Goldman Sachs has forecast a 7 percentage point decline in Big Tech return on equity, attributing the compression to AI capital expenditure that has not yet produced measurable returns.1
The bank released its AI spending and profitability analysis alongside the ROE forecast simultaneously.1 That pairing carries weight. Publishing a bearish equity return projection alongside an AI spending report is an explicit institutional signal that the two are connected — and that the connection is damaging.
Big Tech companies have committed enormous sums to AI infrastructure: data centers, custom silicon, and model development at scale. Goldman's analysis suggests those outlays are front-loaded costs, not yet matched by revenue.1 The result is margin compression that flows directly into equity returns.
Institutional investors are now weighing near-term derating of AI infrastructure stocks.1 Premium valuations built on AI growth narratives become harder to sustain when a major bank quantifies the drag on returns at 7 percentage points.
The pattern has precedent. Large infrastructure buildouts — broadband in the late 1990s, cloud computing in the 2010s — preceded returns by years. AI timelines are more compressed, expectations are higher, and scrutiny of the gap between capital deployment and output is intensifying.
What distinguishes the current moment is institutional weight. Analyst skepticism about AI ROI has circulated since 2024. Goldman's forecast adds a specific number — 7 points of ROE decline — to that skepticism and attaches it to Big Tech as a sector.1
For companies that have used AI infrastructure investment to justify premium multiples, the near-term pressure is to show revenue. Data centers and GPU clusters are depreciating assets. Without measurable output growth, they compress margins quarter by quarter.
Goldman's signal points toward derating as the near-term outcome: investors pricing in margin compression rather than discounting future AI revenue.1 That repricing, if it materializes, would mark a shift from the capital deployment phase of the AI cycle to an accountability phase — where spending requires justification in earnings, not just roadmaps.