AI spending added 0.9 points to third-quarter US GDP growth

Artificial intelligence is no longer a rounding error in the US economy. AI spending added 0.9 percentage points to annualized real GDP growth in the third quarter of 2026, according to figures highlighted in Morgan Stanley’s analysis.

Two years ago, that same contribution was 0.2 percentage points.

The numbers behind the boom

The trajectory is steep. AI investment contributed 0.2 percentage points to growth in 2024, then 0.5 points in 2025, before hitting 0.9 points in the third quarter of 2026.

The broader economy had a strong quarter, too. US real GDP grew at a 3.4% annualized rate in the third quarter, up from 2.2% in the second quarter.

Do the arithmetic and AI accounts for a little more than a quarter of that 3.4% headline pace.

Business investment did much of the heavy lifting. It jumped 12.7% on an annualized basis in the quarter, contributing 1.78 percentage points to overall growth, with data centers and AI-related infrastructure leading the surge.

The detail worth underlining: more than three-quarters of AI’s contribution to the GDP increase came from infrastructure, not applications.

Morgan Stanley’s trillion-dollar framing

Morgan Stanley estimates hyperscaler capital expenditures at $800–900 billion in 2026. The bank expects that figure to climb past $1 trillion in 2027.

Hyperscalers are the giant cloud operators, companies such as Microsoft and Google, that rent out computing power at enormous scale.

Morgan Stanley CEO Ted Pick has described AI as a multi-trillion-dollar economic cycle. He has put AI-related spending at around $1 trillion per year and argued the buildout is still early.

The industry is in the “second or third inning,” Pick has said.

In Pick’s view, total investment over the full cycle could land between $10 trillion and $15 trillion across the next 15 to 30 years.

Why the infrastructure tilt matters

The composition of the spending tells you who is actually getting paid. A boom led by infrastructure means the money is flowing into physical assets: data centers, chips, networking gear and the energy systems that keep them running.

That is a different economy from one where AI software is quietly lifting productivity across every office. The first shows up quickly in GDP because building things counts as investment. The second is harder to measure and slower to arrive.

Right now, the data says the US is firmly in the building phase.

This is also why the numbers can move so fast. A handful of very large companies writing very large checks can shift a national growth figure in a single quarter, which makes the contribution impressive and concentrated in roughly equal measure.

What this means

For investors, Morgan Stanley’s projections point toward continued spending on technology and infrastructure, with hyperscalers and the firms that supply them sitting closest to the cash.

The risks are not hard to spot, though. The research flags concerns over rising Treasury yields and questions about whether the productivity gains will prove durable. Higher yields raise the cost of financing massive projects.

There is also a concentration problem. When more than a quarter of headline growth leans on one theme, any slowdown in hyperscaler budgets would not stay contained to tech. It would show up in GDP prints, in business investment data and in the broader read on the US economy.

The things to watch are fairly concrete. First, whether hyperscaler capex actually lands in Morgan Stanley’s $800–900 billion range for 2026 and pushes past $1 trillion in 2027 as the bank expects. Second, whether the AI share of growth keeps rising from 0.9 points or plateaus as the infrastructure phase matures.

Third, and most important over the long run, whether applications start contributing a bigger slice. Infrastructure spending is the down payment. The economy eventually needs the software layer to generate returns that make the mortgage worth it.