As demand for AI model training and inference grows rapidly, global technology companies continue to commit enormous sums to building large data centers and related computing infrastructure. However, since the U.S. 2026 midterm elections, local governments’ approach to data center development has become a growing focus for markets. In addition to the supply of GPUs and server equipment, land-use approvals, access to power, and support from local communities are emerging as important factors in the expansion of AI computing capacity.
Meta President Dina Powell McCormick recently said the election results could lead to stricter reviews of new data center projects in the United States, while expansions at existing sites may have an advantage. At the same time, Anthropic reportedly spent $3.4 billion to secure orders for 1.7 GW of natural gas generation equipment and related engineering capabilities. This reflects a shift among AI companies from simply purchasing computing equipment toward investing more directly in energy infrastructure.
For investors, competition in AI infrastructure is expanding beyond the ability to procure GPUs to include land-use approvals, power resources, and the efficiency of deploying computing capacity. The scale of future technology-company capital spending may not directly reflect how much AI computing capacity can actually be brought online. Project development timelines and energy-related costs will be important variables in assessing investment returns.
AI Data Centers Face New Local Regulatory Challenges After the U.S. Midterm Elections
Meta Warns That Building New Data Centers Will Become More Difficult
The U.S. 2026 midterm elections took place on November 3. Energy, land-use, and community issues arising from data center development were important topics in some local races. Speaking at The Wall Street Journal’s Tech Forum on November 4, Meta President Dina Powell McCormick said new data center projects could face more stringent local reviews following the elections. Some incoming local leaders had already called during their campaigns for restrictions or a pause on large data center developments, increasing uncertainty for technology companies seeking permits.
Meta, however, believes that expanding existing data centers may be relatively advantageous. Local governments and communities are generally more familiar with the employment, tax revenue, and infrastructure impacts of long-running data centers. Companies may also have already secured some land-use approvals and established arrangements for power and water supply, reducing the cost of negotiations during expansion.
This suggests that the key conditions for data center development may shift beyond land costs and location to include a company’s ability to secure local government support, grid capacity, and a predictable construction schedule. The election results do not mean that every state will impose restrictions on data center development. Policies still need to go through formal processes at the local council, state government, and regulatory-agency levels. Campaign positions may differ from restrictions that are ultimately enacted.
From Virginia to Texas: Data Center Regulatory Risks Are Becoming More Uneven
U.S. data centers are concentrated in areas with established power grids, fiber-optic networks, and cloud-service infrastructure. As AI data centers grow in scale, however, some mature markets are facing constraints on land and grid capacity, pushing development into other states.
| Key region | Role in the data center industry | Development constraints to watch |
|---|---|---|
| Virginia | Northern Virginia is one of the world’s major cloud data center hubs | Zoning, grid expansion, noise, and local opposition |
| Texas | A major hub for large AI data centers and energy infrastructure investment | Power access, reviews of energy facilities, and post-election policy changes |
| Ohio | Meta, Amazon, and other companies continue to expand their presence | Local development pauses and land-use restrictions |
| Georgia | A growing data center market in the southeastern United States | Tax incentives, local development restrictions, and allocation of grid costs |
| Michigan | An emerging destination for large data centers | Local land-use rules and community acceptance |
| Pennsylvania | Data center investment and demand for power supply are increasing | Development approvals, energy infrastructure, and local oversight |
| Arizona | The Phoenix metro area has an established data center industry | Water resources, cooling needs, and zoning changes |
Regulatory concerns vary by state. In mature markets such as Virginia, development constraints are more often related to competition for land, grid capacity, and resources in established communities. In Texas, Ohio, and other emerging markets, developers must also assess large electricity loads, infrastructure development, and changes in local government policy.
As a result, data center site-selection strategies may shift from prioritizing low land and energy costs toward a broader assessment of power availability, certainty of permitting, and speed of deployment.
