The pullback in “time to power” stocks triggered by Meta cloud business rumors is viewed by Morgan Stanley as a buying opportunity

Overview

This report discusses the impact on US “time to power” stocks after Meta reportedly plans to develop a cloud business. Morgan Stanley believes the market selloff in related stocks in response to the news is an overreaction. The core logic is that AI compute demand still far exceeds supply, while power access, data center construction capacity, and energy infrastructure remain expansion bottlenecks. The report views Meta's potential compute leasing as an opportunity to temporarily monetize excess owned compute, rather than as the sole core of Meta's investment thesis.

Core views

Core views include: first, the economic returns from selling compute are attractive, and Meta-related news validates the economics of neocloud transactions and compute leasing; second, timely access to power remains a key bottleneck for AI data center expansion, giving an advantage to companies that can provide power, natural gas pipelines, generation, data center shells, or supporting infrastructure; third, enterprise AI adoption has attractive ROI, and relatively low token costs versus savings will continue to support compute demand; fourth, compute demand growth could significantly exceed supply forecasts for key AI chips from NVIDIA, Google, Huawei, and others; fifth, the recent decline in “time to power” stocks creates a buying opportunity.

Analysis framework

The report combines thematic and equity research, mapping Meta cloud business news, neocloud transaction structures, token economics, enterprise AI ROI, compute supply and demand, data center power contracts, and energy infrastructure projects to related stocks. Its focus is not on earnings forecasts for a single company, but on assessing how AI compute shortages and power bottlenecks affect bitcoin miner conversions, power generators, natural gas infrastructure, equipment manufacturers, data center REITs, and neocloud companies.

Methodology notes

• Thematic investingTime to Power

  Ability to provide power in a timely manner

  This framework focuses on who can most quickly provide usable power, data center shells, natural gas supply, power generation equipment, and related infrastructure to AI data centers. The report believes power availability is a key bottleneck for compute expansion.

• Unit economicsToken Economics

  Comparison of enterprise AI usage costs and benefits

  Using an example in which an enterprise AI task saves approximately $55 while the token cost to execute it is approximately $2-5, the report demonstrates that enterprise AI adoption has favorable economics, supporting token and compute demand.

• Supply and demand analysisCompute Demand vs Supply

  Compute demand growth relative to chip supply growth

  The report compares compute demand growth with AI chip supply forecasts, concluding that demand is growing far faster than current sales forecasts for key AI chips, and that even meeting those forecasts would create power challenges.

• Project returnsPowered Shell Lease Math

  Value creation from powered-shell data center contracts

  The report compares 15- to 25-year renewable energy PPAs with powered-shell data center leases, concluding that the latter offer higher value creation per GW and higher equity investment returns, which the market has not fully priced in.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• META PLATFORMS INC (US.META)

  News trigger and potential cloud business participant

  Strengths

  Owns substantial owned and leased compute capacity and can use temporary leasing of owned capacity to create an EPS bridge; the core investment thesis remains AI products, platform engagement, and revenue growth.

  Weaknesses

  A shift toward a full-scale long-term cloud business could increase 2027/2028 capital expenditures; third-party leased capacity may not be subleased.

  Comparison

  The report emphasizes that Meta is not a pure-play neocloud investment; its compute leasing resembles a temporary stopgap rather than a full hyperscale business.

  Risks

  Uncertainty around cloud business development, higher capital expenditures, and Meta has not confirmed the media reports.

• COREWEAVE INC (US.CRWV)

  Neocloud peer and one of Meta's third-party compute lessors

  Strengths

  Benefits from AI compute shortages and an environment of high leasing prices.

  Weaknesses

  Entry by large technology companies into cloud computing could limit upside in neocloud valuations.

  Comparison

  The report rates CRWV Equal-weight rather than as a preferred Overweight.

  Risks

  Competition from new entrants, limited upside in cloud capital expenditures, and shorter or more flexible lease terms.

