Two Data Center Watch estimates show local opposition touching data center proposals valued at $156 billion in 2025 and about $130 billion in the first quarter of 2026. The public figures combine blocked and delayed projects and do not measure capital lost, while Morgan Stanley expects political resistance to raise costs and change where and how capacity gets built.
Local opposition is changing the terms of the U.S. data center expansion, but the largest number attached to that shift is easy to overstate. The evidence supports a costlier, slower and more selective buildout—not a conclusion that $286 billion has vanished from the AI investment cycle.
Data Center Watch, which tracks political and community resistance to data center development, published two period estimates:
| Tracking period | Projects described as affected | Value assigned by tracker |
|---|---|---|
| 2025 | Dozens | At least $156 billion |
| Q1 2026 | At least 75 | Approximately $130 billion |
The 2025 public summary says opposition blocked or delayed projects with publicly disclosed values totaling at least $156 billion. The same page, however, uses $152 billion in an opening bullet. It does not reconcile the difference.
The tracker’s Q1 2026 summary describes at least 75 projects worth about $130 billion as blocked or delayed during the quarter. It calls that the largest quarterly concentration on record and roughly the scale of all of 2025.
Those estimates add arithmetically to about $286 billion, but that sum is not a measure of economic loss. The public pages do not separate postponements from cancellations, show how much money had already been spent or disclose the project-level data needed to test how continuing disputes are handled across reporting periods. They advertise fuller reports separately. The defensible comparison is therefore between two large estimates—not between $286 billion of supposedly lost capital and another investment total.
That distinction also rules out a tempting comparison with Morgan Stanley’s reported $877 billion estimate for AI-related capital spending in 2026. The tracker figures attach announced values to proposals affected over five quarters; the bank figure is a one-year spending forecast. They do not share a measurement basis.
The tracker’s evidence does show a rapid widening of political risk. It said hundreds of opposition groups were active across 42 states by the end of 2025. By the end of March 2026, active groups had more than doubled and activity extended across 49 states. Its Q1 summary also counted more than 300 state data center bills filed in the first six weeks of the year and statewide moratorium proposals in 14 states.
Those categories cover more than outright bans. They include pauses, zoning limits, environmental review, tax treatment and rules about who pays for new electricity infrastructure. A July 14 report quoting Morgan Stanley’s analysis said New York had announced a one-year statewide pause for large projects above 50 megawatts while it developed energy, water and environmental standards.
Morgan Stanley’s own public discussion frames the likely outcome as conditional expansion. Ariana Salvatore, the bank’s head of public policy research, said a federal ban was unlikely under the current administration and would also be difficult to envision under a future Democratic administration because both parties see AI capacity as strategically important relative to China. She expected contracts to require grid modernization, longer commitments and benefits for host communities.
In that broader discussion of AI investment, the bank’s Stephen Byrd said he expected demand for compute to exceed supply despite competition from cheaper open-source models. The bank is treating local resistance as an execution and supply constraint, not evidence that underlying compute demand has collapsed.
Federal policy points in the same direction, though it does not guarantee the result. The administration’s Ratepayer Protection Pledge calls on hyperscalers and AI companies to build, bring or buy the new generation they need, pay for delivery upgrades and negotiate separate utility rates that remain payable even when the electricity goes unused. Those rate arrangements are voluntary. The pledge expresses the federal government’s preferred allocation of costs; it is not evidence that every utility or state regulator will achieve it.
The scale of expected electricity demand explains why communities, utilities and owners of connected sites have leverage. A 2024 federal assessment said U.S. data center electricity use rose from 58 terawatt-hours in 2014 to 176 TWh in 2023, or about 4.4% of national consumption. It projected 325 to 580 TWh by 2028, equivalent to roughly 6.7% to 12% of U.S. electricity use.
That forecast establishes the size of the power challenge, not the cause of any particular project delay. The same July 14 account said the bank estimated a possible 38-gigawatt U.S. data center power shortfall through 2028 before additional “time-to-power” solutions, with grid interconnection waits reaching five to seven years in some regions.
The bank sees two broad deployment paths. In politically supportive states with vertically integrated utilities, a project can remain on the grid while the developer pays for generation, network upgrades and community benefits. In less accommodating markets, developers may pursue natural-gas turbines, fuel cells and storage onsite to reduce reliance on the grid. Moving power behind the meter does not eliminate local risk: fuel-burning equipment may need air permits, and water-dependent designs still require approval.
The constraint also creates uneven competitive effects. The July 14 account said Morgan Stanley identified onsite-power suppliers Solaris Energy Infrastructure, Innio and Bloom Energy as possible beneficiaries. It described Equinix and Digital Realty as relatively insulated because their colocation facilities generally draw 5 to 10 megawatts, far less than hyperscale campuses. Bitcoin-mining operators with firm interconnection agreements may have another advantage: an energized site that a new developer cannot reproduce quickly.
Construction evidence narrows the political thesis further. A construction-sector report attributed to Baird analyst Justin Hauke a rise in data center cancellations from two in 2023 to six in 2024 and 25 in 2025. But interviews in the same report identified power, permitting and labor as overlapping points of failure.
The projects themselves are also getting larger. Green Street’s David Guarino said a 100-megawatt lease would have been considered massive a few years earlier, while projects above 1,000 megawatts were setting the benchmark. Hut 8 construction executive Brennan Church said medium-voltage switchgear was taking 40 to more than 60 weeks, with transformers and generators also on extended lead times. A site with land and fiber can still stall when the utility cannot commit to energization.
That evidence prevents the opposition tallies from serving as a complete explanation of industry delays. It also complicates any claim that resistance has broken demand: Guarino said larger lease orders from both neocloud companies and hyperscalers were continuing.
