The power super cycle is not just about AI! Morgan Stanley: The "Three A" forces push global power shortages beyond expectations, with about half of market demand yet to be priced in

According to news from Zhitong Finance APP, Morgan Stanley recently released an in-depth global power industry report titled "Powering AI: Agents, Aircons, Allshoring: All Not Priced In," which presents a view different from the mainstream market narrative: Over the past three years, the power consumption upgrade led by data centers and semiconductor wafer fabs has been spreading to the industrial and residential sectors. The "Triple A" forces of Agentic AI, Aircons, and Allshoring (industrial return/supply chain restructuring) will push global electricity tightness beyond expectations. The firm raised its global power consumption growth forecast for 2025–2030 by 10%, predicting that global electricity consumption will reach about 36.3 trillion kWh by 2030—about 60 basis points higher than IEA forecasts and about 15% faster than market expectations.

The core conclusion of Morgan Stanley is: The electricity supercycle is not solely driven by AI. Industrial return, manufacturing expansion, semiconductor localization, and space cooling demand are collectively lifting both baseload and peak load. Currently, the market significantly underprices power generators, midstream natural gas pipelines, grid operators, and coal producers, while growth expectations for upstream power supply chains and renewable energy have already been largely priced in.

Morgan Stanley estimates that about half of the power demand driven by Agentic AI, the Asian industrial supercycle, and new air conditioner/heat pump installations has not yet been reflected in the valuations of power companies, midstream natural gas, or coal producers.

From "Powering AI" to "Powering Everything"

Morgan Stanley regards Agentic AI as the first layer of incremental power demand. The Morgan Stanley U.S. Internet team estimates that by 2029, Agentic AI will form a market of $30 trillion per year, and the token consumption of Agentic AI is 5 to 30 times higher than that of traditional chatbots. Complex enterprise workflows may deploy 5 to 7 agent instances, each consuming about 15 times the tokens of a traditional query, with end-to-end task token consumption reaching between 375,000 and 525,000. The report states that agent workloads already account for about 64% of enterprise output token usage; despite token prices declining, total electricity demand has not decreased, and instead, broader adoption and more complex tasks are driving up absolute computing power and power consumption.

However, the firm emphasizes that AI is not the entire story. Power consumption by the industrial sector will reach about twice that of data centers by 2030; Asian industrial sectors will account for about 43% of the region's incremental power demand, about 2.7 times that contributed by data centers, with China and India accounting for 97% of the global industrial incremental demand. Global industrial power demand is expected to rise from 11.1 trillion units in 2025 to 13.1 trillion units in 2030, with Asia making up 83% of this demand.

The report identifies steel, automotive, semiconductors, defense, refining, chemicals, and related industries as the main carriers of "Allshoring," estimating that by 2030, these may add about 2 trillion units of power consumption, with a compound annual growth rate (CAGR) of around 3.43% from 2025 to 2030, about 1.5 times faster than the past decade.

Space cooling is the third layer of incremental demand. Global power demand for cooling is expected to rise from 2.9 trillion units in 2025 to 4.1 trillion units in 2030, with a CAGR of about 7%, higher than the overall global electricity demand growth of 4.2%. While cooling accounts for about 10% of global electricity consumption, it makes up about 30% of peak power demand. In India, under El Niño conditions in 2026, cooling could account for up to 40% of peak power demand.

The firm expects the stock of installed air conditioners in Asia to increase from about 1.1 billion units to about 1.4 billion by 2030, with electricity demand for cooling rising from about 1.5 trillion units to 2.1 trillion units; India's cooling demand will almost double to about 280 billion units by 2030. In Europe, power consumption for heat pumps is expected to increase from 113 TWh in 2025 to 179 TWh in 2030, elevating the share of Europe's electricity demand from 3.8% to 5.4%. Morgan Stanley believes that the incremental demand for cooling (excluding data centers) could be as large as the additional power demand from data centers.

Pricing Power Shifts from Equipment Vendors to Infrastructure Owners

Morgan Stanley believes that pricing power in electricity markets is undergoing a pivotal shift. After two years of equipment cost inflation, power generators and grid operators have begun to gain stronger pricing power, while equipment suppliers face resistance from consumers and policymakers to high prices. Morgan Stanley expects the average commercial electricity price spread to widen by about 15% by 2028; Asia's wholesale electricity prices have risen about 20% over the past year, with price spreads for generators expanding 2 to 3 times in most regions; globally, electricity prices are expected to rise by about 18% to 20% in 2026 compared to the prior year. The report states that generators can combat high equipment costs by renovating existing equipment and delaying new capacity deployment; expansion announcements for turbines and transformers are expected to alleviate supply chain concerns.