Local Governments Are Introducing Development Moratoriums, Adding Uncertainty to Data Center Timelines
Several Local Governments Have Already Passed Data Center Moratoriums
Restrictions on large data centers did not begin after the elections. Since the start of 2026, several local governments have used development moratoriums to pause the acceptance or approval of certain types of data center applications while they review land-use, energy, and environmental rules.
| Location | Date passed | Moratorium period | Specific development |
|---|---|---|---|
| Oklahoma City | April 21 | Through the end of 2026, or until new rules are passed earlier | Took effect immediately; paused new applications and permits |
| Montgomery County, Maryland | June 12 | 6 months | Executive order took effect immediately while permanent rules are developed |
| Dane County, Wisconsin | June 4 | 18 months | Paused zoning permits for large data centers |
| Linn County, Iowa | July 1 | 18 months | Paused rezoning applications for large data centers |
| Raleigh, North Carolina | October 6 | 6 months | Review of new rules scheduled to be completed by April 6, 2027 |
| Prince George’s County, Maryland | July 7 | Up to 2 years | Paused review of certain data center development plans |
The types of data centers covered, the affected land, and exemptions for existing projects vary by measure. These actions should not be treated as a blanket ban on data center construction. However, they indicate that local governments are shifting from a traditional focus on attracting investment through tax incentives toward closer review of energy use, environmental impacts, and community interests.
Why Are Local Governments Opposing Large AI Data Centers?
Large AI data centers can support construction investment, tax revenue, and some job creation. Compared with traditional industries, however, they can also require enormous amounts of energy and infrastructure, potentially imposing additional costs on local residents.

The main concerns are:
| Issue | Concerns of local governments and residents | Potential impact on developers |
|---|---|---|
| Power demand and electricity prices | Whether new grid costs will be passed on to ordinary users | Higher power-access costs and longer review times |
| Land use | Large industrial facilities near homes; conversion of farmland | Zoning changes may be blocked or sites may need to be reconsidered |
| Water resources | Whether cooling systems will compete with residents and agriculture for water | More environmental reviews and higher cooling-system costs |
| Noise and emissions | Impacts from cooling equipment, backup generators, and construction | Higher spending on noise control, emissions controls, and monitoring |
| Economic benefits | Whether large investments will create enough long-term jobs | Greater scrutiny of tax incentives and community support |
| Transparency | Limited public information on project power demand and local agreements | More time spent on public hearings, litigation, and negotiations |
In the past, data center developers generally obtained power through utility companies, which typically planned and built the related transmission and distribution infrastructure. As demand from large AI data centers rises rapidly, however, the question of whether technology companies or ordinary electricity users should bear the cost of grid upgrades, substations, and new generation capacity has become a growing policy dispute.
Data centers can create many jobs during construction, but staffing needs are relatively limited once they are operational. This has led some local governments to focus more closely on tax revenue, community investment, and long-term public benefits.
The core issue in local regulation is not whether to support AI development, but how to distribute the economic benefits and social costs of large infrastructure investments.
AI Giants Begin Taking on Energy Costs Directly, Making Power a New CapEx Priority
In addition to obtaining power through utility companies, some AI companies are beginning to get directly involved in purchasing generation equipment. According to the Financial Times, Anthropic has reached a deal worth approximately $3.4 billion with U.S. equipment supplier OnePWR Solutions. The agreement includes commitments for large gas-fired generators with a total capacity of approximately 1.7 GW, along with engineering and construction capabilities, to secure power for Anthropic’s AI data centers.
The equipment was originally ordered by OnePWR from Caterpillar. With demand for natural gas generators rising rapidly and delivery times for new equipment potentially stretching to several years, Anthropic reduced the risk of waiting in a new procurement queue by securing existing orders.
The deal reflects an important shift in the AI industry: companies are no longer focused only on GPU procurement and delivery schedules. They are also seeking direct control of generation equipment and energy infrastructure to ensure that future computing-capacity expansion can proceed. However, energy investment also means higher upfront capital requirements and ongoing costs for fuel, maintenance, emissions compliance, and depreciation.
For markets, a full analysis of future AI infrastructure investment will need to include not only GPUs and servers but also generation equipment, substations, transmission and distribution systems, and cooling systems.
Overseas Data Center Investment Picks Up as Technology Companies Diversify U.S. Development Risks

Meta Considers Expanding Overseas Investment to Reduce Local Regulatory Uncertainty
Against the backdrop of potentially tighter local restrictions on U.S. development, Meta has begun reassessing where to locate large data centers. Meta has major data center projects in Louisiana, Ohio, and Texas, but growing uncertainty around local approvals and infrastructure construction could lead the company to reallocate capital among expansions of existing facilities, new campuses, and overseas investment.
Meta management recently said large cloud providers are considering increasing investment in data centers outside the United States to reduce their exposure to development restrictions in any one market. Overseas expansion can diversify local regulatory risks, but it also brings new construction and operating challenges, including cross-border data regulations, the reliability of energy supply, network latency, tax systems, and operating costs. Overseas investment may therefore not reduce total data center construction costs directly; its main value may be to provide additional options for deploying computing capacity.
