San Jose Data Centers Face a Backlash in Silicon Valley

San Jose data centers have triggered organized resistance despite the city’s effort to accelerate the infrastructure behind artificial intelligence. Residents and environmental groups now want stricter reviews before more facilities receive land, electricity, water, or operating permits.

The conflict reached a new stage on September 17, when Reuters documented opposition to proposed projects across California’s third-largest city. San Jose had spent the previous year working with utility PG&E to shorten power-delivery timelines for large energy users.

That strategy promised tax revenue and a stronger position in the AI infrastructure race. It also brought a national political conflict into the home region of Google, Meta, OpenAI, and Anthropic.

The primary contest is no longer technology companies against distant regulators. It is San Jose’s economic development strategy against residents demanding proof that AI infrastructure will not transfer costs and environmental risks onto surrounding communities.

San Jose Invited More Data Centers, Then the Opposition Organized

The backlash follows a deliberate city strategy to attract electricity-intensive development, not an accidental cluster of projects.

San Jose leaders have treated large-load customers as part of the city’s next economic development cycle. A large load is a facility whose electricity demand requires major grid planning, new equipment, or dedicated transmission capacity.

The city and PG&E approved a cooperation framework in 2025 to improve planning for these customers. The agreement sought faster timelines, better coordination, and more certainty around when power would become available.

San Jose’s own growth agenda established concrete targets. Those targets included permanent power for three data centers or other large-energy projects by June 2026. The city also sought five projects in the planning phase and two in building-plan review.

Officials identified another opportunity near the San Jose-Santa Clara Regional Wastewater Facility. The city owns 159 acres there that it has considered for economic development, including data centers and research operations.

This was an aggressive infrastructure bet. A city planning document said two expected transmission lines would provide a combined 2,000 megawatts of capacity. It described that amount as more than twice the electricity San Jose used at the time.

The lines, scheduled to enter service in 2028, were not designed only for AI. Advanced manufacturing and semiconductor facilities could also use the capacity. Data centers, however, became the most visible and politically contested customers.

City officials have argued that each new facility can produce between $3 million and $6 million in annual tax revenue. That money could support police, firefighters, libraries, roads, and parks.

Residents challenging the policy are asking a different question. They want to know whether those revenues compensate for the electricity infrastructure, water demand, pollution, and land-use consequences attached to each project.

Ellina Yin, a leader of the I Love San Jose coalition, told Reuters that companies must explain why they deserve community resources. Her coalition wants stronger public review, environmental and health studies, and clearer reporting on water and electricity consumption.

The opposition has also reached school officials. Frances Herbert, president of the Oak Grove School District board, raised concerns about a proposed facility near a community and a park used by children. She called for a moratorium and a broader discussion about project concentration.

Those demands do not amount to a single citywide rejection of digital infrastructure. They represent a push to change the approval sequence. Residents want impacts documented before officials commit scarce power capacity or approve additional sites.

That distinction matters. San Jose is not debating whether the internet needs physical infrastructure. It is debating who must carry the burden of proving that an unusually resource-intensive development benefits the city.

Why San Jose Data Centers Have Become a Statewide Test

Silicon Valley’s data center backlash matters because the industry is losing the assumption that its home region will automatically welcome more computing infrastructure.

California had more than 200 active data centers in early 2026, according to the California Energy Commission. Their combined peak demand was approximately 1,000 megawatts, equal to about 2 percent of peak demand on the California Independent System Operator grid.

The commission’s state energy forecast projects data center demand reaching 4,500 megawatts by 2040. That would represent about 9 percent of peak electricity demand.

Those figures remain forecasts, not guaranteed consumption. Utilities receive more connection requests than eventually become operating projects. Developers can withdraw, delay construction, reduce capacity, or seek sites elsewhere.

Even so, the direction is clear. AI training and inference are adding proposed loads that can reshape local grid plans. Training creates models from large datasets, while inference uses trained models to answer prompts or perform tasks.

The power question extends beyond how much electricity a completed facility consumes. New projects can require substations, transmission upgrades, and generation resources years before operators reach full capacity.

Utilities must decide which proposed loads are credible and when infrastructure should be built. Regulators must decide who pays if a customer reserves capacity but never completes the project.