The key basis for this view is that ultra-large-scale AI enterprises are not sensitive to electricity prices. According to Morgan Stanley's analysis of a 100MW fully integrated AI hyperscale project (with proprietary power shell, GPU and LLM), when power prices rise from $120/MWh to $140/MWh (up about 17%), project returns are only impacted by about 60 basis points, and ROE remains well above 30%; even if power prices rise to about $160/MWh (up 1.5 times), ROE remains high. In the 100MW project analysis, revenue is $1.93 billion, power costs $116 million, EBITDA $1.802 billion, net profit $714 million, net margin 37%, ROE 33.6%. The firm thus believes that power utilities have the ability to apply differentiated pricing to data centers to fund grid and generation capacity expansion and enhance long-term returns.

However, Morgan Stanley also points out that power systems are facing increasingly steep load curves. India's "duck curve" is transforming into a double peak curve, Thailand's EV charging peak has shifted to night-time, and Australia's deep midday troughs highlight the value of energy storage and flexible generation. The restart of nuclear power in Japan has reduced the grid's flexibility to absorb renewable fluctuations. By 2030, Asia is expected to require about $1 trillion in grid investments, with data centers partly bearing grid upgrade costs through higher electricity prices, while tiered pricing is used for cross-subsidizing residential consumers.

Asia is the Focus, U.S. Shortfalls Spill Over

At the regional level, Morgan Stanley sees Asia as the center of the global electricity supercycle. Asia (including China) is expected to account for about 64% of global incremental electricity demand from 2025–2030; by 2030 Asia will consume about 20 trillion units of electricity, accounting for more than half the world's total. Asia's incremental electricity consumption is projected to be four times that of the U.S., with demand growth 2.5 times that of Europe plus the UK, and 1.2 times that of the U.S. The report projects Asia’s 2025–2030 electricity demand CAGR at about 5%, about 140 basis points faster than the current average for most Asian economies.

By country, India's 2025–2030 power demand CAGR is projected at 6.5%, China at 5.2%, ASEAN at 4.9%, the U.S. at 4.7%, Australia at 3.5%, Latin America at 3.4%, South Korea at 3.0%, Europe plus the UK at 2.7%, and Japan at 0.9%.

Japan's electricity demand will reverse its two-decade-long decline, but inflexible nuclear baseload forces Greater Tokyo to curtail up to 4.41GW of local solar power and import electricity from the Northeast during March-April 2026. South Korean data center power consumption is projected to rise from about 8 TWh in 2025 to about 30 TWh in 2038, a CAGR of about 10%—much higher than the overall 2% electricity demand growth. However, Korea Electric Power Corporation (KEPCO) will see debt reach KRW 21 trillion in Q2 2026. Samsung Electronics and SK Hynix reportedly rejected KEPCO's proposal to prepay KRW 25 trillion in electricity fees over five years to fund grid investment, and grid financing constraints may limit South Korea's ability to take on AI loads.

For the U.S., the firm estimates that by 2028, data centers in the U.S. will face a 43GW power shortfall, accumulating to 80GW by 2029; even factoring in off-grid power, fuel cells, nuclear, and site conversions (so-called "time-to-power" solutions), the net shortfall remains about 43GW in 2028—equivalent to 38% of U.S. data center power demand. The U.S. power team estimates that U.S. data centers will need to add about 116GW of power from 2026–2028. After accounting for projects under construction, available grid capacity, add-on peaking turbines, Bloom Energy (BE.US) fuel cells, and nuclear-powered data centers, the gap remains 70GW, and after adding about 6GW of Asia-Pacific spillover, the net gap remains 56–38GW.

Morgan Stanley expects around 15GW of the U.S. data center shortfall to spill over to Asia, including about 4GW to Malaysia, 1.5GW to Japan, 4GW to the rest of ASEAN, 2GW to Australia, and about 3.6GW to India and other locations; Chinese hyperscale enterprises also plan to add 2–3GW of capacity annually outside China. Thailand has received commitments of about $25 billion in data center investments, while Malaysia has approved about $55 billion in data center projects since 2025; of 84 projects, 57 will require about 6.2GW of power, all supplied by Tenaga.