ByteDance Considers Expanding Its Brazil Data Center
At the same time, ByteDance is reportedly considering expanding its data center project in Brazil’s northeastern state of Ceará. According to Bloomberg, the company is evaluating a second phase and looking for development partners. The existing project involves an investment of around $10 billion and a capacity of 300 MW. A potential expansion could exceed the first phase, but no confirmed investment amount, capacity, or completion date has been announced.
ByteDance’s plans in Brazil should not be attributed directly to local U.S. regulations. Tax incentives, energy conditions, and regional market demand may also be major considerations. Still, the moves by Meta and ByteDance show that global AI infrastructure planning is paying closer attention to differences in energy supply, policy environments, and investment terms. Markets that offer reliable power, clearer development processes, and competitive energy costs may gain an advantage in the global data center investment race.
Industry Analysis: How Will Regulatory and Power Constraints Change the Investment Case for AI Infrastructure?
Development Permits and Power Capacity Become Scarce Resources, Raising Barriers to Entry
In the past, competition in AI data centers focused mainly on GPU procurement, access to capital, and construction efficiency. As local governments increase scrutiny, land and facilities that already have development permits and grid-access arrangements may become more valuable. In mature data center hubs, new projects often need to obtain zoning approval, grid capacity, and local government support. Existing facilities may already have these foundations in place and therefore have an expansion advantage.
This could give companies with existing data center campuses, substantial rights to grid capacity, and established relationships with local governments a competitive advantage in expanding computing capacity. Existing data centers, however, cannot expand without limit. Even with land-use permits in place, companies may still face restrictions on power interconnection, cooling capacity, and equipment configurations. The resources with real competitive value are therefore not simply land and building permits, but data center capacity that can be expanded and powered in practice within a reasonable timeframe. If restrictions on new projects persist, companies with these resources may benefit from greater asset scarcity, raising barriers to entry for new competitors.
More Energy Infrastructure Investment Could Increase AI CapEx and Long-Term Operating Costs
Market discussion of Big Tech’s AI CapEx has traditionally focused on GPUs, servers, networking equipment, and data center buildings. As power-interconnection constraints emerge, however, spending on new generation equipment, substations, transmission and distribution systems, and related engineering may account for a larger share of capital budgets. These investments can give companies more control over energy supply, but they also increase upfront capital requirements and ongoing operating costs. In particular, natural gas generation requires not only construction spending but also fuel purchases, maintenance, regulatory compliance, and depreciation. The actual total cost of ownership for AI computing capacity may therefore be higher than originally expected.
More energy investment does not necessarily mean that the final capital efficiency of AI computing will improve at the same pace. Going forward, investors should compare the capital required to add power capacity, the cost of building computing capacity per unit, and the revenue and cash flow generated after new capacity becomes operational.
AI CapEx May Become Further Decoupled from Computing Capacity That Is Actually Available
Large data center projects depend on multiple interconnected stages, including land acquisition, development permits, construction, energy supply, GPU delivery, and system testing. Even if a company has completed GPU procurement or signed long-term computing-capacity contracts, delays caused by land reviews, grid interconnection, or local opposition may prevent the facility from coming online as planned, leaving the investment unable to generate revenue immediately.
These delays can cause a company to incur financing costs, equipment payments, and other fixed expenses before the facility generates revenue or cash flow. For AI infrastructure operators that rely heavily on new capacity and external financing, project delays may create more pronounced financial pressure.
When analyzing AI infrastructure investment, announced investment should not be treated as a direct proxy for future computing capacity. The full progression—from contracted power capacity to planned data center capacity and then to operational IT load—affects how much computing capacity ultimately comes online. Tracking capacity that is actually operational and the revenue it generates is the most reliable way to assess capital efficiency.
Impact on U.S. Stocks: Which Companies Are Most Sensitive to Data Center Development Restrictions?
