Beneath the seemingly calm surface of U.S. equities, a violent deleveraging process is unfolding behind the scenes.

Rich Privorotsky, Goldman Sachs' Head of Delta-One Trading, described recent market action as "extremely violent" in his latest note. During this storm, systematic funds have suffered significant losses, and the core AI investment narrative that has supported the bull market over the past two years is now facing fundamental disruption and reconstruction.

Privorotsky argues that while the long-term adoption outlook for AI remains positive due to rapidly improving model economics and sharply compressed inference costs, the moats for compute sellers, model access providers, and parts of the upstream "picks and shovels" supply chain are becoming increasingly unsustainable, creating persistent de-rating pressure across the sector.

Systematic Funds Hit Hard as Market Correlations Break Down

Data suggests the actual damage from this correction is far more severe than what major index levels indicate.

According to Marco Laicini, a trader in Goldman Sachs' private wealth business, systematic long/short managers fell 1.4% in a single day, marking their worst daily performance in over two years and exceeding three standard deviations of daily volatility seen over the past three years.

Specifically, the global momentum factor plunged 4.7 standard deviations in one day, while Goldman's U.S. momentum factor dropped nearly 7% in just 48 hours, with losses quickly spreading across major global markets.

On Thursday, all three major U.S. indices closed lower. The Dow fell 703.84 points, and the Nasdaq 100 suffered a fifth consecutive day of losses, down 0.72%. The VIX rose 7.52% to 16.01.

Privorotsky expressed surprise that such a violent shakeout occurred after the significant reduction in positions seen in July. He noted that the current two-way volatility is largely a classic "liquidity trap."

In a typical low-liquidity August environment, leveraged capital that had been making quick profits blindly rushed back into semiconductors and momentum sectors, pushing overall leverage levels in the system to new highs. With this positioning, the sharp decline in sectors like semiconductors directly triggered passive selling by quantitative funds.

Investors should therefore avoid over-interpreting intraday price movements. The current volatility is more a result of capital stampede than a broad fundamental deterioration.

AI Logic Under Reconstruction: Shovel Sellers Face Pressure

However, beyond the capital games, the fundamental logic of the AI sector is indeed at a crossroads of reassessment.

Privorotsky believes the core debate has shifted. The market is no longer questioning whether AI demand is real, but rather whether future growth can continue to match already extremely high expectations.

A key marginal change is underway: the operating economics of large models are improving at an astonishing pace. Models like Qwen are increasingly delivering stronger usable intelligence at smaller sizes and lower costs.

The "usable intelligence output per dollar" is growing explosively, and competition is multiplying. This trend is strongly positive for overall AI adoption and the macro economy, but it sounds an alarm for certain vested interests in the supply chain.

Privorotsky explicitly stated that as the benefits of compressed inference costs are released, companies selling compute access and model API services will find their "moats" increasingly difficult to maintain. Similarly, parts of the upstream "picks and shovels" supply chain face similar re-rating pressures.

The AI sector is likely to experience a strange divergence: while corporate earnings per share (EPS) continue to rise, overall sector valuations are being revised downward. Current valuations falling to the lower end of their range may be reasonable, but with an extended window of uncertainty stretching into 2027-2028, the market will find it difficult to chase highs blindly.

Nvidia's Strong Earnings May Not Spark a Rally

Based on this logic, the upcoming earnings report from Nvidia may struggle to act as a catalyst for a sharp rally.

Privorotsky expects that while the headline numbers will still be impressive, this will likely only serve as a "neutral catalyst."

The market's scrutiny has shifted from the demand side to deeper details, such as funding structures, gross margin changes, and how profits are redistributed across the supply chain.

Memory inflation and supply bottlenecks are becoming key variables affecting profit margins. This means that even if Nvidia signals strong spending, certain segments like the memory supply chain may not react optimistically after digesting the report details.

In contrast, more substantively supportive catalysts may have to wait until the dense industry conference season in September.

U.S. Treasury Buyback Signal: Deeper Implications for Dollar and Gold

Beyond equities and the AI supply chain, another macro-level signal lies hidden in the U.S. Treasury market.

The U.S. Treasury recently announced it will at least double the per-operation cap for long-end liquidity buybacks from $20 billion to $40 billion, covering 10- to 30-year maturities, with operations extending at least through early November.

Privorotsky emphasized the importance of the word "at least," suggesting the policy has no hard ceiling and clearly leaves the door open for further increases if necessary.

Mechanically, this is not quantitative easing (QE), as the Treasury still needs to finance the buybacks and operations do not require full duration neutrality.

But the policy signal is clear: officials are willing to more proactively use their toolkit to stabilize long-end rates. Privorotsky noted that whether the market calls this "quasi-operation twist" or "Operation Twist-lite," the name doesn't matter. What matters is the policy reaction function.

He believes the lasting impact of this policy on the U.S. dollar and gold may be far greater than its effect on Treasury yields themselves.

Because the underlying supply has not disappeared, yields may eventually drift higher again. However, if the market begins to believe that long-end rate pressure will trigger more aggressive buybacks or short-end issuance substitution, then systemic pressure accumulated in the bond market will inevitably migrate to other assets.

The end result will be a flatter yield curve and a weaker dollar. This logic spilling over into broader asset classes would actually constitute a fairly substantial positive for equities and gold.