Natural gas is seen as a key bridge connecting AI with power security. The firm expects Asia (excluding China) to add about 19GW of gas-fired power capacity from 2025–2030, underwriting about 20 million tons/year of LNG imports—about one-fifth of incremental U.S. shale exports; globally, from 2025–2030, 252GW of new coal and 234GW of new gas power are expected. Asia will add about 303GW of coal, 61GW of gas, 51GW of nuclear, and over 2TW of renewables, and around 3TWh of energy storage. The report argues natural gas, renewables, and storage are increasingly collaborative rather than competitive. Profits should rise for midstream gas pipelines, LNG importers, and related infrastructure, but ongoing global conflicts keep LNG costs high and the market is not yet ready to price this change.

In Europe, Morgan Stanley expects power demand to grow 1.7% in 2026, and 2.2% CAGR from 2026–2030—the first multi-year sustained growth since 2014. Data centers will contribute about 90TWh in incremental demand from 2025–2030, accounting for 25% to 30% of Europe's power demand growth, and raising the European data center share of demand from nearly 3% in 2026 to about 5% in 2030 and about 8% by 2035. The report estimates data center demand could drive up European power prices by about €2/MWh by 2030, with the UK possibly exceeding €3/MWh, and bull market conditions could push this to over €4/MWh. Europe's electrification rate in 2024 is just 21%, increasing by just 0.1 percentage points per year since 2010; if it follows the IEA path to reach 28–30% by 2035, Europe will need to add about 1,400TWh of power and around 900GW of generation capacity over the next decade, much of it solar.

Latin American power demand is expected to grow at a 3%–4% CAGR from 2025–2030. In a bull market scenario, data centers will add about 1TWg of demand by 2030 (about 7.9GW in Brazil), equivalent to about 4% of total demand and contributing about 80 basis points to overall growth; however, traditional economy, industry, and electrification remain the main drivers. In the Australian NEM, data center power consumption is projected to increase from about 5TWh in 2025–2026 to 34TWh in 2035–2036, taking the share of grid-supplied electricity from about 3% to around 13%. Australian battery deployments are approaching commercial market saturation, commissioning 3.7GW/9.8GWh of BESS and 6GW/1GWh of home batteries since early 2026; battery power capacity is about 14% of overall system capacity (117GW), likely suppressing future wholesale power prices.

Investment Strategy: 36 Winners and an Avoid List

Morgan Stanley has selected 36 "Powering AI" winners from across the global power value chain, covering reliable power generators, midstream natural gas pipelines, grid companies, integrated energy and storage suppliers, but remains selective on equipment suppliers due to pricing power shifting toward infrastructure owners. The firm notes that current equity valuations only price in about one-third of its baseline case for electricity consumption growth, close to mid-cycle consumption levels over the past decade. Its preferred picks have about 14% upside in 2028 EPS, corresponding to about $500 billion in market value creation.

Morgan Stanley's preferred winners include: EDP (Portugal), RWE (Germany); Asian names such as Gulf Development, Keppel, Adani Power, Yangtze Power, Samsung SDI (Samsung SDI), Hokkaido Electric (Hokkaido Electric); and in the U.S., Bloom Energy, Solaris Energy (SEI.US).

The report also lists names to avoid: in Asia, KEPCO, Tata Power, Global Power Synergy, Huangen Power, Electric Power Development; and in Europe, Enagas, and SNAM.

In terms of specifics, about 35%-40% of EDP’s 2027 EBITDA is exposed to Portugal, one of the fastest-growing electricity markets in Europe, with demand CAGR forecast at 4.5% from 2025–2030; EDP has already signed MOUs with Start Campus and Merlin. RWE is positioned for European data center supply through a combination of renewables and flexible capacity, with around 3GW of available capacity (10 sites of 300MW each) for “powered land” monetization. In the U.S., Bloom Energy is seen as a key beneficiary of “time-to-power” shortages; Morgan Stanley expects 6GW of fuel cells to be deployed from 2026–2028, with recent site expansion hinting at upside. Solaris Energy is also a top U.S. pick. In Asia, Gulf Development, Adani Power, Sembcorp, etc., are providing behind-the-meter power to data centers; Yangtze Power is considered a high-quality defensive hydropower utility, and Samsung SDI benefits from data center backup power and grid-level storage demand for ESS, BBU, and UPS.