PG&E says large customers can broaden the base that supports fixed grid costs. Under that argument, additional revenue from data centers can reduce the amount other customers must contribute.

Critics focus on the opposite scenario. They worry that utilities will build expensive infrastructure around optimistic forecasts, leaving households and small businesses exposed if projects are delayed or canceled.

Both outcomes are plausible because the answer depends on contract design. Minimum payments, deposits, credit requirements, construction milestones, and exit fees determine whether project risk stays with the developer.

California lawmakers have responded with proposals designed to prevent cost shifting. Senate Bill 886 addresses tariffs and financial protections for certain large-load facilities connecting to the grid.

A legislative analysis of the proposed ratepayer protections identified a 90-megawatt San Jose data center as relevant to the regulatory debate. The analysis also noted increasing applications for transmission-level connections in investor-owned utility territories.

Supporters say large customers should cover the infrastructure created for them. Industry and business groups argue that inflexible statewide requirements can duplicate ongoing utility regulation and discourage investment.

San Jose sits at the center of that dispute. Its cooperation with PG&E was designed to make projects move faster. State lawmakers are considering requirements intended to make their financial and environmental consequences more visible.

The city therefore faces pressure from three directions. Developers want speed and reliable power dates. Residents want enforceable protections before approvals. State officials want rules that work across utilities and local jurisdictions.

That makes the San Jose AI data center fight larger than one project. It is an early test of whether California can combine faster construction with credible limits on public exposure.

The AI Infrastructure Promise Is Colliding With Its Physical Costs

AI services feel weightless to users, but the facilities supporting them compete for land, grid capacity, water, and clean air.

The industry’s strongest case begins with demand. Companies are building larger computing systems because developers, businesses, and consumers are using more generative AI services.

California also has strategic reasons to keep that infrastructure nearby. The state hosts leading AI developers, chip designers, cloud companies, and research institutions. Local capacity can support product development, network performance, and associated technical work.

San Jose officials see an additional fiscal opportunity. Data centers can generate substantial tax revenue while occupying sites that might not support housing, retail, or conventional offices.

The weakness in that argument is distribution. A city can collect new revenue while one neighborhood experiences the noise, construction, generator emissions, or land-use effects.

Data centers also create fewer permanent jobs than many factories occupying comparable industrial land. Construction can generate significant temporary employment, but the long-term operating workforce is typically smaller.

That does not erase their fiscal value. It does make the public bargain more dependent on tax receipts, infrastructure payments, and enforceable community protections.

Energy is the largest visible constraint. A modern facility can require enough power to trigger upgrades extending far beyond its property line. The exact demand varies according to equipment, utilization, cooling design, and project scale.

Water use is more complicated. Some facilities rely on evaporative cooling, which can consume significant water during operation. Others use closed-loop systems, air cooling, reclaimed water, or designs that reduce local consumption.

Residents cannot evaluate those differences when project documents offer only general assurances. They need site-specific estimates covering average use, peak use, water sources, drought conditions, and emergency operations.

Air quality creates another point of conflict. Data centers need backup power because an interruption can damage equipment or disrupt important digital services. Diesel generators remain a common way to provide that backup.

Operators describe these generators as emergency equipment rather than a primary energy source. Environmental groups argue that testing, maintenance, outages, and exceptional grid conditions can still create concentrated emissions.

In April 2026, a coalition led by the Center for Biological Diversity petitioned the Bay Area Air Quality Management District for stronger oversight. The groups argued that existing generator rules predated the recent expansion of AI-related infrastructure.

The district said it was evaluating whether additional protections were necessary. Meanwhile, the Data Center Coalition maintained that backup generators are essential for reliability and critical services.

The dispute cannot be resolved through broad claims about diesel technology. The useful questions concern each facility’s generator capacity, testing schedule, recorded operating hours, emissions controls, and proximity to homes or schools.

That is why transparency has become central to the Silicon Valley data center backlash. Communities are not only asking for lower impacts. They are asking for enough information to establish what those impacts actually are.

The conflict also exposes a timing problem. Cities often approve land use before utilities disclose every detail of power delivery. Developers can finalize cooling or equipment choices later in the process.

By the time complete data becomes available, public agencies may already have made decisions that are politically or financially difficult to reverse. Opponents want environmental review and community engagement moved earlier.

