【TMGM Financial Recap】Goldman Sachs Reviews the November 5 Tech Sell-Off: A 1998-Style Liquidity Flush, Rather Than a 2000-Style Bubble Burst

The November 5 sell-off was extreme by the numbers. The Nasdaq 100 closed down 6.4%, wiping out approximately $2.3 trillion in market value in a single day and recording its largest daily decline since March 2020; the Philadelphia Semiconductor Index plunged 12.25%; and the VIX surged 70% to its highest level since early April. Nvidia fell 9.26%, bringing its decline from its October high to 24.1% and officially entering a technical bear market. But Goldman Sachs’ review went beyond the scale of the losses to address a question with direct implications for investment decisions: was this a liquidity-driven liquidation within an otherwise sound growth story, or the beginning of a discovery that the underlying economic logic was flawed? Goldman Sachs’ answer was the former.

A Repeat of 1998 or 2000?

Goldman Sachs compared the current decline with two historical episodes and ultimately selected 1998 as the more appropriate template.

Three arguments support the comparison with 1998. First, the shock originated externally and was not driven by fundamentals. In 1998, the triggers were Russia’s default and the collapse of LTCM, while US economic growth remained intact. Today, uncertainty over data centre approvals following the Democrats’ sweep of both chambers of Congress represents a political shock, rather than the disappearance of AI demand itself. Second, today’s technology leaders have real revenues, cash flows and solid balance sheets, fundamentally distinguishing them from the companies without earnings support in 2000. Third, delaying data centre construction does not eliminate demand for computing power; constrained capacity could actually improve the economics of existing assets.

Goldman Sachs also acknowledged that the most unsettling similarities with 2000 are the enormous scale of spending and the uncertainty surrounding returns. Nevertheless, it leans towards the view that the market is liquidating positions in a sound growth story encountering friction over financing schedules and construction timelines, rather than exposing flaws in the story’s underlying economics.

The investment implication is straightforward: if the 1998 comparison holds, similar liquidity shocks have historically developed into buying opportunities. However, Goldman Sachs added a crucial qualification elsewhere: the rapid market recovery in 1998 depended on Federal Reserve interest rate cuts, while today’s market has no such support. This is the most substantial difference between the 1998 template and the current situation.

Goldman Sachs also believes that forced liquidation by leveraged ETFs ahead of the close amplified declines in semiconductors and the Nasdaq, while dealers’ negative gamma exposure created a self-reinforcing liquidation cycle. Nomura’s Charlie McElligott provided more specific estimates: approximately $47 billion in leveraged ETF rebalancing sell orders ahead of the close, around $61 billion from volatility control funds, and approximately $7 billion in additional selling pressure from dealers’ negative gamma exposure for every 1% market decline. Together, mechanical selling across the market easily exceeded $100 billion.

But what truly turned this liquidation from a “technical correction” into a “stampede” was the positioning structure revealed by Goldman Sachs’ prime brokerage data: underlying demand from long-only institutions was absent, while multi-strategy funds, quantitative funds, long-only institutions and retail investors were all selling in the same direction. “Everyone being on the same side” was itself the problem.

John Flood of Goldman Sachs’ trading desk confirmed the same pattern from an execution perspective. November 4, the day after the election, was one of Goldman Sachs’ busiest trading sessions of the year. Long-only investors had reduced their risk exposure before the election, were caught off guard by Wednesday’s rally and were forced to chase prices higher to rebuild positions. On November 5, those investors joined hedge funds in selling the hyperscaler stocks they had bought just a day earlier. Brokers with a high proportion of retail clients reported a significant imbalance towards selling for a second consecutive session. Goldman Sachs saw “almost no real buying”. Rising correlations, momentum factor underperformance, continued selling in the residual leverage factor, and demand for low-beta and defensive assets formed a combination closely aligned with typical forced de-risking, rather than a fundamentals-driven valuation reassessment.

“Front-running” further amplified the decline. At midday, Morgan Stanley estimated approximately $52 billion in mechanical sell orders ahead of the close. As this information spread rapidly across Wall Street, discretionary traders chose to sell early and wait to buy back at the “usual low” between 2 pm and 3 pm. But that low never materialised. The front-runners themselves became a force driving prices lower, and the market closed at its lowest level of the day.

