US Data Center Regulation Tightening Hits AI Infrastructure Stocks? Morgan Stanley: Limited Impact in 2027, Risks More Concentrated in 2028

According to The Zhihu Financial APP, Morgan Stanley pointed out in its latest research report released on November 9 that the results of the U.S. midterm elections have heightened market concerns about the approval and regulatory policy changes for data center construction, leading to a significant decline in stocks related to artificial intelligence (AI) infrastructure. However, the bank believes that the market may be overreacting to policy risk, as the demand for AI infrastructure remains robust, and risks to the planned expansion of data centers by 2027 are relatively limited, with potential impacts mainly concentrated in 2028, especially for certain projects in Texas and Ohio.

Morgan Stanley noted that over the past week, the AI infrastructure-related stocks it covers have fallen by an average of about 12%, with a cumulative decline of about 28% over the past three months. The hardest hit are emerging AI cloud service providers, who are being impacted not only by regulatory uncertainty but also by rising concerns over their financing capacity.

Following the U.S. midterm elections, the data center development policies in some states face adjustments, with the market worried that local governments may tighten approvals for large-scale data center projects, slowing construction progress. Morgan Stanley cited research showing that some data center projects in Texas and Ohio face greater policy risk, but these risks are mainly concentrated in 2028 and should have little impact on the planned delivery capacities for 2027.

Notably, Morgan Stanley believes that even if the pace of data center construction slows moderately in 2028, it may not necessarily constitute a comprehensive negative for the share prices of related companies. Given the continued strong demand for AI computing power, a slower pace of new supply may keep the supply-demand relationship tight, thus enhancing cloud service providers’ pricing power. At the same time, companies may be able to demonstrate the returns on infrastructure investments to investors more quickly, alleviating concerns over the scale of AI capital expenditures and profitability.

Against this backdrop, Morgan Stanley continues to list Microsoft (MSFT.US) as its top pick for capturing AI infrastructure investment opportunities, believing that its global data center presence, the operational efficiency of existing facilities, and the flexibility of its computing power deployment help to mitigate localized policy risks.

According to the research, Microsoft's data center capacity will increase by more than 2 GW in fiscal 2025, with about 1 GW to be added in just the fourth quarter of fiscal 2026. Morgan Stanley expects Microsoft’s installed data center capacity to grow from approximately 11 GW in fiscal 2026 to over 20 GW by fiscal 2028.

Although Microsoft’s Pecos Data Center Campus and Abilene project in Texas may be affected by local approval policies, its data center coverage extends across multiple states in the U.S. and overseas markets, including Norway and Portugal. Morgan Stanley believes that this geographically diversified layout enables Microsoft to adjust computing power deployment according to project progress, and to offset the impact of delays in new projects by improving the utilization efficiency of existing data centers.

Therefore, Morgan Stanley maintains its capital expenditure forecasts for Microsoft in fiscal years 2027, 2028, and 2029 at about $238 billion, $298 billion, and $358 billion, respectively, corresponding to year-on-year increases of about 64%, 25%, and 20%. It also retains its “Overweight” rating and a target price of $650 for Microsoft, believing that the market has yet to fully price in Microsoft’s AI business leadership and long-term profit growth potential.

For Oracle (ORCL.US), Morgan Stanley notes that, while several of its large U.S. data center projects face delivery delay risks, the company's previously announced long-term performance outlook has already allowed for a buffer for some project delays.

Oracle plans to increase its total data center capacity to about 10 GW by fiscal 2028, up sharply from approximately 3 GW in fiscal 2026. Its five major U.S. data center projects collectively account for about 5 GW in capacity, with the Texas Abilene project delivering approximately 618 MW by the first quarter of fiscal 2027, a completion rate of about 75%.

The remaining four large projects are expected to come online in succession over 2027. Morgan Stanley points out that Oracle management has factored into its guidance the possibility of these projects being delayed by up to two quarters, and that projects are usually delivered in phases, with capital spending and lease commitments increasing gradually with revenues, thus reducing the risk of idle funds due to delays.

In addition, Oracle also holds a large number of smaller data center projects, helping spread the construction risk of its large projects. The bank maintains its “Equal-weight” rating for Oracle, with a target price of $210, but reminds investors to consider Oracle's significant infrastructure investment needs and financing pressures.

For emerging AI cloud service providers CoreWeave (CRWV.US) and NEBIUS (NBIS.US), Morgan Stanley believes that both are unlikely to be directly and significantly affected by changes in local U.S. regulatory policy in the short term, though financing risk is still worth watching.

CoreWeave’s current projects in Texas and Georgia have mostly been put into operation or are nearing completion. The Helios Data Center project in west Texas, however, still warrants close attention. The company has committed to deploying 526 MW of critical IT power capacity at this site, with the first stage of 133 MW delivered in July this year, and further construction scheduled until 2028.

Morgan Stanley points out that tighter local government approvals for data center construction may change project locations, but may not necessarily reduce the overall demand for AI infrastructure. CoreWeave is also accelerating its overseas expansion, including increasing its data center capacity in India and Indonesia, to reduce reliance on certain regions in the U.S. However, the company still faces relatively high debt burden and financing cost risks. Morgan Stanley assigns CoreWeave an “Equal-weight” rating, with a target price of $70.

As for NEBIUS, the company holds data center projects of over 1 GW in Missouri and Pennsylvania, and is laying out computing power infrastructure in overseas markets such as Europe. Since the disclosed projects are not in the main regions affected by the post-election policy risks, Morgan Stanley believes its direct risk is relatively low. However, possible tighter data center development regulations in Pennsylvania still pose some uncertainty for larger local projects. The bank maintains an “Equal-weight” rating for NEBIUS, with a target price of $144.

By contrast, Morgan Stanley believes the market's sell-off of Akamai (AKAM.US) and DigitalOcean (DOCN.US) may be excessive, since their approaches to infrastructure expansion differ from those of large AI cloud service companies.

Akamai previously stated that it has secured the required data center hosting sites and power resources for a major $11.6 billion AI customer contract. This means that the execution risks for its contracted deals are relatively controllable, with these businesses expected to drive revenue growth through 2028.

DigitalOcean, on the other hand, primarily expands its computing capacity by leasing existing data center space rather than building concentrated gigawatt-scale facilities. The company has signed on for 75 MW of new capacity to go online in 2027, plus an additional 30 MW for 2028. Since individual projects are small, its exposure to risks associated with large data center approval restrictions is limited, though inadequate disclosure about future expansion sites still brings some uncertainty.

Morgan Stanley gives DigitalOcean an “Overweight” rating with a $150 target price, noting that its cloud services for AI inference and agent applications are likely to benefit from growing demand, whose potential has yet to be fully reflected in the share price.

Overall, Morgan Stanley emphasizes that the main issue facing the current AI infrastructure market is not a structural decline in demand, but rather the potential changes in project construction timelines and capital spending deployment. Despite persistent policy uncertainty induced by the midterm elections that could continue to affect market sentiment, the sector's sharp pullback has already made the risk-reward profile of some stocks more attractive given continued strong demand for AI computing power.

The bank believes investors should focus more on companies’ global computing power layout, the utilization efficiency of existing facilities, project delivery capabilities, and financing conditions, instead of regarding changes in local approval policy as a signal of deteriorating growth prospects for the entire AI infrastructure industry.