Everything About AI
NextFin News - For the first time in years, artificial intelligence is no longer a profit story confined to the technology sector. As the third quarter of 2026 closes, all eleven S&P 500 sectors are expected to report year-over-year earnings growth, and the engine behind that breadth is the largest capital-spending buildout in modern corporate history: roughly $1 trillion of global AI investment this year, of which about $581 billion is landing in the United States. The question is no longer whether AI can move markets - it is whether the spillover can hold once the hyperscalers slow.
The S&P 500 is on track for its third consecutive quarter of earnings growth above 25%, with the estimated year-over-year earnings growth rate for the third quarter at 28.5% as of early September. That breadth is the real story. For the full year 2026, analysts expect 15.0% earnings growth across the index, and every one of the eleven GICS sectors is forecast to grow profits from a year earlier - five of them at double-digit rates. Behind those figures sits a buildout that Goldman Sachs Research compares to the railways, the interstate highway system, and the internet backbone: an investment impulse large enough to lift the companies that supply the power, the steel, the cooling systems, the financing, and the construction labor for the data-center economy.
The Buildout Has a Transmission Mechanism - and It Runs Through the Real Economy
The first-order effect of the AI boom was obvious: Nvidia, the chip designers, and the cloud platforms that buy their silicon saw earnings explode. The second-order effect is what is reshaping the index now. Every dollar of AI capital expenditure does not vanish into a server rack; it pays an electric utility for power, an engineering firm for design, a construction company for the shell, a copper producer for wiring, and a bank for the financing. That chain is why a spending decision made in a Menlo Park or Redmond boardroom shows up as earnings growth in sectors that have nothing to do with software.
The scale matters. Goldman Sachs Research estimates global AI-related investment will exceed $1 trillion in 2026, with $581 billion of it in the United States. Cumulative AI investment since 2022 is expected to reach $1.8 trillion by year-end. To put that in perspective, AI capex will consume about 1.8% of U.S. GDP this year, rising to 2.5% in 2027 and 2.8% in 2028. Joseph Briggs, who co-leads the Global Economics team at Goldman Sachs Research, notes that these levels are "consistent with the 2%-5% of GDP peak investment impulses observed in prior general-purpose technology buildouts." The phrase "general-purpose technology" is doing heavy lifting: the railroads did not enrich only locomotive makers, and the internet did not enrich only router vendors.
Power is the clearest transmission channel. U.S. investor-owned utilities plan to spend at least $1.4 trillion on grid and generation projects through 2030, a 30% increase from projections made a year earlier, according to the nonprofit PowerLines. A majority of the 51 utilities surveyed named data centers as the primary driver of that spending. That converts a technology narrative into regulated rate-base growth for the Utilities sector - earnings that are contracted, visible, and largely insulated from the consumer cycle.
The Numbers Confirm the Breadth
The earnings data back the mechanism. For the third quarter of 2026, the estimated year-over-year earnings growth rate for the S&P 500 is 28.5%, and if that figure holds it will mark the third straight quarter of growth above 25% for the index, according to FactSet's Earnings Insight report dated September 4, 2026. More importantly for the breadth argument, the estimate has been revised upward during the quarter at a time when it normally falls. The bottom-up EPS estimate for Q3 rose 1.2%, to $89.69 from $88.64, between June 30 and August 31. Over the past five years, the average change in the bottom-up estimate during the first two months of a quarter has been a decline of 1.7%; over twenty years, a decline of 3.1%. Estimates are rising when history says they should be falling.
At the sector level, four of the eleven sectors saw their bottom-up EPS estimates increase from June 30 to August 31, led by Energy at +11.8%. Seven sectors recorded decreases, led by Materials at -9.1% - a reminder that the buildout's beneficiaries are not evenly distributed and that commodity names can be whipsawed by trade policy and demand timing even as the long-term thesis holds. For the full year, the CY2026 bottom-up EPS estimate rose 6.1%, to $361.38 from $340.49, over the same two-month window, and seven of eleven sectors recorded an increase.
