Tech stocks have suffered a violent selloff, but whether this is a leverage unwind or a disproof of business models will determine completely opposite trading directions.

According to Goldman Sachs' FICC and equities trading desk in a November 6 report, the Nasdaq fell 5.5% in a single day, the Philadelphia Semiconductor Index plunged 12%, and the S&P 500 dropped 4.4%. Goldman Sachs cash equities business data showed that US cash equities saw US$3 billion in net selling that day, with the information technology and communication services sectors accounting for US$2.2 billion. Cumulative net selling in the tech sector has now expanded to US$27 billion, exceeding the US$21 billion technical selloff in July. At the same time, the VVIX posted its largest single-day move since the "Volmageddon" episode on February 5, 2018.

Rich Privorotsky, head of Goldman Sachs' FICC and equities business, made clear in the above daily update that he leans toward characterizing this decline as a 1998-style liquidity crisis rather than a precursor to a 2000-style tech bubble burst. His core judgment is that the AI growth engine itself may remain intact, and what the market is violently repricing is the financing and timing of buildout, not underlying demand. This judgment is crucial for investors' positioning implications — if the analogy holds, such episodes in history have ultimately turned into buying opportunities.

Leverage self-reinforcement, not valuation reset

The mechanical characteristics of this selloff point to forced deleveraging rather than fundamentals-driven valuation resets.

Privorotsky noted that forced liquidation of leveraged ETFs before the close amplified declines in semiconductors and the Nasdaq, compounded by market makers' negative gamma exposure forming a self-reinforcing liquidation loop. From a fund flow perspective, Goldman Sachs prime brokerage data showed that the core pressure in the semiconductor sector lies in the absence of long-only institutional (LO) base demand, with multi-strategy funds, quant funds, long-only institutions and retail investors selling in the same direction — "everyone standing on the same side" is itself the problem.

Correlations rising, momentum factors underperforming, residual leverage factors continuing to sell off, while low-beta, defensive and quality assets are being chased — this combination of readings is highly consistent with typical forced de-risking characteristics. Looking at market evolution in the two days after the election, AI stocks underperformed on the first day while cyclical stocks rebounded on expectations of political centralization. On the second day, there was a sharp catch-down: if data center construction slows substantially, nominal growth expectations will face downward revisions, and the cyclical stock rebound itself faces repricing pressure.

1998 or 2000: two templates, two outcomes

Privorotsky compared this decline side by side with two historical nodes, ultimately leaning toward the 1998 template.

The reasons supporting the 1998 analogy are: that year's Russian default and Long-Term Capital Management (LTCM) crisis were external shocks, but US economic growth itself remained intact; once liquidity recovered and the Fed turned accommodative, the market quickly recovered its losses. Corresponding to the present, the AI growth logic may similarly remain intact, and what the market is repricing is buildout financing and the pace of progress.

Moreover, current tech leaders have real revenue, cash flow and balance sheets, which is fundamentally different from companies during the 2000 bubble period; delaying data center construction does not eliminate computing power demand, and constrained capacity may actually improve the economics of existing assets.

The most similar discomfort to 2000 lies in the massive scale of spending and the uncertainty of returns. The core question is: is the market liquidating a sound growth story, or is it discovering that its economic logic itself is flawed? Privorotsky leans toward the former, but emphasizes that demand, utilization and return data must remain solid.

Four types of AI buildout concerns, the market is focused on the wrong one

Goldman Sachs sorted out four types of market concerns supporting this selloff: first, capital is insufficient to support the scale of buildout; second, revenue is insufficient to justify investment; third, power and permitting capacity is insufficient to support computing power expansion; fourth, the overall spending trajectory is unsustainable.

Privorotsky believes the market is currently most focused on the third — power and permitting issues — but this is precisely the least important of the four concerns. Constrained supply does not equal disappearing demand. The question that truly needs answering is: can the revenue side ultimately justify the capital invested?

On political variables, Texas Governor-elect Gina Hinojosa's remarks were more moderate than the market interpreted. She clearly stated: "We can achieve economic development while protecting Texans, and these two things are not mutually exclusive," and reiterated "I don't hate data centers." Privorotsky believes this is more of a regulatory reset than an ideological rejection, and actual spending may turn out better than what current market discounts imply.

Technical levels and prerequisites for buying

From a technical perspective, the Nasdaq has broken below its 200-day moving average, while the S&P 500 is just holding above that average, and whether this support level holds is crucial. The negative gamma environment means the market has room for violent swings in either direction.

On quality preference, Privorotsky explicitly named Taiwan Semiconductor Manufacturing Co Ltd (NYSE: TSM) and ASML Holding NV (NASDAQ: ASML), arguing that under current uncertainty, exposure should tilt toward high-quality tech assets.

He also listed three conditions needed to change the market narrative: AI investment forming a positive political feedback loop, financing markets stabilizing, and empirical evidence that underlying spending and demand remain intact. The disproof conditions are equally clear: if AI capital expenditure is substantially delayed and the revenue side disproves the investment logic, the 1998 analogy fails and a shift to the 2000 template is needed; if the financing market is the first to experience refinancing failures, the nature of the decline will fundamentally change.

From a longer cycle perspective, Privorotsky cited historical data noting that in the 12 months following all 19 midterm elections since 1950, the S&P 500 recorded gains every time — but he emphasized this is a 12-month observation, not a one-month prediction, and short-term timing remains difficult. On energy, oil prices rebounded about 4% in a single day, and the market had previously been overly optimistic in pricing a rapid easing of Iran tensions. Negotiations remain unresolved, and the complicating impact of energy shocks on rates, growth expectations and risk appetite remains an important variable in the stabilization process.