How Goldman Sachs Views the Political Variable

Goldman Sachs’ treatment of the political shock deserves attention. It did not interpret the Democrats’ victory as an ideological rejection of AI. Goldman Sachs specifically cited comments from Texas Governor-elect Gina Hinojosa: “We can achieve economic development while protecting Texans. The two are not mutually exclusive,” alongside her explicit statement that “I do not hate data centres.” Goldman Sachs views this as more of a regulatory reset than an ideological rejection, suggesting that actual spending could exceed the expectations implied by current market discounts.

At the same time, Goldman Sachs acknowledged that political developments have changed the parameters of market pricing. Its trading desk explicitly noted that the debate surrounding the AI trade has shifted from “long-term demand for AI computing power” to “the speed and cost of deployment”. Higher financing costs, political pressure on data centres and post-election regulatory concerns have jointly weighed on sentiment. On the day of the sell-off, Goldman Sachs warned that the Democrats’ “blue tsunami” could leave approximately 45% of planned US data centre capacity expansion exposed to moratoriums.

This assessment implies that the November 5 sell-off was pricing in “a longer timeline for AI infrastructure construction”, rather than “the disappearance of AI demand”. For assets valued on the basis of future cash flows, a longer timeline itself warrants lower valuations, but it represents a fundamentally different shock from demand falling to zero.

Goldman Sachs also systematically outlined four categories of market concerns underpinning the sell-off: insufficient capital to fund the scale of construction; insufficient revenue to justify the investment; insufficient power and permitting capacity to support computing expansion; and an unsustainable overall spending trajectory.

Its assessment is that the market is currently most focused on the third category, power and permitting constraints, which it considers the least important of the four. The reasoning is that constrained supply does not mean demand has disappeared. The real question is whether revenues can ultimately justify the capital invested.

This assessment shifts the focus of pricing from “political noise” back to “revenue validation”. Political shocks create uncertainty over timelines; whether revenues materialise ultimately determines whether the 1998 comparison holds.

When Would the 1998 Template Break Down?

Goldman Sachs went beyond characterising the sell-off to identify three conditions needed to change the market narrative, alongside two clear developments that would invalidate its assessment. The three conditions for restoring the narrative are: a positive political feedback loop from AI investment, stabilising financing markets, and concrete evidence that underlying spending and demand remain intact.

Two developments would invalidate the 1998 comparison: if AI capital expenditure is materially delayed and revenue outcomes disprove the investment thesis, the appropriate template would shift towards 2000; if refinancing failures emerge first in funding markets, the nature of the downturn would fundamentally change.

At the trading level, Goldman Sachs’ desk had already observed early signs of dip-buying on the day of the sell-off, confirming investor interest in pullbacks in Nvidia, TSMC, Microsoft and CoreWeave. At the same time, it emphasised that “it is difficult to know when selling pressure will ease, and the rebound when it does could be sharp”.

Goldman Sachs’ review offers a framework distinct from the “AI bubble” argument: the core driver of the November 5 plunge was a liquidation of positions jointly triggered by forced selling from leveraged ETFs, negative gamma feedback loops and the absence of demand from long-only institutions, rather than a sudden disappearance of underlying AI demand. The political shock, uncertainty over data centre approvals following the Democrats’ congressional sweep, provided the catalyst. Its effect, however, was to extend construction timelines rather than eliminate demand itself.

The explanatory strength of this framework is that it accommodates two seemingly contradictory facts: the severity of the declines, 12% in semiconductors and 6.4% in the Nasdaq 100, did reach crisis levels; yet the cumulative $27 billion in net technology selling recorded by Goldman Sachs’ cash equities desk, together with positioning in which “everyone was on the same side”, points to the mechanics of a liquidity event rather than a fundamental one.

But this framework also has a clear limitation. A central implicit assumption behind the 1998 comparison is that “a backstop will arrive”, which in that year took the form of Federal Reserve rate cuts. Goldman Sachs acknowledges that this backstop is absent today. This means that even if its assessment is correct, the speed and path of the market’s recovery from the liquidity shock will not simply replicate the 1998 playbook. Goldman Sachs has clearly outlined the conditions for recovery: positive political feedback, stabilising financing markets and concrete evidence from revenues.