AI Trading Has Become a 'Rubber Band,' Goldman Sachs Top Trader: The Question is 'When Will It Snap'
The Head of EMEA Equity Trading at Goldman Sachs points out that hyperscale cloud providers are over-committing to AI infrastructure, but near-term returns face significant uncertainty. As open-source models close in on closed-source ones, the market value chain is rotating from hardware vendors to "toll booth" platform companies that control distribution and workflows. Computing power may become commoditized in the future, with Q2 earnings reports serving as a key near-term litmus test for the AI return path.
Between the massive investment in AI infrastructure and the uncertainty of returns, a rubber band is being stretched tighter and tighter.
A new analysis from two core trading heads of Goldman Sachs' EMEA equities business indicates that hyperscale cloud providers like Microsoft, Amazon, Alphabet, and Meta are betting on AI infrastructure at a pace that exceeds their own operating cash flow. Whether this unprecedented wave of capital can generate sufficient returns has become the market's most critical unresolved question.
Mark Wilson, Head of EMEA Equity Hedge Fund Business, and Rich Privorotsky, Head of EMEA Equity Flow Intermediation at Goldman Sachs, believe the core contradiction lies in: the historically unprecedented speed of AI build-out versus the high uncertainty of near-term monetization prospects and investment returns.
Privorotsky summarized the current situation: "The AI market has become a 'rubber band'—the question now is how much further it can stretch." He also noted that last week showed the first signs of cracks in the AI economic model: accelerated proliferation of frontier models and a sharp decline in inference costs. Meanwhile, the debt markets for hyperscalers remain under pressure, and if one of them leads in cutting capital expenditure, it could trigger a chain reaction.
Unprecedented Scale of Capex, Return Path Still Unclear
Wilson and Privorotsky point out that hyperscale cloud providers are pouring hundreds of billions of dollars into AI infrastructure, often at a scale exceeding their own operating cash flow, aiming to seize the strategic high ground of the next-generation computing platform—a logic identical to past bets on cloud computing and search.
Optimists believe this is a crucial "option" purchased for the future explosive growth of AI agents, enterprise tools, and new applications; as costs decline and adoption scales, returns will ultimately materialize. Pessimists worry about lagging near-term returns, potential oversupply, rising pricing pressure, and the severe test of whether enterprise monetization can land in time if cheaper alternatives emerge globally—raising market concerns about valuation bubbles and concentrated positioning risks.
According to the Epoch Capability Index cited by Apollo Global Management economist Torsten Slok, which aggregates multiple AI benchmark tests, the gap between the capability level of open-source models and closed-source frontier models has now narrowed to about four months, and the catch-up momentum of other major models cannot be ignored.
Value Chain Shifts Upstream, "Toll Booth" Logic Replaces Hardware Narrative
Privorotsky admits he initially expected token price declines to compress returns across the entire AI supply chain—two competitive API and model providers have already emerged outside cloud infrastructure, sparking aggressive price wars—but the market's current pricing logic puzzles him.
He analyzes two possible market signals: if the market believed "cheaper intelligence will expand the addressable market, drive exponential growth in enterprise adoption, and pull up overall computing demand," hardware stocks should be leading the gains; but the actual price action is not so. He thus infers the market might be expressing another judgment: value is migrating upstream in the supply chain, and platform companies that control customer relationships, distribution channels, and workflows will enjoy a durable economic moat, not hardware suppliers.
He also notes that the current price action largely reflects positioning adjustments and sector rotation, rather than a fundamental shift in the underlying business.
S&P 500 Index "Admirable," Sector Rotation Bears All the Weight
Privorotsky frankly admits he "has to admire" the resilience of the S&P 500: the index not only ignored geopolitical and energy tensions but also remained strong against a backdrop of clear pullbacks in semiconductor and hardware stocks. He said if someone asked him a month ago "could the market rise even if chip stocks fall," his answer would have been "impossible."
The current market characteristic is: correlation breakdown, with sector rotation becoming the main force propping up the index. The S&P 500 ex-AI-related weightings index has already broken out, with funds rotating from hardware suppliers to "toll booth" companies for AI adoption—platforms that control demand, software, cloud infrastructure, and distribution channels—highly consistent with the long-term economic logic of the technology.
Wilson and Privorotsky believe computing power will eventually be commoditized as supply expands, and hardware stock valuation multiples will likely contract before earnings forecasts are revised down, while those who control the demand side are the true destination for long-term economic value.
Second Quarter Earnings to Serve as Key Near-Term Litmus Test
The two trading heads characterize the upcoming Q2 earnings season as a key near-term test for whether the "AI investment return path" can make substantial progress. Recent unwinding of momentum strategies has caused market pain, but they believe the direction of sector leadership change over a longer cycle is correct.
Overall, the Goldman Sachs trading team characterizes this cycle as: high risk and high volatility are inevitable, but if hyperscalers can secure the "operating system" position of the AI ecosystem, they possess solid long-term potential. The tone remains cautious on valuation and timing, while maintaining belief in the upside potential of technological transformation.
Comments