San Jose’s original strategy emphasized speed to power. The emerging political response emphasizes proof before power. Reconciling those goals is now the central challenge.

Faster Connections and Public Consent Are Pulling in Opposite Directions

San Jose’s economic promise depends on rapid approvals, while public trust depends on reviews that can slow or reshape projects.

Mayor Matt Mahan has rejected the idea that the city must stop all data center development. He has instead argued for responsible construction and acknowledged the legitimacy of residents’ concerns.

Mahan told Reuters that people were right to ask whether projects would increase energy costs, consume clean water supplies, or reduce quality of life. That statement places the mayor between two increasingly organized constituencies.

Technology and infrastructure companies want predictable approvals. Their projects involve equipment orders, utility studies, land commitments, and construction schedules that can span several years.

Community groups want a process capable of producing a different outcome when risks are too high. Consultation without the possibility of denial or redesign would offer participation without meaningful leverage.

The Data Center Coalition says it supports responsible growth. Its members include Google, Meta, OpenAI, and Anthropic, alongside major infrastructure operators.

The coalition has criticized California legislation for singling out data centers instead of applying equivalent rules to other large commercial and industrial customers. It also warns that added requirements can redirect investment, employment, and tax revenue toward other states.

That objection deserves serious consideration. Semiconductor plants, advanced manufacturing sites, hospitals, and other facilities can also require substantial electricity. A rule based only on the label attached to a building can produce uneven treatment.

However, data centers have several characteristics that make the current wave unusual. Projects can be exceptionally large, operate around the clock, rely on extensive backup generation, and expand in phases after initial approval.

Their demand is also difficult to forecast. The California Energy Commission says uncertainty surrounds application volumes, completion rates, development schedules, and the load each completed project will ultimately require.

San Jose’s approach must handle this uncertainty without treating every application as inevitable. A proposed project should not receive the same planning weight as one with financing, equipment commitments, permits, and binding utility obligations.

The city also needs a clear standard for cumulative impact. Reviewing each facility independently can miss what happens when several projects occupy the same district or draw from the same infrastructure.

This is especially important near schools, parks, residential areas, and communities already carrying significant pollution burdens. One compliant generator installation can look different when combined with several neighboring facilities.

California’s broader environmental debate reflects that concern. Senate Bill 887 seeks to clarify review requirements and establish conditions for data centers seeking streamlined treatment under state environmental law.

The proposal’s environmental rules include expectations concerning clean electricity, interconnection costs, and grid investments. The legislative analysis also acknowledges uncertainty about whether operators would accept its conditions.

That uncertainty points to the real policy tradeoff. Requirements that no developer can meet will move projects elsewhere without improving local operations. Weak requirements will preserve speed while leaving residents with unresolved risks.

A workable framework needs measurable standards rather than promises. Those standards can include full interconnection payments, enforceable water limits, public generator records, community engagement milestones, and penalties for unused reserved capacity.

The framework must also distinguish verified demand from speculative requests. Otherwise, utilities may plan around a queue that exaggerates future consumption.

Public consent does not require universal agreement. It does require a process that shows who benefits, who bears each cost, and what happens when a company fails to meet its commitments.

What the Backlash Still Does Not Prove

The resistance establishes a trust and governance problem, but it does not establish that every proposed facility will raise bills or exhaust local resources.

Opponents have identified legitimate questions, yet many project-specific answers remain unavailable. The effects of San Jose data centers will vary according to location, design, contracts, and operating practices.

A facility paying its full interconnection expenses presents a different financial risk from one whose costs enter the general rate base. A project using reclaimed water differs from one drawing potable supplies during hot periods.

Backup generators equipped with modern emissions controls differ from older systems with limited monitoring. A site in an established industrial district differs from one next to homes, schools, or public parks.

The city’s tax estimate also needs careful interpretation. The projected $3 million to $6 million in annual revenue is an estimate per facility, not a guaranteed net public benefit.

Gross tax receipts do not automatically account for added fire protection, road work, permitting, grid coordination, water infrastructure, or environmental monitoring. A credible public analysis should identify those costs separately.

Claims about electricity bills require equal caution. PG&E’s argument that new large customers can spread fixed costs is economically coherent. It becomes convincing only when contracts prevent other customers from absorbing abandoned or underused infrastructure.

