The AI investment surge is getting even bigger

• Alphabet, Amazon, Meta and Microsoft are together anticipating capital spending of around $650 billion this year, per Bloomberg's calculations, up from "only" $359 billion in 2025. A decade ago, that number was only $31 billion.
• These numbers are staggering — and don't even include the many other companies spending big on data centers, semiconductors, software development and the electricity infrastructure to power it all.

State of play: For context, total U.S. nonresidential investment spending — the money spent on every office building, factory, machinery and software deployment — was running at a $4.3 trillion annual rate in the most recent reading.

• So a small handful of hyperscalers are likely to account for a massive share of all investment this year.

Between the lines: As a general rule, it's a bad idea to try to map the actions of any one company onto overall macroeconomic variables like GDP. The U.S. economy is just too big and sprawling for even the largest companies' actions to move the dial much.

• For example, Walmart may be the biggest retailer by far, but its U.S. sales equate to only about 2% of total personal consumption expenditures.
• The AI boom is looking like an exception, with a handful of companies deploying enough cash to move the dial on GDP, even beyond their major contribution in 2025.

What they're saying: "Capex intentions by the large technology firms exceeded expectations for 2025 by an unusual order of magnitude," RSM chief economist Joe Brusuelas tells Axios.

• "Should reality match planned outlays, this strongly implies upside risk to the consensus 2.2% estimate of GDP for the upcoming year."

Reality check: It's not a simple matter of saying that every billion dollars a mega-cap tech company spends translates into an extra billion dollars of GDP. The effects are more complex.

• The spending may be displacing other potential investments, both by pushing up borrowing costs and by claiming finite physical resources. A piece of land where a new data center is going up might instead be the site of a new warehouse, absent AI spending.

• To the degree the investment relies on imported equipment, such as semiconductors from Taiwan, it creates an offsetting subtraction from growth. GDP aims to capture domestic output, so spending on imported goods is subtracted to make the arithmetic work.
• Moreover, it is uncertain how much the AI investment spend will mean for job creation, as it is a highly capital-intensive pursuit but not very labor-intensive.

The intrigue: Both AI investment itself and the returns it aims to generate across the economy imply a significant boost to productivity in the years ahead.

• Indeed, that was already evident last year, given a mix of weak job growth and elevated GDP growth.
• "Should productivity-enhancing investment arrive close to planned capex outlays," Brusuelas says, "that would imply another year of robust increases in productivity."
• Kevin Warsh, President Trump's nominee to lead the Federal Reserve, is betting on exactly that as a justification for lower interest rates.