On July 30, 2026, the U.S. stock market experienced a wave of structural trends driven by AI computing power—not chips themselves, but the power and cooling infrastructure supporting computing power. As global tech giants accelerate the expansion of AI data centers, electricity consumption has risen exponentially, with stocks of nuclear power, natural gas, and liquid cooling companies collectively strengthening, becoming the new main line after AI chips.
AI Data Center Power Gap Widens
According to the latest report by the International Energy Agency (IEA), global data center electricity consumption increased by 45% year-on-year in the first half of 2026, with AI training and inference accounting for over 60% of the new electricity consumption. Taking the GPT-5 mega-cluster as an example, the peak power consumption of a single cluster has exceeded 500 megawatts, close to the output of a medium-sized nuclear power plant. Data from the U.S. Department of Energy shows that the power load in Virginia's "Data Center Corridor" has exceeded 80% of the local grid's capacity, forcing many utility companies to postpone new project grid connections.
Against this background, power supply has become a key bottleneck restricting AI computing power expansion. Tech giants are turning to nuclear power and natural gas to ensure stable, low-carbon baseload power. Driven by this, U.S. listed nuclear power operators Constellation Energy (CEG) and Talen Energy (TLN) have risen 112% and 89% respectively this year, while natural gas power generators CGE and NRG Energy have also recorded gains of over 50%. Analysts point out that Amazon, Microsoft, Google, etc. have signed multiple power purchase agreements (PPAs) to lock in nuclear power output for the next 10 years. The nuclear renaissance theme has officially moved from concept to performance realization.
Liquid Cooling: Essential Technology for High Power Consumption
Simultaneously with the surge in power demand, AI server cooling technology is iterating. Traditional air cooling can no longer meet the heat dissipation needs of GPU chips exceeding 1000W/TDP. Liquid cooling, especially immersion liquid cooling, has become the standard for new data centers. According to research firm IDC, the global liquid cooling data center market will exceed $20 billion in 2026, with a compound annual growth rate of 58%. In the U.S. stock market, liquid cooling leaders Vertiv (VRT) and CoolIT Systems (unlisted but held by related ETFs) have hit record highs, while LiquidStack (private enterprise) specializing in immersion liquid cooling is also expected to go public via SPAC.
In addition, data center power equipment suppliers also benefit. The order backlog of American Power Conversion (APC, under Schneider) and Eaton (ETN) increased by 80% year-on-year, mainly from uninterruptible power supplies (UPS) and power distribution systems for AI data centers. It is worth noting that Taiwan-based ODM manufacturers such as Quanta and Wistron have increased their share in liquid cooling server shipments, but U.S. stock related targets are more concentrated in power management and cooling components.
Investment Logic: Paradigm Shift from 'Computing Power' to 'Electric Power'
Looking back over the past two years, the AI investment main line has always revolved around GPU chips and optical modules. However, as chip manufacturing processes approach physical limits, power consumption growth far exceeds performance improvements. The capital expenditure share of power infrastructure is rapidly rising from 15% to over 35% of data center costs. According to the latest report from Morgan Stanley, global AI infrastructure investment will exceed $1.5 trillion from 2026 to 2028, of which power and cooling related spending will exceed $400 billion, becoming the fastest-growing sub-track.
For U.S. stock investors, currently investing in AI power and liquid cooling stocks is equivalent to grasping the 'pick-and-shovel' logic of computing power expansion. Compared to the high volatility of chip stocks, power stocks and cooling equipment stocks have stronger defensive attributes and benefit from long-term contracts and recurring revenue. However, caution is needed: nuclear power approval cycles are long, and commercialization of small modular reactors (SMRs) will take time; liquid cooling technology paths are still evolving, and some startups face the risk of being eliminated.
Outlook and Risk Warning
In the short term, the third-quarter earnings season of U.S. stocks is about to begin. Power and cooling companies are expected to deliver better-than-expected results, and sector momentum is likely to continue. In the long term, the growth of AI computing power's electricity demand is highly certain, but investors need to be wary of government regulations on carbon emissions and natural gas price volatility risks. Investors can pay attention to AI data center themed ETFs, such as the Global X Data Center Infrastructure ETF (VPN) or the First Trust Nasdaq Clean Edge Green Energy Index Fund (QCLN), which include core nuclear power and liquid cooling targets.
In summary, in the second half of 2026, the U.S. stock AI computing power sector is undergoing an expansion in connotation—from chips to power, from computing density to cooling efficiency. Regardless of what AI positions you hold, you cannot ignore the underlying currents of 'electricity' and 'heat'.