The impact of regulatory and power constraints varies by company, depending mainly on its business model, capital structure, existing data center resources, and reliance on new capacity. Overall, project developers and leasing companies, along with emerging AI cloud providers, may be more vulnerable to delays in individual projects than established large cloud platforms.
| Company | Potential impact level (research assessment) | Main channels of impact |
|---|---|---|
| IREN (IREN) | High | Progress of Texas AI data center projects, power capacity, and customer delivery |
| Applied Digital (APLD) | High | Large data center development, lease revenue, and project financing |
| CoreWeave (CRWV) | High | GPU deployment efficiency, data center delivery schedules, and capital costs |
| Nebius (NBIS) | Medium-high | New AI cloud capacity, construction costs, and funding needs |
| Core Scientific (CORZ) | Medium-high | Conversion of existing facilities, power access, and HPC project construction |
| Meta (META) | Medium-high | AI CapEx allocation, overseas expansion, and approval of new campuses |
| Microsoft (MSFT) | Medium | New Azure computing capacity and delivery by third-party partners |
| Amazon (AMZN) | Medium | AWS capacity additions, energy infrastructure, and community development costs |
| Alphabet (GOOGL) | Medium | Google Cloud expansion and spending on energy infrastructure |
| Nvidia (NVDA) | Medium, indirect | Customer data center deployment speed and GPU purchasing schedules |
Note: Potential impact levels are qualitative research assessments based on each company’s business model. They are not quantitative forecasts of share-price declines or the probability of delays to specific projects.
AI Data Center Developers: Watch Capacity Delivery and Financing Costs
For companies such as IREN, Applied Digital, CoreWeave, Nebius, and Core Scientific, revenue growth is often closely tied to the amount of new data center capacity they bring online. If local regulations delay projects, expected lease or computing-service revenue may be pushed back, increasing the amount of capital tied up during construction and the burden of interest costs.
Companies that have already purchased GPUs in advance or signed long-term equipment contracts may also face a mismatch between hardware delivery and data center readiness. Investors should therefore monitor whether capacity covered by customer contracts can be delivered on schedule, as well as the impact of project delays on liquidity and free cash flow.
Big Tech: Financial Strength May Become a Greater Advantage, but Investment Efficiency Remains a Challenge
Meta, Microsoft, Amazon, and Alphabet have more mature data center networks, energy partnerships, and sources of funding. They may be better positioned to reduce their exposure to any one project through existing-site expansions, overseas investment, and direct energy investment.
However, these companies may also need to spend more to secure power and construction resources, further raising the total cost of AI infrastructure investment. In particular, while AI-related revenue has yet to fully offset the associated capital spending, additional energy investments could put further pressure on free cash flow.
Nvidia: Demand May Remain, but GPU Orders and Deployment Schedules Could Be Affected
For Nvidia, restrictions on data center development do not necessarily signal weaker demand for AI computing. They could, however, affect the timing of customers’ GPU purchases, deliveries, and deployments. If planned data centers cannot secure power on schedule, technology companies may reschedule hardware purchases or adjust how computing capacity is distributed across regions.
When assessing Nvidia’s future data center business, investors should therefore track not only demand for AI models and customer CapEx but also end-market data center power availability, construction progress, and whether major customers are adjusting GPU delivery plans.
What to Watch Next: Shift from CapEx Scale to Deployment Efficiency
As data center development faces growing energy and local regulatory constraints, investors need to broaden their analysis beyond AI CapEx growth rates and assess infrastructure investment efficiency more fully.
| Area | Key indicators to track | Investment relevance |
|---|---|---|
| Local regulation | Time to obtain permits, moratoriums, and zoning approvals | Assesses the risk of delays to new capacity |
| Power supply | Contracted MW, MW that can be physically connected, and interconnection timelines | Assesses whether computing capacity can be deployed |
| Construction progress | Planned capacity, capacity under construction, and capacity already online | Shows the gap between announced investment and actual supply |
| Energy costs | Power-interconnection cost per MW; generation and transmission CapEx | Assesses cost pressure from energy investment |
| Capital efficiency | Construction cost per unit of available computing capacity; asset turnover | Assesses how efficiently new capital spending is used |
| Operating performance | Computing utilization, revenue per unit of capacity, and operating cash flow | Measures the commercial returns from new capacity |
| Financial risk | Free cash flow, financing costs, capital commitments, and cash on hand | Assesses the company’s ability to absorb project delays |
One of the most important indicators is whether the gap between committed investment and capacity that is actually operational continues to widen.
If technology companies continue to increase AI CapEx but power connections and deployment do not keep pace, more of the new funding may be going to infrastructure that is not directly related to computing, such as energy supply, land approvals, and engineering. These investments may be necessary over the long term, but they could extend the capital-recovery period and reduce short-term returns on invested capital.
By contrast, companies that can secure reliable power at lower construction cost and with shorter approval timelines, then quickly commercialize their computing capacity, may have higher capital efficiency and a stronger competitive position.

