Is the tightening of US data center regulations impacting AI infrastructure stocks? Damo: Limited impact in 2027 Risks focus more on 2028

The Zhitong Finance App learned that Morgan Stanley pointed out in the latest research report released on November 9 that the US midterm election results have heightened market concerns about data center construction approvals and changes in regulatory policies, and driven a sharp decline in stocks related to artificial intelligence (AI) infrastructure. However, the bank believes that the market may be overreacting to policy risks, demand for AI infrastructure is still strong, and the risks faced by the 2027 data center expansion plan are relatively limited. The potential impact is mainly concentrated in 2028, especially some projects in Texas and Ohio.

Damo said that in the past week, the AI infrastructure-related stocks it covers fell by an average of about 12%, with a cumulative decline of about 28% over the past three months. Among them, the impact on emerging AI cloud service providers was the most obvious. In addition to uncertainty about regulatory policies, market concerns about their ability to finance have also heated up.

After the US midterm elections, data center development policies in some states faced adjustments. The market feared that local governments might tighten the approval of large-scale data center projects, leading to a slowdown in construction progress. Citing relevant research, Damo pointed out that some data center projects in Texas and Ohio face greater policy risks, but these risks are mainly concentrated in 2028 and have little impact on the scale of computing power planned to be delivered in 2027.

It is worth noting that Damo believes that even if the pace of data center construction slows moderately in 2028, it will not necessarily be completely bad for the stock prices of related companies. As demand for AI computing power is still strong, the slowdown in new supply may keep the supply-demand relationship tight, thereby enhancing the pricing capabilities of cloud service providers. At the same time, companies may also be able to show investors the return on infrastructure investments more quickly as a result, and ease market concerns about the scale and profitability of AI capital expenditure.

In this context, Morgan Stanley continues to list Microsoft (MSFT.US) as the preferred target for seizing investment opportunities in AI infrastructure, believing that its global data center layout, operational efficiency of existing facilities, and flexible computing power deployment capabilities will help to resist local regional policy risks.

According to research reports, Microsoft added more than 2 gigawatts of data center capacity in fiscal year 2025, and added about 1 gigawatt in the fourth quarter of fiscal year 2026 alone. Damo expects Microsoft's installed data center capacity to grow from about 11 gigawatts in fiscal 2026 to more than 20 gigawatts in fiscal 2028.

Although Microsoft's Pecos data center campus and Abilene project in Texas may be affected by local approval policies, its data center layout covers many US states and overseas markets, including Norway and Portugal. Damo believes that this geographically diversified layout enables Microsoft to adjust computing power deployment according to project progress, and at the same time make up for the impact of delays in some new projects by improving the utilization efficiency of existing data centers.

As a result, Damo kept Microsoft's capital expenditure forecast unchanged for the 2027, 2028, and 2029 fiscal years, which were approximately US$238 billion, US$298 billion, and US$358 billion, respectively, corresponding year-on-year increases of about 64%, 25%, and 20%. The bank also maintains Microsoft's “plus” rating and a target price of $650, believing that the market has yet to fully reflect its AI business leadership position and long-term profit growth potential.

Regarding ORCL.US (ORCL.US), Damo believes that although there is a risk of delays in delivery of many of its large-scale US data center projects, the long-term performance outlook previously announced by the company has reserved buffer space for delays in some projects.

Oracle plans to increase the total data center capacity to about 10 gigawatts by fiscal year 2028, a significant increase from about 3 gigawatts in fiscal year 2026. Together, its five large-scale data center projects in the US involved approximately 5 gigawatts of capacity, of which the Abilene Texas project had delivered approximately 618 megawatts as of the first quarter of fiscal year 2027, completing about 75%.

The remaining four major projects are expected to be put into operation one after another in 2027. Damo pointed out that Oracle management has considered the possibility of delaying these projects for up to two quarters in the performance guidelines, and that projects are usually delivered in stages, and capital expenses and leasing commitments will also gradually increase along with revenue, helping to reduce the risk of capital idle due to delays.

In addition, Oracle also has a large number of smaller data center projects, which can diversify the construction risks of large-scale projects. The bank maintains Oracle's “hold and see” rating, with a target price of $210, while reminding investors to pay attention to its huge infrastructure investment needs and financing pressure.

Regarding the emerging AI cloud service providers CoreWeave (CRWV.US) and NEBIUS (NBIS.US), Damo believes that the direct impact of changes in US local regulatory policies in the short term is relatively limited, but financing risks are still worth paying attention to.

Most of CoreWeave's existing projects in Texas and Georgia are in operation or close to completion. Among them, the Helios data center project in western Texas still requires close attention. The company has promised to deploy 526 megawatts of critical IT power capacity at the project. The first phase of 133 megawatts was delivered in July this year, and subsequent construction plans will continue until 2028.

Damo pointed out that local governments are tightening approval for data center construction, which may change the site selection for the project, without necessarily weakening the overall demand for AI infrastructure. CoreWeave has also recently been speeding up overseas expansion, including expanding data center capacity in India and Indonesia to reduce its dependence on parts of the US. However, the company still faces the risk of high debt burdens and financing costs. Damo gave CoreWeave a “wait and see” rating, with a target price of $70.

On the NEBIUS side, the company has more than 1 gigawatt data center projects in Missouri and Pennsylvania, and has deployed computing power infrastructure in overseas markets such as Europe. Since the projects currently disclosed do not cover some of the major regions where policy risk increased after the current gubernatorial election, Damo believes that its direct risk is low. However, Pennsylvania may tighten data center development regulations, which still brings some uncertainty to large-scale local projects. The bank maintains the NEBIUS “hold and see” rating, and the target price is $144.

In contrast, Damo believes that the market may have reacted excessively to the sell-off of AKAM.US (AKAM.US) and DigitalOcean (DOCN.US) because the infrastructure expansion model adopted by the two companies is different from that of large AI cloud service providers.

Akmai said earlier that it has implemented the required data center hosting site and power resources for a large-scale AI customer contract worth 11.6 billion US dollars. This means that the execution risk of its signed contracts is relatively manageable, and related businesses are expected to continue to accelerate revenue growth until 2028.

DigitalOcean mainly expands computing power by leasing existing data center space, rather than concentrating on the construction of gigawatt-scale facilities. The company has contracted 75 megawatts of additional capacity to be put into use in 2027, and an additional 30 megawatts of capacity to be put into use in 2028. Due to the small size of individual projects, the risk of being restricted by large data center approvals is relatively limited, but insufficient disclosure of future expansion locations still brings some uncertainty.

Damo gave DigitalOcean an “gain” rating, with a target price of $150. It believes that its cloud service business for AI reasoning and smart applications is expected to benefit from increased demand, and the relevant potential has not yet been fully reflected in the stock price.

Overall, Morgan Stanley emphasized that the main problem currently facing the AI infrastructure market is not a structural decline in demand, but rather a possible change in the construction time and capital expenditure deployment pace of some data center projects. Although policy uncertainty brought about by the midterm elections may continue to affect market sentiment, against the backdrop of strong demand for AI computing power, the recent sharp correction in the sector has made the risk-reward ratio of some companies more attractive.

The bank believes that investors should pay more attention to the company's global computing power layout, utilization efficiency of existing facilities, project delivery capacity, and financing conditions, rather than viewing changes in local approval policies as a sign that the growth prospects of the entire AI infrastructure industry are deteriorating.