Goldman Sachs Says US Data Center Growth Remains Intact Despite Opposition
The U.S. data center boom is facing growing political and community resistance, but Goldman Sachs expects the expansion to continue with limited near-term disruption.
In a research note released Sunday, Goldman raised its forecast for U.S. data center capacity at the end of 2026 by 5 gigawatts (GW) to 64 GW, while cutting its 2027 forecast by 5 GW to 90 GW. Despite the adjustment, the bank said its broader growth outlook through 2027 remains largely intact.
The conclusion matters well beyond the data center industry. Artificial intelligence is driving enormous demand for computing infrastructure, and data centers are becoming an increasingly important source of electricity consumption. That is creating opportunities for technology companies, utilities, power generators, equipment suppliers and data center operators — while also creating political pressure over electricity prices, water use and local development.
Why Goldman Still Expects Data Center Growth
Goldman's latest assessment suggests that opposition is creating delays and changes to individual projects, rather than stopping the broader U.S. buildout.
The bank's revised estimates imply that developers are still bringing substantial amounts of capacity online, even after accounting for cancellations, delays and political resistance.
Goldman expects U.S. data center power demand to grow by 38% in both 2026 and 2027. That translates into an additional 12 GW of demand in 2026 and 17 GW in 2027.
For investors, this is an important distinction. A project can be delayed or moved from one state to another without destroying the underlying demand for computing capacity.
If AI companies continue to require more computing power, developers may simply shift investment toward locations where electricity, land, permitting and grid connections are easier to obtain.
AI Remains the Core Growth Engine
The fundamental driver behind the U.S. data center boom remains artificial intelligence.
Data centers provide the computing infrastructure needed to train and run AI models. The rapid expansion of AI services by companies such as Microsoft, Amazon, Alphabet, Meta and xAI is therefore translating into demand for additional servers, networking equipment, cooling systems and electricity.
Goldman has previously forecast that U.S. data center power demand could rise from 31 GW in 2025 to 41 GW in 2026 and 66 GW in 2027. Its earlier analysis assumed U.S. data center capacity could reach roughly 95 GW by the end of 2027, although the bank has since revised its capacity outlook.
The latest revision is therefore better understood as a recalibration rather than a fundamental reversal of the AI infrastructure investment cycle.
That distinction is particularly relevant as investors debate whether the enormous capital spending on AI can continue at its current pace.
Political Opposition Is Becoming a Real Constraint
The bullish data center outlook does not mean the political backlash is insignificant.
Public opposition has grown substantially as communities confront the consequences of large computing facilities, particularly their electricity and water requirements.
A June Reuters/Ipsos poll found that only 33% of Americans supported the rapid pace of data center construction, while 64% opposed it. Even more strikingly, just 14% said they would support a data center being built in their own community for technology companies.
That resistance has already translated into policy.
New York became the first U.S. state to impose a one-year moratorium on construction of large new data centers, targeting facilities using at least 50 megawatts of power. The state cited concerns about electricity costs, water supplies and environmental impacts.
More broadly, numerous states and local governments are considering restrictions, new taxes, environmental reviews or rules governing who pays for the additional electricity infrastructure.
This is why Goldman's updated forecast is significant: it suggests the political response is currently more likely to reshape the geography and timing of the buildout than eliminate the underlying demand.
Electricity Has Become the Key Bottleneck
For investors, the most important part of the data center story may no longer be simply how many buildings are planned.
It is whether those facilities can actually obtain enough electricity.
Goldman previously identified power availability, pricing, policy and other constraints as major bottlenecks. In some parts of the U.S., connecting a new data center to the regional electricity grid can take years.
This creates an important competitive advantage for locations that already have available generation and transmission infrastructure.
Goldman's earlier research identified significant regional differences. The Mid-Atlantic, Mid-Continent and Northwest markets faced elevated reliability risks because planned generation additions were relatively limited, while Texas and Georgia had more generation capacity in development.
That means the U.S. data center boom is unlikely to be evenly distributed.
States with abundant electricity, suitable land, faster permitting and strong transmission networks could attract projects that are delayed or cancelled elsewhere.
Why Utilities Could Be Major Beneficiaries
The growth in data center power consumption has potentially important implications for the U.S. electricity industry.
