Goldman Sachs has run the numbers. The bank estimates that $7.6 trillion could be spent on artificial intelligence over the next five years. That is a staggering sum. Most of it will go into computing power and data centers. Those data centers need energy grids. They need water supply. The infrastructure bill alone is enormous.
Investment in AI in the United States is expected to hit $750 billion this year. That is double last year’s $375 billion. Next year, the figure could top $1 trillion. The money is flowing from companies developing AI technology and its applications. They are pouring unprecedented amounts into training models and building the physical backbone of the industry.
This raises a straightforward economic question. Does all that spending push inflation up or pull it down? The answer hinges on productivity.
AI could raise productivity rates faster than the costs of deploying it. If that happens, the technology is disinflationary. More output per worker, per dollar spent, means prices can fall or at least rise more slowly. Lower inflation opens the door for lower interest rates. That is the optimistic scenario.
But there is a darker version. The costs of deploying AI could outweigh its productivity benefits. Companies spend billions on hardware, energy, and water. They get back less in efficiency gains than they paid out. That scenario is inflationary. It pushes prices up. Interest rates stay high or rise further.
The AI sector is still in its early phase of deployment. No one knows which path it will take. The next few years will be decisive. The spending numbers are already huge, and they are accelerating. The US is on track to spend more on AI this year than it did in the entire prior two years combined.
Consider what $7.6 trillion buys. That is roughly the size of the entire German economy. It is being directed into a single technology sector over five years. The sheer scale of the bet is what makes the productivity question so urgent. If the bet pays off, the economy gets a permanent boost in efficiency. If it does not, the money is sunk into assets that generate less value than they cost.
Goldman Sachs is not the only firm watching this closely. The investment bank’s estimate has become a reference point for analysts trying to gauge the macroeconomic impact of AI. The figure covers spending on computing power, data centers, and the energy and water infrastructure that supports them. Those are not soft costs. They are concrete, physical investments that show up in GDP accounts and inflation statistics.
The productivity question is not abstract. It will be measured in real output per hour worked, in the prices of goods and services, and in the interest rates set by the Federal Reserve. If AI makes the economy more efficient, the Fed can keep rates lower. If it just adds cost, the Fed will have to keep rates higher to contain inflation.
Right now, the evidence is mixed. The spending is surging. The productivity data has not yet caught up. That is the tension at the heart of the story. The AI boom is real. The money is real. The question is whether the returns will be real too.