For calendar year 2026 as a whole, FactSet expects 15.0% earnings growth, well above the trailing ten-year average annual growth rate of 8.6%. All eleven sectors are predicted to report year-over-year growth, and five - Information Technology, Materials, Industrials, Communication Services, and Consumer Discretionary - are projected to grow at double-digit rates. Revenue growth of 7.2% is also above the ten-year average of 5.3%, and the estimated net profit margin of 13.9% would be the highest annual figure since FactSet began tracking the metric in 2008. The AI buildout is not just lifting revenue; it is lifting profitability across the index.
Who Captures the Spillover - and Who Does Not
The beneficiaries cluster in the physical layers of the buildout. Industrials gain from electrical equipment, data-center construction, and the manufacturing upturn that AI capex is helping to sustain. State Street Global Advisors' Q3 2026 sector outlook describes a "manufacturing-led upturn" and notes that "less price-sensitive AI CapEx" is among the factors supporting the recovery, upgrading Industrials to a positive view alongside Materials and Financials. Materials captures demand for copper, steel, and specialty chemicals; copper in particular has become a proxy for electrification, with data centers and grid expansion competing for supply.
Financials benefit through a different channel: the financing of the buildout and the associated pick-up in capital-markets activity. FactSet's first-quarter preview projected the Financials sector would report the third-highest growth rate among the eleven sectors, at 15.1%, with banks and capital-markets firms both contributing. When corporations borrow to fund data centers and utilities issue debt to fund transmission lines, fee income and net interest margins follow.
Energy sits at the front of the pack for the third quarter. FactSet lists Energy alongside Information Technology, Communication Services, and Materials as the leaders in Q3 earnings growth, helped by upward estimate revisions of 11.8% over the summer. But the relationship with the AI buildout is indirect and contested: natural gas fuels a large share of marginal U.S. power generation, so data-center load supports gas demand, yet the sector's own full-year revenue is forecast to decline as supply growth weighs on prices. Energy's earnings growth is a margin and cost story as much as a demand story.
Not every sector is a clean beneficiary. Consumer Discretionary and Consumer Staples face a consumer whose real disposable income has declined year over year, with consumption growth expected to decelerate toward 2.0% from 2.6% last year, according to State Street. The AI economy is lifting corporate profits faster than it is lifting household purchasing power - a divergence that matters for the durability of the breadth.
Cyclical Wave or Structural Shift? This Is Both - and That Is the Risk
The central judgment: the AI buildout is a structural shift riding on a cyclical wave, and the two must be separated. The structural leg is the technology itself. AI is a general-purpose technology with a decade-long adoption curve ahead; the investment impulse of 1.8% to 2.8% of GDP is consistent with prior technology buildouts that reshaped the economy rather than merely boosted a cycle. Adam Benjamin, manager of the Fidelity Select Technology Portfolio, put it plainly:
AI is the most powerful and far-reaching of all the cycles of innovation and disruption I've seen in my 25 years following tech.
A buildout of that character does not reverse on its own.
But the earnings breadth it is producing in 2026 contains a large cyclical component that will mean-revert. The third quarter's 28.5% growth rate is being compared against a depressed year-earlier base in several sectors, particularly Energy and Materials, where 2025 earnings were compressed by the manufacturing slowdown and weak commodity prices. The estimate revisions running counter to seasonal norms - up 1.2% when they typically fall 1.7% to 3.1% - are a cyclical tailwind, and cyclical tailwinds expire. When the base effect rolls over and the pace of incremental capex growth slows, the earnings growth rate will compress even if the absolute level of spending keeps rising.
This distinction determines the conclusion. The structural shift means the beneficiaries of AI - power, grid, data-center infrastructure, and the industrial supply chain - will see a higher earnings floor for years. The cyclical wave means the 28.5% growth print is not the new normal, and investors extrapolating it forward are making a mistake. The opportunity is in the infrastructure layer; the trap is in the growth-rate extrapolation.
