Francesca Hong’s Wisconsin Primary Loss Gives AI Data Centers a Reprieve

Francesca Hong lost Wisconsin’s Democratic gubernatorial primary by less than one percentage point, removing the race’s strongest advocate for pausing new AI data centers. That result makes the google news story bigger than an election postmortem. It changes which data-center policies are likely to reach the governor’s office.

Milwaukee County Executive David Crowley defeated Hong after reentering the race with support from outgoing Governor Tony Evers. Hong had been the only major Democratic candidate supporting a statewide moratorium on new AI data-center construction.

Her defeat does not authorize a single server building, transmission line, or power plant. Local governments, state regulators, utilities, and courts retain important authority. Yet it removes a potential governor who promised to place a broad pause at the center of state policy.

That is the reversal. Wisconsin voters have become increasingly skeptical of data centers, while the state has committed substantial tax support to their expansion. The candidate offering the clearest statewide brake still lost.

Microsoft, Meta, Oracle, Vantage Data Centers, and other developers therefore gained political breathing room. Their victory remains provisional because Wisconsin’s data-center fight has already moved beyond one candidate and one election.

Hong’s loss removed the clearest statewide brake

The primary changed who can exercise executive power over Wisconsin’s expanding data-center market.

The primary result placed Crowley against Republican Representative Tom Tiffany in the general election. Hong conceded after an exceptionally close contest, ending the possibility of a recount.

The Associated Press reported that Crowley’s margin remained below one percentage point. Hong could have requested a recount, although her campaign would have needed to cover its cost.

About 28,000 Milwaukee absentee ballots were also reported late because of a human error involving data uploads. That delay added uncertainty during the count, but Hong’s concession ultimately settled the Democratic nomination.

The data-center consequences come from the difference between the candidates, not merely the narrow margin. At a June forum, six leading Democratic candidates were asked whether they supported a one-year construction moratorium.

Hong raised the only green sign. Every other candidate opposed the proposal.

Her plan would not have halted projects already under construction in Mount Pleasant, Port Washington, or Beaver Dam. Instead, it would have paused additional approvals while Wisconsin created statewide rules.

Those proposed rules addressed four recurring concerns. They included utility costs, public subsidies, environmental safeguards, and union labor requirements.

Hong argued that Wisconsin lacked enough enforceable protections for the speed and scale of proposed development. She also wanted affected communities to hold meaningful approval power.

Her earlier legislative proposal had sought similar conditions. Large data centers could not operate until standards covering energy costs, environmental effects, labor, and local consent were satisfied.

That legislation failed. Competing Democratic and Republican regulatory bills also failed to reach Evers’ desk during the previous legislative session.

A Hong governorship would have changed that political equation. She could have made a moratorium an executive priority, shaped agency appointments, and pressed lawmakers to revisit restrictions.

Her loss closes that particular route. It does not erase the underlying proposals, but it deprives them of their most prominent statewide sponsor.

Google News aggregation can make this look like a simple story about progressives losing momentum. That frame misses the technology-policy consequence.

Hong tied a traditional election to decisions about power generation, grid investment, tax exemptions, water, land, and local authority. Those decisions determine whether AI infrastructure can scale inside Wisconsin.

Her defeat therefore matters to companies far beyond the campaign. It reduces the immediate probability of a governor-led construction pause beginning in 2027.

The word “immediate” matters. Crowley has not received a blank check from voters, and the general election remains unsettled. He must also answer a constituency that came remarkably close to choosing Hong.

Her near victory demonstrated that data-center opposition can support a competitive statewide campaign. Developers avoided the strongest proposed intervention, but they did not eliminate its political base.

Wisconsin already made a large bet on data centers

The election occurred after Wisconsin committed public policy, utility planning, and industrial land to several immense computing projects.

Wisconsin created its qualified data-center sales and use tax exemption through its 2023-25 budget. Certified projects can avoid taxes on many purchases connected to construction and operations.

State eligibility depends on investment thresholds and county population. According to the Wisconsin Economic Development Corporation, qualifying investment starts at different levels across large, medium, and smaller counties.

The exemption matters because data centers require far more than buildings. Their taxable inputs can include servers, networking equipment, electrical systems, cooling equipment, backup generation, and large quantities of electricity.

A Legislative Fiscal Bureau estimate placed the potential public cost above $2 billion. Across projects involving Microsoft, Oracle, Meta, and Epic Systems, Wisconsin could forgo $1.5 billion during construction.

The same estimate projected another $369 million in forgone sales tax revenue each year after the projects become operational. These figures are estimates, not final invoices, because construction schedules and equipment purchases can change.

