A rotation within the artificial intelligence sector accelerated in the second quarter of 2026, as capital flowed from high-flying chip designers toward companies building the physical infrastructure required for AI. Data from the quarter ending June 30, 2026, shows significant outperformance by firms focused on power grid expansion, semiconductor manufacturing equipment, and data center construction. This divergence highlights a maturation of the AI investment thesis, moving from pure-play software and hardware to the essential utilities of the digital economy. While the Nasdaq Composite advanced 4.2% in the quarter, infrastructure-centric stocks like Quanta Services and Applied Materials posted gains exceeding 15%.
Context — [why the AI trade is broadening now]
The AI investment theme initially concentrated on a narrow cohort of companies, most notably NVIDIA, which saw its valuation increase over sevenfold from 2022 to early 2026. This hyper-focus on graphics processing units (GPUs) mirrored early internet investing, which initially favored web portals before capital flowed to internet infrastructure providers like Cisco Systems. The current shift is driven by a tangible bottleneck: the immense energy demands of AI data centers. A single large-language model training run can consume more electricity than 100 homes use in a year, forcing a reckoning with global power capacity.
The catalyst for the recent rotation is a wave of revised capital expenditure forecasts from major cloud providers. In May 2026, multiple hyperscalers announced aggregate spending projections on data center construction and energy procurement that exceeded $400 billion over the next five years. This committed capital signals a long-term, physical build-out phase for AI, moving beyond the initial purchasing of servers. Concurrently, rising bond yields have increased pressure on richly valued tech stocks, making the more reasonably valued industrial and utility segments attractive for new allocations.
Data — [what the numbers show]
Performance data for Q2 2026 illustrates the clear divergence within the AI ecosystem. The iShares Semiconductor ETF (SOXX) returned a modest 5.1% for the quarter. In contrast, the Global X Data Center REITs & Digital Infrastructure ETF (VPN) gained 11.8%. Quanta Services, a leading electrical grid contractor, saw its stock price increase 18.5% in the quarter, while semiconductor equipment maker Applied Materials rose 16.2%. NVIDIA’s stock was nearly flat, appreciating just 1.5% over the same period.
A comparison of valuations further explains the rotation. NVIDIA trades at a forward price-to-earnings ratio of approximately 35x, based on 2027 earnings estimates. Quanta Services and Eaton Corporation, a power management company, trade at forward P/E ratios of 22x and 21x, respectively. The capital intensity of the AI build-out is also evident in order books. Leading chip fabrication plant builder KBR reported a 40% year-over-year increase in its infrastructure backlog, reaching $12.5 billion. This tangible demand contrasts with the more speculative growth projections for some AI software applications.
| Company / ETF | Q2 2026 Return | Primary AI Catalyst |
|---|
| NVIDIA (NVDA) | +1.5% | GPU Demand |
| Quanta Services (PWR) | +18.5% | Grid Expansion |
| Applied Materials (AMAT) | +16.2% | Chip Manufacturing Tools |
| SOXX ETF | +5.1% | Broad Semiconductors |
Analysis — [what it means for markets / sectors / tickers]
The capital rotation signifies a second-order effect where the enablers of AI adoption are capturing significant value. Primary beneficiaries include electrical component manufacturers like Eaton Corp and Vertiv Holdings, utility companies with dense power grids, and engineering firms specializing in data center construction. The demand for high-voltage transformers and power conversion systems is projected to grow at a 15% compound annual growth rate through 2030, directly benefiting industrial conglomerates. This trend also supports commodities like copper, essential for electrical transmission, where prices have risen 12% year-to-date.
A key risk to this infrastructure thesis is execution. Building physical assets is subject to permitting delays, supply chain constraints for components like transformers, and labor shortages. These projects also carry higher interest rate sensitivity due to their long-duration cash flows. If the pace of AI adoption slows unexpectedly, these infrastructure investments could face underutilization. Despite these risks, positioning data from major prime brokerages indicates net inflows into industrial and materials sectors throughout Q2, while technology sector flows turned neutral. Hedge funds have been actively establishing long positions in the infrastructure supply chain while shorting the more crowded, high-multiple AI software names.
Outlook — [what to watch next]
The trajectory of this trend will be tested by several imminent catalysts. Second-quarter earnings reports, starting in mid-July 2026, will provide critical updates on the capital expenditure timelines from Microsoft, Amazon Web Services, and Google Cloud. Guidance from utility companies like NextEra Energy on their grid modernization investments will be equally important. The Department of Energy’s report on national grid reliability, expected August 15, 2026, could influence regulatory support for infrastructure spending.
Market participants should monitor the relative performance ratio of the Industrial Select Sector SPDR Fund (XLI) against the Technology Select Sector SPDR Fund (XLK). A rising ratio confirms the rotation into industrials is ongoing. Key technical levels for Quanta Services include the $285 price area as near-term support, a 15% retracement from its recent highs. For the trade to sustain, 10-year Treasury yields need to stabilize; a sharp spike above 4.5% could pressure the valuation of all long-duration assets, including infrastructure stocks.
Frequently Asked Questions
What are the best AI infrastructure stocks to watch?
Beyond well-known names, investors are monitoring companies in the middle of the supply chain. This includes Cadence Design Systems and Synopsys, which provide electronic design automation software essential for creating new AI chips. It also includes cooling technology firms like CoolIT Systems, critical for data center efficiency, and uranium producers like Cameco, as nuclear power is increasingly considered a stable baseload energy source for AI data centers. The investment universe extends to rare earth element miners vital for permanent magnets in electrical systems.
How does this AI infrastructure build-out compare to the dot-com era?
The scale of capital expenditure is larger and more concentrated than during the dot-com bubble. In the late 1990s, telecom companies overbuilt fiber optic capacity, leading to a crash. Today, the spending is led by a handful of cash-rich hyperscale cloud providers with proven demand for their AI services. The key difference is that current infrastructure is being built to solve a proven capacity constraint—energy availability—rather than a speculative bet on future demand, potentially making it more durable.
