Is the S&P 500 in an “earnings bubble”? Goldman weighs in

Goldman Sachs expects S&P 500 EPS growth to decelerate rather than collapse, forecasting 11% growth in both 2027 and 2028 to $415 and $460, respectively, and a 12-month index return of 14% to 8,700. While S&P 500 earnings rose 51% in Q2 and 26% over the past four quarters, Goldman notes valuations are elevated relative to trend earnings and warns that slowing AI capex, semiconductor price declines, and fading mega-cap equity gains could pressure profits. In an adverse semiconductor-price scenario, S&P 500 earnings could fall by about 10%.

Analysis

The relevant regime shift is from broad AI-capex beta to return-on-invested-capital discrimination. As depreciation catches up with infrastructure spending, hyperscalers will face greater pressure to prove AI revenue conversion; this favors asset-light monetizers such as APP over hardware vendors whose earnings are more exposed to unit pricing, inventory digestion, and customer capex timing. SMCI remains especially vulnerable to a normalization in accelerator availability or server ASPs because working-capital intensity can magnify even a modest demand miss into a cash-flow and multiple event.

The market can sustain a higher index level if earnings breadth expands beyond the AI complex, but that requires cyclicals and financials to replace some of the incremental growth currently concentrated in a narrow capex chain. GS benefits if elevated equity issuance, M&A, and capital-markets activity persist, yet its upside is more dependent on realized deal volumes than on AI spending itself; it is a cleaner beneficiary of a soft-landing/broader-growth outcome than a continuation of hardware euphoria. A semiconductor price reset would likely be transmitted first through supplier inventories and gross-margin guidance, then through enterprise-server and networking orders over the following one to two quarters.

Consensus appears too comfortable treating a deceleration as benign while assigning high multiples to companies with the greatest sensitivity to revised AI spending assumptions. The downside need not be an aggregate earnings recession: a 5-10% reduction in high-multiple technology earnings expectations, combined with a lower terminal growth rate, could drive materially larger equity declines in crowded AI infrastructure names. Conversely, evidence that AI workloads are producing incremental software/advertising revenue rather than merely cost savings would extend the cycle and invalidate the rotation thesis.

Over the next several days, the principal risk is positioning-driven upside in AI proxies and a broad risk-on response to optimistic index targets. Over one to three months, hyperscaler capex guidance, semiconductor lead times, cloud revenue growth, and SMCI inventory/receivables trends are the key catalysts. Over six to eighteen months, the investable question is whether AI depreciation and power costs outrun revenue realization; that is the point at which quality-of-earnings dispersion should widen materially.

Key Decisions for Investors

• Establish a 3-6 month pair: long APP / short SMCI in equal dollar exposure. APP offers comparatively asset-light AI monetization optionality, while SMCI carries higher sensitivity to server pricing and working-capital normalization; target 15-20% relative outperformance, with a stop if SMCI reports sustained gross-margin expansion and declining inventory days while APP ad-revenue growth decelerates materially.
• Use SMH or SOXX puts, 4-6 months to expiry and 5-10% out-of-the-money, as a portfolio hedge rather than a standalone directional short. The catalyst window is the next two earnings cycles; close or reduce if hyperscaler capex guidance rises again and semiconductor pricing/lead-time data remain firm.
• Accumulate GS on broad-market pullbacks over the next 1-3 months rather than chase AI hardware beta. The thesis requires improving announced M&A, IPO, and equity-underwriting volumes; reassess if capital-markets fees fail to inflect by the next two reporting periods or if credit losses begin to rise.
• Monitor a watchlist trigger rather than add broad S&P 500 beta: initiate a quality rotation only if major platform companies begin cutting forward capex or explicitly flag AI depreciation pressure. That development would favor profitable software and financials versus AI infrastructure, but absent such guidance the rotation is premature.