Goldman Sachs puts hyperscaler AI spending at $1.2 trillion in 2027

Goldman Sachs projects that hyperscaler AI capital spending will reach $1.2 trillion in 2027, above consensus², with the top five US hyperscalers named as the source of the increase¹. The bank sees spending rising further to $1.4 trillion in 2028 and frames the cycle as a $7.6 trillion spending boom⁶. Growth in the capex line cools after 2026, and financing pressures are part of the reason¹.

Alongside the spending forecast the bank published a break-even test. Goldman estimates hyperscalers need some $300 billion in annual AI revenues to break even on their 2026 and 2027 capex³, a figure a second account reports as a $300 billion annual revenue requirement². A separate summary of the same work describes Goldman calculating the revenue needed to justify $1.7 trillion of hyperscaler capex⁵.

Why it matters to investors

The hyperscalers carry the largest weights in the US equity indices, so the gap between the spending number and the revenue number is a market-level question rather than a sector one. Goldman’s own framing is that the reset depends on revenue catching up with capex³.

The bank sets out the bull path explicitly. “Visibility into improving FCF as revenues accelerate and capex decelerates should be supportive for the stocks,” the report said, in a scenario under which hyperscaler stocks rebound⁴. That sequence requires two things to happen at once: revenue acceleration and a slowdown in spending, which is consistent with the cooling growth rate after 2026¹.

The credit side is being positioned differently. Goldman Sachs Asset Management is underweight the largest AI borrowers on expectations of heavy hyperscaler bond supply⁷, which is a bet that the financing of this capex, rather than the capex itself, is where the strain shows.

What to watch

Free cash flow, not announcements. The rebound case rests on capex decelerating while revenues accelerate⁴, so the quarterly disclosures that matter are the ones showing spending growth rolling over¹ rather than new commitments.

The other measure is progress against the break-even bar. Goldman’s $300 billion annual revenue figure is a stated threshold^2,3, which makes reported AI revenue the number to track against it, and the bond supply that funds the gap is the signal to watch in credit^5,7.