AI infrastructure capex spending surges 116% year-over-year as S&P 500 Q3 earnings forecast rises 27% to record.

U.S. companies are entering what could be a record-breaking earnings season. Goldman Sachs forecasts S&P 500 third-quarter earnings will grow approximately 27% year-over-year, with revenue up about 12%. If realized, this marks the third consecutive quarter of U.S. corporate earnings growth exceeding 25%—the strongest performance since 2021. Société Générale's chief U.S. equity strategist Manish Kabra notes that outside of post-financial-crisis recovery periods, U.S. equities have never posted three consecutive quarters of earnings growth above 25%.

The robust earnings outlook has enabled U.S. stocks to absorb upward pressure from interest rates. The 10-year U.S. Treasury yield recently breached 5%, pushing long-term financing costs to their highest levels in more than two decades, yet the S&P 500 and Nasdaq still hit record closing highs.

AI infrastructure buildout remains the primary driver of earnings expansion. Goldman Sachs estimates hyperscale cloud companies' Q3 capital expenditures grew approximately 116% year-over-year, accelerating from 87% in Q2. This creates massive demand across data centers, AI servers, high-end chips, memory, power, cooling systems, and industrial equipment. Federated Hermes chief equity investment officer Steve Chiavarone describes AI capex as producing the strongest corporate earnings and margin growth environment of his career.

The impact has spread beyond technology into other sectors. FactSet projects all 11 S&P 500 sectors will deliver year-over-year earnings growth in Q3; 72 S&P 500 companies have issued positive earnings guidance—a record high—while negative guidance has fallen to a five-year low. The energy sector is outperforming notably: West Texas Intermediate crude rose roughly 30% in Q3 on Middle East conflict concerns, and FactSet expects U.S.-listed energy companies' Q3 profits to jump approximately 114% year-over-year, well above the 79% forecast at end-June. Refiners such as Marathon Petroleum and Valero Energy are expected to be primary beneficiaries.

Despite earnings expansion broadening across sectors, U.S. profit growth remains highly concentrated. Goldman Sachs estimates Nvidia (NVDA.O) and Micron Technology (MU.O) alone could contribute more than one-third of all S&P 500 Q3 earnings growth. Specifically, Micron is expected to contribute 19% of S&P 500 Q3 earnings growth and Nvidia 15%—together roughly equal to the combined contribution of the other 490 constituents. The next largest contributors include Meta (META.O) at 7%, Alphabet (GOOGL.O) at 6%, and Broadcom (AVGO.O) at 5%. The top ten contributors are projected to account for 68% of Q3 earnings growth, up sharply from 47% in Q1 and 48% in Q2.

By theme, Goldman Sachs expects "AI infrastructure (ex-hyperscalers)" to contribute 54% of Q3 earnings growth, with hyperscalers adding another 19%—together accounting for roughly three-quarters of S&P 500 earnings growth.

This represents the most glaring structural contradiction in today's U.S. market. Big tech's capital spending continues translating into revenue for chipmakers, equipment suppliers, power companies, and industrial firms, sustaining rapid U.S. corporate profit growth. Yet an increasingly large share of earnings contribution is concentrated in a handful of AI companies, making index performance ever more dependent on this capex cycle.

U.S. market breadth has narrowed to levels not seen since the dot-com era. The S&P 500 sits just 2% below its all-time high, yet the median stock is down 17% from its own peak. Goldman Sachs data shows the average realized correlation among S&P 500 constituents has fallen to 0.06—the lowest in the firm's decade-long dataset. Morgan Stanley strategist Mike Wilson argues the divergence between the index near record highs and a large swath of individual stocks in bear-market territory "will eventually converge one way or another," with bond market volatility likely serving as the ultimate "referee."

Bank of America Merrill Lynch strategist Michael Hartnett points out that roughly 400 S&P 500 constituents have broken below their 50-day moving averages, and about 300 below their 200-day. His "AI Big 10"—the Magnificent 7 plus Broadcom, AMD, and Micron—now commands a market-cap concentration of 42%, exceeding the "Nifty Fifty" era's 40%, the TMT bubble's 41%, and the 1920s' 36%.

For investors, the most important number this earnings season may be not profits themselves but how much big tech plans to spend next. AllianceBernstein head of equities Nelson Yu says the market first needs to see whether hyperscale cloud companies continue raising capex plans, since that determines revenue growth across the entire AI supply chain in coming quarters.

Market forecasts for hyperscaler capex have consistently lagged actual growth. In early 2024, the market expected 19% full-year growth, but actual growth came in at 54%; in early 2025, the forecast was 22%, actual was 73%; earlier this year, the forecast was 36%, and the current run rate is already near 96%. Goldman Sachs therefore believes the market's 37% capex growth forecast for 2027 may also prove too low, projecting 2027 capex growth will exceed 50%, with total spending surpassing the current consensus of roughly $1.1 trillion.

Accelerating cloud revenue growth is underpinning this capex cycle. Cloud revenue growth at Amazon, Google, Microsoft, and Oracle accelerated to 48% in Q2, and Goldman Sachs expects it to climb further to 55% in Q3. The combined cloud backlog at Amazon, Google, and Microsoft now approaches $1.7 trillion.

The high-rate environment adds urgency to these questions. The Federal Reserve delivered its first rate cut in three years in September, yet long-term U.S. Treasury yields have risen sharply, increasing financing pressure on financials, real estate, utilities, small and mid-sized businesses, and households. If AI capex continues growing at triple-digit rates, this investment can offset a meaningful portion of the economic drag from high rates. But Pimco economist Tiffany Wilding cautions that once AI investment growth decelerates meaningfully, U.S. economic resilience will depend on whether AI productivity gains materialize quickly enough to fill the gap left by slowing capex.

Micron already reported earnings on September 30, delivering strong results: adjusted EPS of $33.42, beating the consensus estimate of $31.83; adjusted revenue surged to $54.2 billion from $11.2 billion a year earlier; gross margin reached 87%, far above the prior year's 45.7%. The company's next-quarter revenue guidance range of $60 billion to $63 billion also came in well above the consensus estimate of $56.77 billion. Yet the blowout results did not trigger a commensurate stock rally. Morgan Stanley analyst Joe Moore noted before the report that the market discussion around Micron "has clearly shifted from 'how good can it get' to 'how long can this momentum last.'" Micron shares are up more than 231% year-to-date.

As the largest S&P 500 weight and the second-largest contributor to Q3 earnings growth, Nvidia will report earnings in the third week of November. In the current environment of extreme earnings concentration, its results and guidance on future AI capex and demand could be the pivotal moment determining whether this U.S. earnings rally can continue.

The upcoming Q3 earnings season may once again deliver very strong profit numbers, but the market is truly testing something else: how much longer the AI capex cycle—the force driving profit growth, stock gains, and part of U.S. economic investment expansion—can sustain its current pace.