• NEBIUS GROUP NV (US.NBIS)

  Neocloud peer and one of Meta's third-party compute lessors

  Strengths

  Benefits from tight compute supply and demand and expanding enterprise AI demand.

  Weaknesses

  Large, well-funded new entrants could weaken neocloud pricing power.

  Comparison

  The report rates NBIS Equal-weight and is more cautious on it than on data center REIT EQIX.

  Risks

  Intensifying competition, declining contract prices, and weaker-than-expected demand growth.

• WMB

  One of the preferred expressions of behind-the-meter power solutions for AI data centers

  Strengths

  Has integrated capabilities spanning natural gas pipelines, gas supply, generation, fiber, batteries, and carbon capture; strong protection from long-term take-or-pay contracts; management has leading capabilities in gas turbine reservations and project execution.

  Weaknesses

  Growth capital expenditures could rise significantly, and the market needs further project clarity.

  Comparison

  The report views WMB as a safer way to express the AI growth theme through long-term contracted natural gas pipelines and power infrastructure.

  Risks

  Project announcement timing, construction execution, customer concentration, financing structure, and regulatory approvals.

• EQIX

  Preferred data center REIT

  Strengths

  Demand is driven by AI inference and agentic compute; 60% of the largest transactions in 1Q26 were AI-related, while 8 of the top 10 AI model providers and 4 of the top 5 neocloud companies are expanding at its facilities.

  Weaknesses

  Data center REIT exposure is mixed amid the cloud capital expenditure cycle and new competitive dynamics.

  Comparison

  The report rates EQIX Overweight and calls it Cameron McVeigh's Top Pick.

  Risks

  New large cloud entrants suppressing upside in cloud capital expenditures and capacity expansion changing pricing.

• CVX

  Energy company expanding power supply to data centers

  Strengths

  Project Kilby will provide Microsoft with 2.67GW of natural gas power generation under a 20-year PPA, supported by Permian natural gas and infrastructure.

  Weaknesses

  The project is large and does not start until 2028, creating execution and construction-cycle risks.

  Comparison

  Together with XOM, EQT, and others, it represents an energy-company pathway to serving AI data center power demand.

  Risks

  Project permitting, costs, natural gas prices, customer demand, and engineering delivery.

• XOM

  Low-carbon natural gas generation plus CCS data center solution

  Strengths

  Its Gulf Coast footprint and liquid CO₂ pipeline system support the expansion of gas-plus-CCS solutions.

  Weaknesses

  The first low-carbon data center project remains in customer negotiations and before FID.

  Comparison

  Compared with CVX's announced Microsoft PPA, XOM's solution is more oriented toward an integrated low-carbon power platform.

  Risks

  FID delays, CCS commercialization, customer contracting, regulatory issues, and project costs.

• LBRT

  Power as a Service and on-site generation provider

  Strengths

  Plans to deploy approximately 3GW of capacity by 2029 and has reached a 1GW agreement with Vantage Data Centers; its owned mobile generation, natural gas logistics, and innovative culture support execution.

  Weaknesses

  The value of the new business has not yet been fully validated by the market, and its technology approach is oriented toward gas reciprocating engines.

  Comparison

  The report estimates that each 1GW of deployed generation capacity could generate approximately $5-6 per share of value after net capital expenditures for LBRT.

  Risks

  Contract acquisition, equipment delivery, customer concentration, execution capability, and capital expenditures.

