The effects will lag. By the time we know whether AI is deflationary or inflationary, hundreds of billions will already have been spent.
Consider what is being built. Data centers are not just warehouses of servers. They are energy-intensive facilities that require dedicated power plants and water for cooling. The infrastructure build-out itself is a major source of economic activity and, potentially, price pressure. Goldman’s estimate of $7.6 trillion over five years includes all of that.
The productivity question is the hinge. If AI raises output per worker faster than the cost of the machines and energy needed to run it, the economy gets a net benefit. If the cost side runs ahead, the net benefit shrinks or disappears.
There is no consensus yet. The AI boom is real. The spending is real. The uncertainty about its macroeconomic impact is equally real. That is the story the numbers tell. They do not yet tell us where it ends.