Still, the scale reframes the election. Hong was not campaigning against a hypothetical sector. She was challenging a state development strategy that had already attracted named projects.

The Department of Revenue’s certified project list included Microsoft in Mount Pleasant, Epic Hosting in Verona, and Degas LLC in Beaver Dam. It also listed Oracle America Cloud Services in Port Washington.

That list was current through October 31, 2025. New proposals and changes can emerge after such a certification date, so it should not be treated as a complete map.

The projects also differ. Some support cloud computing, some target AI workloads, and some serve enterprise applications. Treating every facility as identical would obscure their separate customers and infrastructure demands.

Their shared characteristic is scale. Large computing campuses need dependable power, extensive grid connections, cooling capacity, fiber access, and long construction timelines.

Developers want predictable approvals because uncertainty can strand land purchases and utility work. Communities want information before those commitments make alternative choices difficult.

Wisconsin’s incentive structure favors early certainty for developers. Hong’s moratorium would have shifted more uncertainty back onto the companies while statewide safeguards were negotiated.

Her loss keeps the existing sequence largely intact. Projects can pursue approvals under current law while legislators debate whether broader rules should follow.

Business advocates say the incentives helped Wisconsin compete with other states. Thirty-seven other states also offered data-center incentives, according to reporting that cited the National Conference of State Legislatures.

Supporters also argue that construction spending, supplier orders, income taxes, and new generation can return value to Wisconsin. Local property taxes could eventually support schools and public services.

That return is neither automatic nor immediate. Tax incremental financing districts can reserve added property revenue for project costs before schools and other taxing bodies receive the full benefit.

This distinction explains the primary opponent in the debate. It is not simply Hong against technology companies. It is rapid project approval against public rules established before communities assume long-term obligations.

The google news headline captures the short-term winner. Existing developers retain momentum because the candidate most committed to reversing that sequence will not become the Democratic nominee.

Yet those companies still need to show that promised economic gains exceed public costs. Tax exemptions make that burden more important, not less.

What Google News misses about the public backlash

Hong lost even though skepticism toward data centers has become a broad Wisconsin concern rather than a narrow ideological cause.

A statewide policy report cited Marquette University Law School polling that found roughly 70 percent of Wisconsin voters saw more costs than benefits. That sentiment extends beyond Democratic socialist voters.

Communities have questioned electricity demand, water use, construction noise, diesel backup generators, land conversion, tax agreements, and confidential negotiations. Each project presents a different combination of those issues.

Hong gave those concerns a simple statewide vehicle. Her proposed pause promised time to develop standards before additional projects advanced.

Business organizations opposed that approach. Dale Kooyenga, president of the Metropolitan Milwaukee Association of Commerce, argued that a moratorium would restrain economic growth.

He also maintained that Wisconsin might have lost major investments without its tax incentive. The Wisconsin Data Center Coalition emphasized manufacturing suppliers that can benefit from construction and operation.

Those arguments identify real economic activity. Data-center campuses purchase generators, cooling systems, electrical components, construction services, and maintenance work.

However, capital investment does not settle how benefits are distributed. Construction employment can be temporary, while transmission assets and generation contracts can operate for decades.

A statewide number can also hide local tradeoffs. One municipality might value property development, while neighboring residents experience transmission construction without receiving the same revenue.

Hong’s campaign placed that distributional question beside the headline investment totals. Who receives the upside, and who remains responsible when forecasts change?

Her defeat does not show that voters rejected data-center regulation. The primary included many other issues, including affordability, healthcare, public safety, and candidate electability.

Crowley also benefited from a chaotic field. Major candidates withdrew, Crowley left and reentered, and Evers endorsed him after his return.

National progressive leaders, including Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez, did not endorse Hong. Democratic voters also weighed which nominee might defeat Tiffany in November.

It would therefore be inaccurate to read the result as a referendum approving every Wisconsin data-center project. Voters did not receive a ballot question asking whether Microsoft or Oracle should obtain specific incentives.

The stronger conclusion is narrower. Data-center opposition helped Hong build a competitive coalition, but it did not carry her across the statewide primary threshold.

That outcome creates a warning for both sides. Activists cannot assume widespread concern automatically produces support for a moratorium. Developers cannot assume Hong’s loss means that concern has disappeared.

The moratorium debate already demonstrated the pressure. Hong was the only Democratic candidate supporting a statewide pause, but no major candidate embraced unlimited development.

That middle position now becomes decisive. Policymakers can oppose a moratorium while supporting rate protections, disclosure rules, environmental standards, or limits on incentives.

