Goldman Sachs Warns: The First Tech Giant to Cut AI Spending Will Trigger a Market-Wide Repricing
Claire Weston
Goldman strategist Rich Privorotsky warns that AI capex is priced on a single assumption — spending only goes up. The moment any major player cuts back, the entire AI value chain faces a full repricing.
What bet is the market making right now?
AI capex pricing rests on one premise: inference demand grows, spending never falls.
This means → the market has priced in zero probability of even a slight cut.
In plain terms = everyone is betting the same direction with no hedge. If that direction reverses, the sell-off will be far larger than a straight-line model would suggest.
What signals are already flashing?
Privorotsky compares the current tension to a stretched rubber band: for weeks, the market has ignored virtually every negative signal in the AI capex trade.
A structural divergence is widening — Amazon, Alphabet, and Meta keep raising spending commitments, yet their stocks are underperforming the broader market. Meanwhile Nvidia and TSMC, the hardware sellers, are rallying.
This reflects a market already voting with its feet: money is flooding toward the shovel-sellers while shying away from the shovel-buyers — the spenders and the earners are being priced in opposite directions.
Why are low-cost Eastern models making the West nervous?
China's GLM-5.2 large language model was trained entirely on Huawei processors — 100,000 Huawei chips, with zero Nvidia involvement.
This means → frontier AI capabilities can be built in the East at a fraction of Western cost.
In plain terms = if someone else can build comparable AI for far less money, the massive spending Western giants are committing right now risks becoming over-investment. Privorotsky puts it bluntly: "The biggest capital allocators are also the most exposed to over-investment risk."
How real is the market-level pressure?
South Korea's Kospi index plunged roughly 10% the day after hitting a closing high. Samsung and SK Hynix each fell over 12%.
Nasdaq futures dropped about 2.5%; Micron slid more than 7% pre-market; Intel fell 6.5%.
This reflects extreme sensitivity in the semiconductor sector to any wobble in demand expectations — no actual spending cut is needed; a mere shift in outlook is already enough to produce sharp losses.
What other pressures does JPMorgan see?
JPMorgan strategists flagged multiple headwinds weighing on tech: overstretched expectations, euphoric sentiment, thin free cash flow, CTOs pushing back on soaring token costs, a shift toward cheaper models, tough restrictions from Washington, and rising debt and equity supply.
Progress in orchestration — coordinating multiple AI models — model fusion, and quantization — compressing models to cut running costs — also warrants attention.
This means → efficiency keeps improving, and the flip side of that improvement is clear: vendors' pricing power could erode — if you need less compute, the people selling compute lose bargaining leverage.
How does Goldman think the "breaking point" arrives?
Goldman believes the breaking point in the AI capex cycle will likely come from a rational awakening at one core spender — the moment it realizes returning cash to shareholders beats pouring more into AI.
The first giant to hit the brakes becomes the spark that triggers a market-wide repricing.
In plain terms = from chipmakers to cloud platforms, the entire AI value chain is priced on a shared assumption that everyone keeps charging forward. The moment one player stops, that consensus shatters.
Content is for reference only, not financial advice.