The financial consequence may be more about timing and risk allocation than near-term output. The July 14 account said Morgan Stanley estimated that a similar pace of delays would trim less than 5 basis points from real U.S. GDP growth in both 2026 and 2027 after accounting for imported equipment. The bank’s larger stated concern was that slower compute deployment could postpone AI adoption and potential productivity gains beyond 2028.
The report quoting the bank said sustained resistance could extend the capital-spending cycle or reduce eventual financing needs, while near-term capital-spending front-loading could intensify debt issuance pressure. That is a forecast, not an observed outcome. It nevertheless identifies who carries the risk: developers must fund more infrastructure and carry schedule risk, while utilities, regulators and communities gain bargaining power over cost and siting.
The central question cannot be resolved by adding more proposal values. It requires project-level outcomes: how many delayed sites obtain permits and financing, how many are canceled, and how many relocate. Disclosure of the tracker’s underlying records would also show whether affected projects recur across reporting periods and how announced values are assigned.
Utility proceedings are the second test. Separate tariffs, minimum-payment contracts and developer-funded upgrades would support Morgan Stanley’s conditional-buildout case if they keep costs away from households and still produce permits. If developers instead move generation onsite, air, gas and water approvals will show whether the strategy removes a bottleneck or merely changes it.
Until those results are visible, the evidence supports a narrower conclusion: opposition has become a material development risk, but the two tracker estimates measure proposals under pressure—not capital already lost or proof that the AI infrastructure cycle has reversed.
Zhipu reportedly reached $1 billion in annual recurring revenue in July, roughly four times a March estimate, but the unconfirmed run rate is not annual sales and still sits far ahead of recognized cloud revenue while margins remain thin.
Anthropic’s 20-year lease gives TeraWulf a customer for 401 megawatts of future AI infrastructure, but rent starts only after delivery and the proposed utility deal passes power and grid costs to TeraWulf, leaving financing and project margin unresolved.
OpenAI has put conversations and Projects back in its redesigned ChatGPT desktop app and enabled cloud Work threads to move across devices, correcting the launch's biggest usability failures without merging local Work or Codex histories.
Twenty-nine countries signed an agreement creating WAICO as an independent intergovernmental organization, while China paired the launch with capacity-building offers that are not yet confirmed as WAICO programs.
Meta reportedly plans to put departing AWS compute executive Dave Brown to work on its data-center buildout, adding hyperscale operating experience while leaving any customer-facing cloud business conditional and undefined.
Moonshot AI has made Kimi K3 available through its apps and API, pairing a 2.8-trillion-parameter architecture with early frontier-level results, but the model's open-weight claim cannot be tested until its weights and technical report arrive.
CIA Director John Ratcliffe said US intelligence is consistent with an estimate that Russian recruits last 20 to 30 minutes on Ukraine’s battlefield, but the public trail leads to an unsourced claim about assault troops and does not establish a representative average.
China has paired a five-year AI training offer for developing countries with cooperation centers, a weather-warning rollout and a new 29-country organization. The package gives Beijing a platform for influence, but no budget, selection rules or delivery timetable has been published.
Nebius has arranged its first senior secured facility against an operating GPU deployment and one customer’s cash flows. The deal adds project-level debt to its expansion toolkit, but an unnamed customer and a larger rival financing limit what it proves about the rest of Nebius’s backlog.
Huawei publicly displayed a 16-cabinet Atlas 950 configuration rated at 1 EFLOPS in FP8, providing tangible evidence of its system-scale AI strategy while leaving price, power use and sustained workload performance undisclosed ahead of the full system's planned fourth-quarter release.
UK testing places leading open-weight models four to seven months behind selected closed-model cyber results, yet longer attack chains, U.S. benchmarks and mixed cost comparisons show why that interval is a warning signal rather than a universal capability clock.
Moonshot AI's Kimi K3 added to a global technology selloff with near-frontier performance, but unreleased weights, mixed cost comparisons and a recommended 64-accelerator deployment leave its effect on chip demand unresolved.
Two binding EU decisions require Google to give rival AI services comparable access to 11 Android features and offer eligible search competitors a restricted, anonymized dataset, but phased deadlines, certification, pricing and privacy safeguards leave the competitive effect unproven.
Databricks has signed a term sheet for a Coatue-led financing at a $188 billion valuation, while unidentified sources put the round at $3 billion. The proposed capital would deepen its push into AI governance, data agents and operational databases, but the transaction remains open and the company supplied no new operating figures.
OpenAI says a handful of GPT-5.6 Sol file-deletion reports most commonly involved Full Access without sandboxing or Auto-review. Its own evaluations show a more complicated risk picture, and the company has not yet published an incident rate or evidence that its promised safeguards stop the failure.
Netflix says generative AI workflows were used on roughly 300 titles in 2026, but its only quantified example tied to that disclosure covers 17 minutes and does not establish how much finished material, spending or labor changed across the slate.
A rogue browser extension can trigger Claude for Chrome’s fixed Gmail, Docs, Calendar and business workflows; the attack is constrained by default approvals but can run silently for users who enabled unattended action.
Gemini 3.5 Pro missed its expected June rollout. An anonymous-source account says a late-June data change fell short of Google's coding goals, but Google has confirmed only partner testing—not the reported cause, a new date, or public results and pricing.
Gold Eagle has begun collecting and prioritizing AI-discovered software vulnerabilities, but the voluntary federal clearinghouse has not disclosed results, operating rules or the resources that would turn findings into deployed fixes.
Google is bringing Gemini Omni editing and a reusable face-and-voice avatar to Vids, but the sharper distinction is account-level identity across Vids and Gemini rather than a new category of AI video; Vids already offered Veo generation and customizable avatars, while specialist rivals already sell digital presenters.