Goldman expects data centers' share of total U.S. peak summer power demand to rise substantially. Its earlier forecast put the share at 8.5% in 2027, compared with 4.1% in 2025.
That creates demand for new generation, transmission infrastructure, transformers, grid equipment and potentially natural gas, nuclear power, renewables and battery storage.
For utilities, the opportunity comes with an important qualification: regulators need to determine how the cost of new infrastructure should be allocated.
That issue has already become politically contentious.
On September 30, the U.S. Senate blocked the Ratepayer Protection Act, which would have required state utility regulators to consider whether large electricity users such as data centers should pay incremental infrastructure costs rather than shifting them toward households. The vote was 57-43, short of the 60 votes required to advance the measure.
The debate highlights one of the biggest risks surrounding the AI infrastructure boom: who ultimately pays for the electricity required to run it?
Data Center Developers Are Having to Respond
The industry is already adapting to the backlash.
Amazon Web Services announced last week that it would invest more than $1 billion over five years in U.S. communities where it operates data centers. The programme includes education, job training, energy affordability and water and energy preservation initiatives.
AWS said it had invested $276 billion in data centers between 2011 and 2025 and noted that more than 100 data center moratoriums were being considered across the country.
This illustrates how the economics of data center development are changing.
Companies can no longer focus only on securing land and electricity. They increasingly need to demonstrate that local residents and governments will benefit from the projects.
What Goldman’s Forecast Means for AI Stocks
Goldman's view is broadly supportive of the companies tied to the AI infrastructure ecosystem.
If data center construction remains strong through 2027, demand should continue for:
AI accelerators and servers
Networking equipment
Cooling systems
Data center real estate
Electricity generation
Transmission and grid equipment
Natural gas and other power sources
But investors should not assume every AI-related company will benefit equally.
The key issue is whether the growth in computing demand translates into profitable investment. Data centers require enormous amounts of capital, and higher electricity costs, construction delays, financing expenses or weaker-than-expected AI monetization could reduce returns.
Goldman itself has previously warned that delays and cancellations could materially affect the pace of capacity additions. Its earlier research estimated that only around 50% to 60% of capacity scheduled for the following one to two years might come online on time.
The Midterm Elections Add Another Layer of Risk
The data center debate is also becoming an electoral issue.
The backlash is emerging ahead of the November 3 U.S. midterm elections, creating pressure on politicians to address rising electricity costs and the local impact of large infrastructure projects. Reuters reported that the issue poses a political challenge for Republicans, who have generally supported rapid AI infrastructure development.
That does not necessarily mean Washington will slow the AI buildout dramatically.
The Trump administration has continued to support rapid AI development, viewing it partly through the lens of economic and national-security competition with China.
The more likely outcome may be a growing focus on cost-sharing, permitting, environmental standards and local economic benefits rather than an outright national restriction on data centers.
What Investors Should Watch Next
The most important indicators for the AI infrastructure trade are increasingly moving beyond chip demand.
Investors should monitor:
U.S. data center capacity additions
Actual versus planned project completions
Electricity demand growth
Grid connection queues
Data center moratoriums and permitting rules
Utility capital spending
Power prices
Hyperscaler capital expenditure
AI demand and monetization
Financing conditions for large infrastructure projects
The biggest signal will be whether strong AI demand continues to overcome the physical constraints surrounding power and construction.
Bottom Line
Goldman Sachs still sees U.S. data center growth remaining largely intact through 2027 despite rising political and community opposition. The bank now expects 64 GW of U.S. data center capacity by the end of 2026 and 90 GW by the end of 2027, while forecasting 38% annual growth in data center power demand in both years.
The message for investors is not that opposition has disappeared. It is that the AI infrastructure cycle appears strong enough, for now, to absorb delays, cancellations and regulatory resistance.
The bigger question is how the industry solves its electricity problem. If developers can secure reliable power and governments establish rules that prevent households from bearing disproportionate costs, the data center boom could continue supporting a broad investment cycle across technology, utilities and energy infrastructure.
For Indian and global investors, the U.S. data center buildout remains an important theme to watch — but increasingly, power availability and policy may matter just as much as AI chip demand.
Follow our blog for more AI, technology, global market and business updates.
This article is for informational and educational purposes only and should not be considered investment advice.