Opponents sometimes describe AI data centers as though all electricity serves model training. In practice, facilities can support cloud computing, enterprise software, storage, streaming, communications, and other workloads.

That distinction does not remove the pressure created by AI. It does make simple labels unreliable. Regulators need operating data and defensible forecasts, not assumptions based on marketing descriptions.

There is also a risk that local restrictions displace impacts instead of reducing them. A project rejected in San Jose can move to a community with cheaper land, weaker disclosure requirements, or greater water stress.

California therefore needs coordination across jurisdictions. Otherwise, cities offering the fastest approvals can capture revenue while neighboring regions inherit transmission, generation, or environmental consequences.

A statewide framework can reduce that competition, but only if it preserves room for local conditions. California’s communities differ in grid capacity, water availability, air quality, land use, and economic needs.

The current debate should not be reduced to supporters of AI against opponents of technology. Many residents rely on digital services and still want stronger conditions on the buildings supporting them.

Likewise, city officials can support AI development without accepting every proposal. Responsible growth becomes meaningful only when officials can require redesign, delay a project, or reject it.

The most important verification gap concerns outcomes. San Jose has announced ambitious targets and planning arrangements, but approval counts alone reveal little about community benefit.

The city will need to publish actual electricity demand, infrastructure payments, water consumption, generator operation, tax receipts, and permanent employment after facilities open.

Until those figures arrive, both sweeping promises and sweeping warnings should be treated cautiously. The backlash proves that the previous level of disclosure is no longer politically sufficient.

Three Signals Will Decide the Future of Silicon Valley Data Centers

The next phase will be determined by state legislation, San Jose’s operating standards, and the contracts governing large electricity connections.

The first signal is Governor Gavin Newsom’s decision on the data center bills awaiting action. Reuters reported that he had until the end of September 2026 to sign or veto the pending measures.

A signature would strengthen the argument that California is moving from voluntary commitments toward statewide obligations. A veto would leave more responsibility with utility regulators and individual cities.

The details will matter more than the headline. Rules addressing deposits, minimum payments, exit fees, and interconnection costs can directly limit ratepayer exposure.

Reporting requirements can also establish a common evidence base. Public information about energy, water, and backup generation would let communities compare facilities instead of arguing from incompatible estimates.

The second signal is San Jose’s local standards. The City Council has directed staff to develop consistent guidance for data centers, including environmental mitigation and community engagement.

Those standards will show whether the city’s response changes project requirements or merely improves communication. Specific thresholds, review schedules, and enforcement mechanisms would strengthen the responsible-growth position.

A moratorium would mark a sharper change. It would show that officials believe existing rules cannot manage the volume or concentration of proposed development.

The absence of a moratorium would not mean the opposition failed. Stronger conditions imposed through permitting can materially alter a project without stopping it.

The third signal is the structure of large-load utility agreements. These contracts determine whether developers or other customers carry the financial risk of new grid construction.

Watch for binding demand commitments, upfront contributions, credit support, cancellation penalties, and milestone-based capacity reservations. Those provisions will test PG&E’s claim that large customers can lower rather than increase pressure on bills.

They will also reveal how much confidence developers have in their own forecasts. A company willing to guarantee payments over several years presents a stronger demand signal than one seeking capacity with limited obligations.

The same principle applies to water and emissions. Operators willing to accept enforceable limits and public reporting will have a stronger case than those offering only general sustainability pledges.

For developers, enterprise buyers, and AI users, this dispute affects more than local permitting. Infrastructure constraints can influence where computing capacity is built, when it becomes available, and how much reliability providers can promise.

For residents, the issue is whether the benefits of the AI economy arrive with transparent and enforceable protections. Tax revenue alone will not settle that question.

San Jose now has an opportunity to define a more credible bargain. It can require companies to pay the costs they create, document resource use, and earn faster treatment through measurable performance.

The alternative is a cycle of rushed approvals followed by escalating resistance. That path would make every new proposal slower, less predictable, and more politically expensive.

The future of San Jose data centers will therefore depend on evidence rather than branding. Will California make infrastructure developers carry identifiable risks, and will San Jose publish enough data for residents to judge the results?