’Our phones are exploding’: Goldman Sachs dissects historic Wall Street crash

The NASDAQ plunged 5.5% into correction territory on Thursday, while the S&P 500 fell 4.3% and the Dow dropped 3.4%, erasing Wednesday’s gains; all three had their worst intraday performances since April 4, 2025. Goldman Sachs’ trading desk attributed the sell-off primarily to macro-product selling and rapid reversals in investor positioning, with the tech move roughly five times the implied move in Wednesday’s one-day straddle. Investors are reassessing AI deployment pace and costs amid rising financing costs, data-center political pressure and regulatory fears; tentative dip-buying in Nvidia, TSMC, Microsoft and CoreWeave emerged late in the session.

Analysis

The key distinction is between a flow-driven air pocket and a change in the AI earnings path. Macro selling and rapid position reversals can push correlated names below fundamentals, creating rebound potential—but they also make intraday price action a poor signal of where durable demand sits. The more consequential shift is toward deployment economics: higher funding costs, permitting or political friction, and uncertain utilization can delay customer capex even if long-run compute demand remains intact. That moves valuation risk from chip demand alone to the return on data-center investment and the timing of orders.

Over the next few sessions, forced selling and dealer/volatility positioning can amplify moves in both directions; a violent rebound would not, by itself, disprove the repricing. Over 1–3 months, watch hyperscaler capex commentary, data-center approvals and financing conditions. Over 6–18 months, slower deployment would spill from infrastructure operators to chip suppliers through order timing and inventory, while cash-rich buyers may gain bargaining power. The contrarian opportunity is that broad macro liquidation can indiscriminately punish higher-quality exposures; the counterpoint is that buying the first bounce risks confusing short-covering with renewed fundamental conviction. A key missing input is whether capex plans and committed power capacity are actually being cut or merely delayed.

Key Decisions for Investors

• Keep broad AI beta hedged near term rather than chase post-selloff volatility: use Nasdaq index futures to trim exposure if selling resumes, and unwind the hedge only after breadth and intraday volatility stabilize. Risk is a sharp short-covering rebound; thesis is falsified by sustained stabilization alongside improving participation, not a single late-session bounce.
• Stage into Nvidia and TSMC only after evidence of stabilization, preferably in tranches rather than a one-day dip buy. The upside is recovery from flow-driven dislocation; the downside is order deferral if hyperscaler capex plans weaken. Reassess on supplier order commentary and inventory trends.
• Consider a conditional relative-value pair: long cash-generative, established AI platform exposure such as Microsoft against short more financing-sensitive AI infrastructure exposure such as CoreWeave, but only after verifying debt terms, funding access, customer concentration, and committed capacity. This targets a widening deployment-economics gap; avoid the trade if financing exposure is not materially different or if infrastructure demand indicators reaccelerate.
• Do not chase long volatility after an outsized realized move. Prefer defined-risk downside hedges only if implied volatility and skew normalize while policy, rates, or capex risks remain unresolved; otherwise premium decay can outweigh protection value. Revisit if volatility reprices higher again or hyperscaler guidance confirms a genuine capex slowdown.