Key data

• Estimated increase in Meta-owned IT capacityApproximately 2GW in 2026 and 3.5GW in 2027Based on the report's estimate of incremental Meta-owned operated IT capacity, with an approximately 3GW base at the end of 2025.
• Estimated Meta third-party leased capacityApproximately 2.5GWThe report believes Meta may not be able to sublease capacity from third parties such as Coreweave, Nebius, GCP, and ORCL, but this capacity increases its flexibility to lease out owned capacity.
• Meta compute leasing EPS sensitivityEach 250MW leased for one year at $40/Watt could add approximately $3, or about 8%, to 2028 EPSUsed to illustrate the potential earnings bridge from the short-term monetization of scarce compute resources.
• Meta capital expenditure assumptions$145bn in 2026, $175bn in 2027, and $205bn in 2028If Meta develops a long-term cloud business more aggressively, capital expenditures in 2027/2028 could have upside bias.
• Examples of recent pullbacks in related stocksWULF -19%, CIFR -22%, SEI -15%, VST -10%The report says these stocks declined significantly last week due to Meta cloud business news.
• Enterprise AI task economics exampleAverage savings of $55 per AI use case, with execution costs of approximately $2-5Assumes 5-7 AI Agents, 375,000-525,000 tokens, and an average token price of $3-10.
• Future chip efficiency assumptionNVIDIA's new chips improve exaFLOPs by 250% per generation while power consumption increases by only 60%Used to explain how higher-performance chips could reduce token prices and trigger Jevons Paradox-style demand expansion.
• Powered-shell contract returnsCosts of approximately $10-12/Watt, 15- to 25-year fixed-price contracts, and unlevered FCF yields of 15-19%The report believes bitcoin-to-data-center companies have not fully reflected this value creation.
• WMB target price and potential returnTarget price of $98, implying a potential one-year total return of +36.9%, including a 2.9% dividend yieldWMB is listed as a high-conviction Overweight within North American Midstream & Renewable Energy Infrastructure.
• WMB project scale and paceCommercialization of 3-4 new projects annually, with each project approximately $1-2bn, approximately 500MW, and delivered in 18-24 monthsThe report expects WMB may announce another 2-3 new data center projects in the near term.
• CVX projectProject Kilby: 2.67GW, 20-year PPA, expected to commence in 2028CVX signed a power purchase agreement with Microsoft for a West Texas natural gas power generation data center project.
• LBRT target deploymentApproximately 3GW of power capacity by 2029LBRT has reached an agreement with Vantage Data Centers for a 1GW power solutions project, including a 400MW firm reservation by 2027.

Impact & implications

The investment implication is that Meta's entry into cloud computing should not be simply interpreted as a negative competitive shock to neocloud and AI infrastructure stocks. Morgan Stanley focuses instead on its validation of the value of scarce compute and power: if selling compute can generate high returns and enterprise AI demand continues to expand, the scarcest links may remain power, data center construction capacity, and energy infrastructure. In the near term, the related stock pullback caused by the news may provide buying opportunities; in the medium term, power and infrastructure companies able to sign long-term contracts with high credit quality and fixed prices or escalation mechanisms have room for re-rating.

Risks

• Meta has not commented on the media reports, and its cloud business plans remain uncertain.
• Large, well-funded, technically capable new entrants could enter the cloud compute market, constraining upside in neocloud and cloud capital expenditures.
• Data center expansion is constrained not only by power but also by skilled labor, permitting, and political factors.
• If enterprise AI adoption, token usage, or compute demand growth falls below expectations, returns from the “time to power” theme could decline.
• Related projects are generally capital-intensive and have long construction cycles, creating execution, financing, customer credit, and contract-structure risks.
• If Meta develops a full-scale long-term cloud business, its 2027/2028 capital expenditures could increase.

What to watch

• Whether Meta formally confirms plans for a cloud business, API/model access, or raw silicon offerings.
• Whether Meta raises its 2027/2028 capital expenditure guidance and expands long-term owned data center capacity.
• Announcements in coming quarters regarding WMB's behind-the-meter data center power projects, contract duration, and new customers.
• The pace of contract execution for CVX's Project Kilby, XOM's low-carbon data center power project, EQT's Well-to-Watt, and LBRT's Power as a Service.
• Whether neocloud transaction prices, terms, exit provisions, and customer composition continue to indicate compute scarcity.
• Whether AI inference, AI Agents, and enterprise AI adoption continue to drive strong token demand growth.
• Whether power, gas turbines, permitting, and data center construction cycles continue to be supply bottlenecks.