When companies keep spending regardless of price signals, economists start to get nervous. Morgan Stanley’s latest research poses an uncomfortable question: has the AI investment boom made the US economy structurally less elastic?

The bank projects AI infrastructure capital expenditure will hit roughly $800 billion in 2026, with estimates for 2027 climbing to approximately $1.16 trillion.

Spending that doesn’t flinch

In a May 11, 2026 podcast segment, Morgan Stanley’s Andrew Sheets zeroed in on what he called the “inelasticity” of AI spending. Hyperscale tech firms are pouring money into data centers, chips, and power infrastructure at a pace that doesn’t slow down when input costs rise. Financing gets more expensive? They keep building. Chip prices climb? Still building. Supply chains get disrupted? You guessed it.

Morgan Stanley expects business capital expenditure growth of 7% year-over-year in Q4 2026, accelerating to 8% overall in 2027.

A stabilizer or a blind spot?

Morgan Stanley estimates 3.2% global GDP growth for 2026, and the steady drumbeat of AI infrastructure investment is a major reason why that number holds up.

Data center investments alone contributed roughly 25% to US GDP growth in 2025.

Goldman Sachs offers a more sobering data point. After accounting for imports, AI’s direct contribution to US growth was estimated at only about 0.2 percentage points of the total 2.2% growth recorded in 2025. A lot of the money being spent on AI infrastructure is flowing to overseas chip manufacturers and equipment suppliers, meaning the domestic economic benefit is thinner than the headline spending figures suggest.

From consumer wallets to concrete and silicon

Morgan Stanley’s analysis points to a structural shift from consumer-driven growth to physical infrastructure investment as the primary engine of expansion. A growing share of economic activity is being driven by companies building massive data centers, laying fiber, securing power generation capacity, and stockpiling GPUs.

Morgan Stanley’s outlook contrasts sharply with more optimistic narratives about AI transforming productivity across the entire economy. The bank isn’t dismissing AI’s potential. It’s pointing out that the current phase of the AI buildout looks more like a traditional construction boom, with all the concentration risks that come with it, than the broad-based productivity miracle that Silicon Valley has been promising.

What this means for investors

The investment implications are layered. On one hand, the sheer volume of AI capital expenditure creates a durable demand floor for companies in the semiconductor, power generation, and data center construction supply chains. If spending truly is inelastic, these firms have unusually predictable revenue streams for the next several years.

The reduced economic elasticity also raises questions about monetary policy effectiveness. If the Fed raises rates to combat inflation but AI-driven capex keeps roaring, the central bank may need to be more aggressive than historical models would suggest.