Goldman Sachs: AI Stocks Oversold, Democratic Sweep May Actually Accelerate Capex
nashnova research
AI infrastructure stocks have fallen 50%-60% from their peaks. Goldman's head of European equity trading argues the sell-off is overdone and offers a counterintuitive case — tighter regulation could raise the value of existing compute and pull capital deployment forward.
What triggered this sell-off?
Democrats won both chambers in the midterms. In Texas, Democratic candidate Gina Hinojosa defeated incumbent Republican Governor Greg Abbott — and she backs a moratorium on AI data-center construction.
This means → Texas accounts for 27% of planned U.S. data-center power capacity (Wells Fargo estimate). A moratorium would choke the country's largest build-out pipeline.
Pre-market reaction was immediate: CoreWeave fell ~3.3%, Applied Digital ~3%, IREN 4.3%.
Why does Goldman say the selling went too far?
Goldman's head of European equity trading, Rich Privorotsky, wrote in a client memo that AI stocks have been "sold to death," yet fundamentals remain solid, election uncertainty is gone, and positioning is clean.
Two supporting points: valuations are deeply compressed, so shorts must take profits — that alone creates rebound pressure. And AMD's Q4 guidance showed a 27% sequential re-acceleration, proving the AI capex cycle has not broken.
In plain terms = prices crashed, but companies are still pouring real money into AI. The sellers' thesis and the buyers' dollars don't match.
How can tighter regulation be bullish?
Privorotsky argues that if future data-center expansion is constrained, already-deployed compute becomes scarcer — and more valuable.
This means → the theoretical winners are hyperscalers that already run massive fleets — Amazon, Microsoft, and Alphabet.
In plain terms = building restrictions act as a moat for incumbents. Compute shifts from "anyone can build" to "some can't."
Why might capital spending get pulled forward?
Privorotsky expects corporate decision-makers to move deployments into 2027, locking in permits and deal windows before the regulatory framework shifts.
His reasoning: a new presidential term in 2029 could bring tax changes, heavier administrative oversight, and tougher M&A reviews — giving hyperscalers an incentive to "move faster now."
This reflects a game-theory stance — spend while the window is open rather than wait for policy to close it.
What about picks-and-shovels stocks and emerging cloud players?
Data-center delays could hit upstream suppliers — chipmakers, cooling and electrical-equipment makers, energy providers. But Privorotsky notes the Philadelphia Semiconductor ETF has already dropped ~20% in the past month; some of the bad news may be priced in.
For emerging cloud players like CoreWeave and Nebius, he says the picture is more nuanced. The key variable is whether their compute capacity is already under contract or physically deployed — contracted capacity is safer; uncommitted capacity carries more risk.
In plain terms = upstream has already taken a hit, so the worst may be in the price. Newer players have to be evaluated contract by contract — no blanket call works.
Could a VIX drop catalyze a broader rebound?
Privorotsky argues that with election uncertainty removed, the current VIX level is hard to sustain and volatility may "drop quickly."
This means → VIX moves inversely to equities. A volatility decline typically lifts indices — and AI and AI-adjacent stocks make up nearly half of major benchmarks, so they stand to benefit directly.
The ultimate test still rests on two variables: whether Democrats' regulatory intent actually becomes legislation, and whether hyperscalers truly choose to deploy capital early — these are the checkpoints that will validate or break Goldman's call.
市场有风险，内容仅供研究参考，不构成投资建议。