Goldman Sachs derivatives traders warn the era of uniform AI stock moves is ending, with the market entering a more selective phase of theme differentiation. Meanwhile, the shift in Federal Reserve expectations from "fear" to "complacency" has left broad market index risk-reward profiles looking lackluster.

Shawn Tuteja, head of ETF and custom basket volatility trading at Goldman Sachs, noted in a recent market assessment that the high correlation among AI trades is breaking down. While AI sub-themes moved almost in lockstep during the April-to-June rally and the July sell-off, clear divergence has emerged over the past two weeks. Optical networking, data centers, and emerging neocloud providers are leading the rebound, while memory and AI power themes lag significantly.

Tuteja also warned that market sentiment toward the Fed has shifted from "fear of rate hikes" in July to an almost "rosy" optimism that "nothing can go wrong." This shift has eroded the asymmetric opportunities that previously existed. He maintains his S&P 500 year-end target of 8,000 but expects the index to trade in a relatively narrow range over the next six weeks, making overall risk-reward less attractive. In this context, he advises a more selective approach to AI stocks rather than broad, high-beta exposure.

Strong Earnings Season Fails to Lift Stocks as Market Turns Range-Bound

On the surface, this earnings season has been robust. Tuteja notes that 90% of S&P 500 companies have reported second-quarter results, with 64% beating consensus earnings per share estimates by at least one standard deviation. This is among the highest levels in history and has driven a 2% upward revision to 2027 earnings forecasts.

However, the market's reaction to these beats has been unusually muted. Stocks that beat estimates by one standard deviation or more posted a median excess return of just 33 basis points relative to the S&P 500 the next day, far below the historical median of 95 basis points since 2010. The tech sector performed even worse, with beating stocks underperforming the S&P 500 by an average of 99 basis points the following day.

Tuteja attributes this to a structural shift in market sentiment. In July, clients broadly reduced risk exposure, with global fundamental long-short hedge funds experiencing one of their worst monthly drawdowns in four years. Currently, net client positions are at the 67th percentile of their five-year history, while gross positions are at the 89th percentile, indicating the market is no longer under-positioned.

Last Tuesday, the daily volume of S&P 500 call options reached a record 4 million contracts, and the short-term put/call skew saw its largest two-day decline in nearly a decade. These signals point to a rapid warming of sentiment.

Fed Sentiment Shifts from "Fear" to "Complacency," Boosting Hedge Appeal

Tuteja believes current market pricing of the September Federal Reserve meeting reflects a "win-win" optimism. Either the Fed stabilizes the long-end yield curve with a "dovish rate hike," or it holds steady, allowing the "rotation and broadening" trade to continue amid strong earnings. Both scenarios imply a significant decline in oil prices.

Ahead of the CPI release, buy-side institutions' consensus estimate for core inflation was around 21 to 22 basis points, with the probability of a September rate hike at roughly 50/50. Fund flows suggest markets are more heavily positioned for a dovish outcome than hedged against a hawkish surprise.

Tuteja notes that this shift from "fear" to "complacency" is the core reason for the deteriorating risk-reward profile of broad market indices.

He suggests that if the market rallies after CPI, investors could consider buying put spreads on the S&P 500 excluding AI stocks (SPXXAI) as a hedge against macro uncertainty ahead of the September meeting. His recommended structure is a SPXXAI September 90/97% put spread, with an indicative cost of around 0.93% and a 19-day duration.

AI Trading Diverges: Fundamental Disagreements Begin to Drive Pricing

Tuteja attributes the divergence within AI stocks to both fundamental and technical factors.

On the fundamental side, Goldman Sachs tech analyst Peter Callahan points out that while July's sell-off was largely technical in nature, the rise of open-source models and ongoing debate about AI capital expenditure "bottlenecks" has quietly triggered a shift in the fundamental narrative. As 2027 capex trajectories become clearer following second-quarter earnings, investors are more closely scrutinizing the earnings sustainability of each sub-theme.

Callahan highlights two emerging points of debate. First, the "inference economy" concept is gaining traction in the software sector, with stocks like Snowflake Inc (NYSE: SNOW), Datadog Inc (NASDAQ: DDOG), Palantir Technologies Inc (NYSE: PLTR), CrowdStrike Holdings Inc (NASDAQ: CRWD), Palo Alto Networks Inc (NASDAQ: PANW), Okta Inc (NASDAQ: OKTA), and Twilio Inc (NYSE: TWLO)—which are less sensitive to large model and compute bottlenecks—attracting more interest. Second, the memory (DRAM, NAND, HDD) sector's investment logic is shifting from "earnings upgrades driven by average selling prices and margins" to "valuation expansion driven by stability, long-term contracts, and capital returns," sparking increased controversy.

Additionally, the power theme faces extra pressure from policy uncertainty ahead of midterm elections and regulatory issues at ERCOT.

Leveraged ETF Redemptions and Elevated Implied Volatility Hinder AI Rally Momentum

From a technical perspective, Tuteja cites data showing that assets under management (AUM) in US semiconductor leveraged ETFs are currently around $99 billion, down sharply from a peak of $157 billion.

Notably, during the July sell-off, the market saw about $15 billion in "excess" buying—where investors added to leveraged ETF positions relative to the drop in spot prices. However, in August, the only source of AUM growth has been the rebound in net asset values as spot prices recovered. In reality, investors are using the rally to redeem shares and reduce exposure.

The implied volatility market reflects that clients are taking a wait-and-see approach to re-entering these stocks. This is likely because position structures have become healthier and overly bullish positions in leveraged ETFs have decreased.

Since July 29, the weighted average implied volatility of the Goldman Sachs AI Leaders Basket (GSTMTAIP) has fallen over 10 volatility points, a decline of 14.3%. In comparison, the implied volatility of SPXXAI has fallen only 3 volatility points, a decline of 11.8%.

Despite this, the premium of AI sector implied volatility over the broader market remains at its highest level since early 2023 (excluding the past two months). This makes expressing bullish views or buying upside protection through options still costly, creating a barrier for clients to re-enter.

Tuteja advises that at current implied volatility levels, buying the stock and overlaying a call ratio spread, or funding a call spread by selling puts, offers better value than broad high-beta exposure. His recommended structure is selling the GSTMTMEM December 31 85% put and buying the 115/150% call spread, with an indicative net cost of about 1.45% and a 50-day duration.

Bull Case Remains, but High-Speed Rally Phase May Be Over

Overall, Tuteja's core judgment is not bearish, but rather a call for lower expectations on the pace of gains.

He maintains his S&P 500 year-end target of 8,000 but believes the high-speed, high-beta phase of the AI trade is largely over. Optical networking, data centers, and emerging neocloud providers remain attractive, but memory and power themes face more fundamental questions and require more careful selection.

On a broader level, as the market shifts from "fearing the Fed" to being "complacent about the Fed," the asymmetric opportunities that once existed have narrowed significantly. Investors need to reassess their AI exposure under a framework of more refined theme selection and more disciplined risk management.