Companies should expect those narrower interventions to become more attractive after Hong’s loss. They offer politicians a response to voter skepticism without stopping construction statewide.

This is where the google news framing can mislead readers. Elections tend to produce binary winners, while infrastructure policy moves through overlapping institutions.

Hong lost. The opposition movement did not.

Developers won time, procedural continuity, and a less hostile Democratic nominee. They did not win permanent public consent.

Utility rules show why the victory is only temporary

Wisconsin regulators have already imposed protections that resemble parts of Hong’s argument, even without adopting her moratorium.

Data centers can require electricity on the scale of industrial facilities. Utilities must build or acquire generation and transmission capacity before that demand fully materializes.

This creates a forecasting risk. If a customer scales back, delays construction, or leaves early, existing ratepayers could face costs from infrastructure built around the original projection.

We Energies proposed special tariffs after very large data-center customers entered its service territory. A tariff is the regulated set of prices, terms, and service conditions applied to customers.

The Public Service Commission reviewed the proposal for about one year. It then approved substantial revisions intended to limit cost shifting and improve transparency.

The commission lowered the tariff’s demand threshold from 500 megawatts to 100 megawatts. That change brought smaller, but still enormous, customers within the special structure.

Regulators also extended the minimum initial term to 15 years. Longer commitments reduce the risk that ordinary customers inherit expenses after a large customer exits.

The commission removed an option under which a data center would have covered only 75 percent of certain generating-facility costs. It instead required very large customers to pay the full costs under the approved model.

Additional reporting requirements addressed agreements between utilities and large customers. The commission also ordered revisions covering possible transmission-cost shifting.

These utility protections weaken any claim that Wisconsin faces a choice between a total pause and no safeguards. Regulators can intervene through rates even when construction continues.

They also reveal the limits of that approach. The commission regulates electric, gas, and water utilities, but it does not approve data-center construction or operations.

It cannot replace local zoning, environmental permits, tax policy, or statewide development law. Rate design addresses one major risk while leaving others unresolved.

The Port Washington campus illustrates the fragmented process. Wisconsin’s Department of Natural Resources has handled wetland, wastewater-planning, and air-permit questions connected to that development.

The DNR issued an air construction permit in June 2026 covering 45 diesel-fired emergency generators. Backup generators support reliability, but their permitting raises local air-quality and operational questions.

One agency can assess those generators while another reviews utility rates. Local governments handle land use, and the state’s economic-development system administers tax eligibility.

Hong’s moratorium tried to place those decisions inside a single political story. Her loss returns the fight to separate proceedings that receive less public attention.

That fragmentation generally helps projects continue. Opponents must engage with multiple agencies, technical filings, municipal meetings, and legislative committees.

Developers can also advance one approval while another remains pending. A failed bill does not automatically stop a permit, and a disputed local decision does not necessarily change statewide tax law.

However, fragmented oversight can create delays and litigation. Oracle has challenged aspects of Wisconsin’s financial-security requirements for large electrical customers, showing that approved projects still face regulatory conflict.

The remaining uncertainty is therefore not whether oversight exists. It is whether separate safeguards can manage the cumulative impact of several projects expanding together.

Power forecasts offer one test. If demand projections hold, long customer commitments can finance dedicated infrastructure without shifting costs.

If forecasts fall, regulators will need contractual protections that survive slower AI growth or changing hardware efficiency. The election did not answer that problem.

The phrase “for now” belongs in any serious analysis. Hong’s loss avoids a statewide brake, but utility rules are already moving toward stricter accountability.

Developers gained time, not an unrestricted mandate

The election protects the current development path while leaving its economic and environmental claims open to verification.

Supporters describe data centers as anchors for construction, manufacturing, cloud services, and property development. Those benefits deserve measurement against the subsidies and infrastructure involved.

The most useful comparison is not between technology and no technology. It is between projects operating under enforceable commitments and projects relying on optimistic projections.

Jobs provide one example. Construction can employ large crews, but permanent data-center staffing is usually more limited than employment at traditional manufacturing plants.

The appropriate question is how each project’s payroll, local purchasing, and tax contributions compare with its public support. Aggregate investment alone cannot answer that.

Energy is another test. A campus can finance new generation while still changing regional planning, transmission routes, and fuel choices.

Contract language determines who pays if costs exceed forecasts. Public reporting determines whether residents can evaluate those arrangements before liabilities become fixed.

Water use also depends on cooling design, weather, facility load, and operational choices. A statewide slogan cannot substitute for project-level information.

The same applies to backup generation. Permitted generator capacity does not reveal how often units will operate, but communities need enforceable limits and monitoring.