The Counter-Thesis: A Capex Bubble With No Return
The strongest case against this view is also the simplest: the hyperscalers are spending money they may never earn back, and when the return-on-investment reckoning arrives, the entire spillover reverses. The bear argument runs that AI revenue is not yet covering the capital cost of the buildout, that the leading indicators Goldman Sachs tracks - GPU rental prices, memory prices, semiconductor equipment imports - can turn quickly, and that a cut in hyperscaler capex guidance would hit the utilities, industrials, and materials names that have front-run the spending.
This is not a strawman. It is the view held by investors who see capex running ahead of monetization, and it has a clear transmission channel: if the four largest U.S. technology companies - Amazon, Alphabet, Meta, and Microsoft, which have signaled combined capital expenditures of about $650 billion for 2026 - collectively slow their spending, the $1.4 trillion utility capex plan loses its anchor and the industrial order books thin out. The concentration is the vulnerability: a handful of companies are deciding the earnings trajectory of eleven sectors.
The answer to the bear is that the buildout has already passed the point of no return for the infrastructure layer. Data-center power demand is contracted and rate-based; utilities do not build generation on speculation. Even if AI application revenue disappoints, the physical capacity is being paid for through regulated returns and long-term power purchase agreements. But the bear wins on one condition, and that condition defines the falsifying signal: if hyperscaler capex guidance for 2027 falls by more than 15% from current consensus while GPU utilization rates stay below 60%, the structural thesis for the spillover is broken and the cyclical leg turns down hard.
What Comes Next: Three Horizons
In the short term - the next one to two quarters - earnings momentum favors the index. The Q3 print near 28.5% growth, combined with 87% of S&P 500 companies reporting positive EPS surprises in the second quarter, creates a high beat-rate environment that supports valuations even as the forward 12-month price-to-earnings ratio of 19.5 sits above its ten-year average of 19.0. Sentiment and liquidity can carry the breadth trade further than fundamentals alone would justify.
In the medium term - six to eighteen months - the story splits. Companies with contracted AI-exposed revenue - utilities with data-center load, industrials with multi-year order backlogs, copper and electrical-equipment suppliers - should continue to deliver. Consumer-facing sectors face a harder path as slowing real income growth meets still-elevated interest rates. This is where the divergence between fundamentals and market performance, which State Street flags, should narrow - or where it should widen into a rotation.
In the long term - beyond 2027 - the structural question dominates. PwC projects global data-center spending could reach $31.6 trillion through 2050, with the United States capturing nearly half at $15.1 trillion. If that trajectory holds, the AI buildout is a multi-decade reconfiguration of the capital stock, and the sectors that supply it will compound earnings through multiple cycles. If it does not, the 2026 breadth will be remembered as the peak of a capex bubble.
Three scenarios frame the path. The base case: hyperscaler capex grows through 2027 at a slowing but positive pace, earnings growth normalizes to high single digits, and the infrastructure beneficiaries continue to outperform on visible order books. The upside case: AI monetization accelerates, capex estimates are revised upward again, and the breadth extends into consumer sectors as productivity gains lift wages and spending. The downside case: hyperscaler guidance cuts exceed 15%, GPU utilization disappoints, and the spillover reverses into a synchronized earnings downgrade across utilities, industrials, and materials.
What to watch is specific. First, the Q4 2026 and Q1 2027 capex guidance from Amazon, Alphabet, Meta, and Microsoft - a combined cut above 15% from current consensus is the break signal. Second, the FactSet bottom-up EPS revision trend for Q1 2027 - if estimates resume their typical seasonal decline of 1.7% or more during the quarter, the revision tailwind has expired. Third, the PowerLines utility capex survey in 2027 - if the $1.4 trillion plan is revised downward by double digits, the demand anchor is slipping.
The AI boom began as a technology trade. It is becoming an industrial one, and the market is only partway through repricing that shift. The investors who get this right will not be those who own the most AI exposure, but those who own the right kind: the power, the pipes, and the permits that no algorithm can replicate.
Explore more exclusive insights at nextfin.ai.
