Damao AI Energy Summit warns of “power crisis”: the 2027-2028 high-risk period is approaching, and turbine and computing power service providers may benefit

The Zhitong Finance App learned that on December 4, 2025, Morgan Stanley held the 2nd “AI Energy Summit” in New York. The core findings revolved around electricity shortages, solution iteration, and investment directions, providing a key reference for industry development and capital layout.

The summit focused on energy support challenges and opportunities for AI development in North America. The number of participants reached 400, an increase of nearly 4 times over last year, reflecting an explosive increase in attention in the field of AI infrastructure. Key points include that US data center developers may face serious AI power shortages, the rise of off-grid solutions (off-grid solutions), preferential terms for power+data center transactions, political risks, labor and equipment shortages; the value of “time to power” solutions (time to power solutions) is prominent, and demand for natural gas is expected to grow.

The summit's first core finding was that US data center developers will face severe power shortages. Morgan Stanley predicts that the electricity gap for US data center developers may reach 10-20% in the next few years. The risk will gradually become apparent from 2026, and 2027-2028 will enter a high-risk period.

This gap is due to a nonlinear increase in AI computing power demand — the current data center size has jumped from 100-500 MW 1.5 years ago to 1 GW (4 GW in a Louisiana park), while traditional grid access capacity and power supply growth are seriously lagging behind, compounding the political controversy caused by the data center's high electricity consumption and water consumption, further increasing the power supply pressure.

In response to the power dilemma, off-grid solutions have become an industry consensus. As grid-connected projects face political risks such as rising electricity prices, complicated approval, etc., and grid access requires higher security deposits, data center developers' preferences for off-grid solutions continue to heat up.

The mainstream technology portfolio includes natural gas turbines, reciprocating engines, fuel cells, and battery energy storage. This type of solution not only avoids grid-connected disputes, but also gradually shifts from transitive power supply to permanent solutions.

The terms of the power and data center development deal have improved, but execution risks are still significant. Given the significant excess demand for such products (and the bank believes that the imbalance between supply and demand is still increasing), power and data center developers can expect better terms (such as project risk allocation). This progress is positive because project execution faces many risks, covering various aspects such as supply chain and labor.

The bank believes that 2026 will be the “year of execution,” and the performance of stocks related to “time to power” solutions will increasingly depend on the development results of power and data center projects.

In terms of investment direction, Damo said it is fundamentally optimistic about the huge benefits brought by artificial intelligence applications, and it is expected that as the large-scale language model (LLM) continues to iterate at a nonlinear pace in 2026, the scale of this revenue will expand rapidly. Numerous data shows that even if sufficient data center infrastructure can be deployed (which is extremely challenging in the US), demand for computing power will still exceed supply.

The bank proposed the core strategy of “laying out AI deployment restrictions”, which are specifically divided into three categories: the first is a “time to power” solution, including utility companies that quickly provide grid access, companies related to the transformation of Bitcoin mines into HPC data centers, turbine manufacturers such as GeverNova (GEV.US) and Siemens Energy, fuel cell companies such as BloomEnergy (BE.US), and service providers for the entire gas industry chain such as EQT Energy (EQT.US) and Williams Companies;

The second is “computing power service providers”, covering chip ecosystem companies such as Nvidia (NVDA.US), and emerging cloud service providers such as data center REITs and CoreWeave (CRWV.US); the third is the field of “bottleneck mitigation”, including engineering construction, transformers, and transmission equipment.

Brookfield presented its views on AI power supply. Brookfield plans to provide a complete integrated data center solution (covering land, electricity, and chips), focusing on the North American and European markets. The company relied on its global real estate assets to gain a competitive advantage in this business.

Electricity is still the core bottleneck in data center construction, although other factors such as labor and land are also limited: when customers bring their own graphics cards (GPUs), electricity costs only account for 5% of the total cost of the data center (Brookfield believes that existing GPUs contain a large amount of residual value).

The real constraint does not come from the gas turbine original equipment manufacturer (OEM) itself, but rather the packaging and integration required for the entire value chain — turbine manufacturers rely on intermediaries to assemble the complete solution and deliver the final power solution package to the customer.

Williams (WMB.US) presented its views on indoor electricity solutions, with a long-term contract period of 10-15 years (the ratio of enterprise value/profit before interest, tax, depreciation and amortization (EV/EBITDA) during the construction phase is 5 times). Whether to renew the contract depends on customer needs, but it is expected that customers will continue to use the power co-location model. The company believes that grid access is beneficial to all parties. Under these circumstances, the turbines can provide incremental megawatts of power capacity as needed; alternatively, these turbines can also be migrated to its core business (compression stations).

Ohio (home of the company's first indoor power project, “Socrates”) is an advantageous region for indoor power generation: local legislation supports it (the approval process is clear and stable), which enables rapid construction and rapid commissioning.

Galaxy Digital (GLXY.US) published an opinion on the conversion of cryptocurrency facilities to data centers. The company believes that market concerns about CoreWeave's credit quality have been exaggerated based on evaluations of CoreWeave-like lenders before listing; considering that CoreWeave has reduced its dependence on Microsoft/OpenAI and diversified its business, its credit situation may have improved since then.

Galaxy remains optimistic about the recent approval of new electricity for the Helios project. Galaxy has set aside a margin of safety in its construction schedule/budget, which makes its guidance on capital expenditure per megawatt slightly higher than other announced transactions from cryptocurrency facilities to high-performance computing facilities.