Hong argued that Wisconsin should establish these protections before approving more projects. Industry advocates argue that an immediate pause would send investment elsewhere.

Crowley’s victory favors the second side’s timing. Development can proceed while policymakers seek narrower protections.

That does not prove the second side’s assumptions. It merely prevents Hong from using the governor’s office to reverse the order.

A meaningful industry victory would require several outcomes. Data centers would need to meet construction schedules, pay contracted energy costs, and deliver promised local revenue.

They would also need to avoid transferring grid risk to households and small businesses. Environmental compliance would need to hold once facilities operate at scale.

Finally, communities would need access to important information before approving land, financing, or utility arrangements. Confidential negotiations can weaken trust even when final agreements appear favorable.

The public-cost estimates deserve similar caution. The projected $1.5 billion construction exemption and $369 million annual figure depend on assumptions about spending and operation.

Actual forgone revenue could differ. The state should publish updated figures as projects acquire equipment and electricity.

The election creates no reason to lower that reporting standard. If anything, the close primary strengthens the case for clearer measurement.

Nearly half the decisive coalition supported a candidate offering a statewide pause. Ignoring those voters would invite the issue to return through local referendums, legislative races, or future statewide campaigns.

Google News readers should therefore distinguish political access from operational success. Companies preserved access to Wisconsin’s development process, but they still must perform under public scrutiny.

A project can survive an election and fail a permit challenge. It can secure tax treatment and still face utility disputes.

It can begin construction and later revise its capacity. It can also meet every legal requirement while producing benefits that residents judge insufficient.

That range of outcomes is why Hong’s loss is best described as a temporary win. The regulatory contest continues through evidence rather than campaign promises.

Three signals will decide what happens next

The next phase depends on the general election, project-level cost evidence, and the revival of statewide legislation.

The first signal is Crowley’s platform against Tiffany. Crowley must convert a narrow primary victory into a position that can hold Democratic voters while reaching independents.

Watch whether he proposes specific data-center standards before November. A detailed plan covering utility liability, tax disclosure, and local participation would narrow the policy distance from Hong.

Silence would strengthen the industry’s current advantage. It would suggest that statewide restrictions lack a champion among the leading candidates.

Tiffany’s response matters too. Data-center skepticism crosses party lines, especially when residents fear higher bills or limited local control.

If both candidates propose protections, Hong’s loss will have changed the form of regulation rather than ended the debate. The contest would move from a moratorium toward conditions on development.

The second signal is the first substantial public accounting of costs under the new utility tariff. The Public Service Commission required reporting and created mechanisms for future adjustments.

Those records should reveal how dedicated generation, transmission, and customer commitments work in practice. They will test whether large users truly pay their full incremental costs.

Evidence of effective cost isolation would support the industry’s case. Unexpected exposure for existing customers would revive Hong’s argument that Wisconsin acted before establishing sufficient protections.

Regulators should also watch customer forecasts. AI infrastructure plans can change as chips become more efficient, model demand shifts, or companies move workloads between regions.

A contract designed around one load projection must remain protective when the customer uses less electricity than expected. Long terms help, but enforcement and financial security remain essential.

The third signal is a renewed legislative package. Both parties previously introduced data-center measures, yet neither proposal reached the governor.

A new bill could address renewable energy, labor standards, disclosure, local consent, tax limits, or reporting. Its committee progress will show whether voter concern can overcome lobbying and partisan disagreement.

This signal can weaken or strengthen the “industry won” thesis. A comprehensive bipartisan bill would show that the election accelerated regulation through another route.

Another stalled session would give developers additional time under current rules. It would also make local fights and agency proceedings more consequential.

The tax exemption deserves particular attention. Other states have reconsidered open-ended incentives or added time limits and performance conditions.

Wisconsin lawmakers can preserve an incentive while requiring clearer returns. They can also connect benefits to verified investment, employment, clean energy, or community agreements.

None of these actions requires Hong to become governor. Her campaign helped create the political demand, but other officials can translate that demand into narrower policy.

This is the final reversal behind the google news headline. The candidate lost, while several elements of her critique have already entered mainstream regulatory debate.

AI data centers escaped the most direct proposed restriction. They did not escape the questions that made the restriction politically viable.

For developers, the next move should be disclosure backed by enforceable commitments. For regulators, it should be consistent measurement across projects and utilities.

For Wisconsin voters, the useful question is no longer whether Hong won. It is whether the state can verify who pays, who benefits, and who decides before the next campus advances.

Track the general-election platforms, the first tariff reports, and the next legislative session. Those three signals will show whether this was a durable industry victory or only a brief reprieve.