A large investment announcement does not tell a community whether a data center incentive is a good bargain. The useful comparison includes what the government gives up, what it receives and who carries the cost if the project changes. A Senate Democrats' report released on October 9, 2026 puts those questions at the center of AI infrastructure growth.
The inquiry covered Amazon, Google, Meta, Microsoft, CoreWeave, Digital Realty and Equinix. The senators said the companies continue to seek incentives while supplying incomplete comparable evidence of permanent employment. This is the report's finding, not a standardized nationwide audit proving that every incentive fails.
Separate investment from additional investment
A subsidy has additional value when it changes behavior that would otherwise not occur. A company buying equipment it already intended to replace may meet a spending threshold without demonstrating that the exemption caused the investment.
The completed source cites a Washington review of four existing urban data centers. Auditors could not determine how much spending resulted from the preference and found no new facilities under that program. That is evidence about the reviewed preference, not every state incentive.
Location decisions still depend on power, networks, land, permits and competing offers. Supporters argue that incentives can attract projects and strengthen local property revenue. Construction work can also be substantial. Those benefits deserve measurement alongside forgone revenue, rather than dismissal or automatic acceptance.
Keep separate ledgers for separate effects
An equipment exemption, a property agreement and an electricity tariff do different things. A site can receive one concession and still make other significant payments. Conversely, collecting local taxes does not show that residents are protected from grid costs.
Record the recipient, concession, duration and estimated fiscal effect. Separately record direct taxes, construction work, permanent positions, wages, local purchasing and public service costs. Use consistent periods and distinguish a proposed expansion from an operating phase.
The comparison should also account for ordinary tax treatment. Some critics of the Senate report argue that exempting business inputs avoids repeated taxation through production. That does not validate every negotiated package, but it means that the baseline needs to be stated before labeling an amount a subsidy.
Ask who pays when demand changes
Companies broadly support paying for infrastructure dedicated to their sites. The harder dispute concerns shared improvements that would not have been needed without a new large load. Such assets may help other users while also being triggered by the project.
A reduced or canceled campus can leave underused capacity. Deposits, minimum payments, exit conditions and dedicated rate arrangements can allocate that risk, but actual contracts and regulatory decisions determine the protection. A promise to pay full costs is less useful than a clear definition of covered assets and scenarios.
Disclose the bargain before it becomes binding
Confidentiality can protect land negotiations and sensitive engineering information. Public support nevertheless requires enough disclosure for residents to assess costs, power and water demand, jobs and enforceable obligations before decisive votes.
A good agreement separates temporary construction from permanent employment, defines measurement dates and specifies repayment or other consequences if commitments are missed. Independent review can test optimistic forecasts and identify benefits that depend on later phases.
The practical test is not whether all incentives are good or bad. It is whether this agreement changes investment decisions, delivers measurable public value and assigns downside risk clearly. For businesses relying on future AI capacity, those same conditions influence where projects can proceed and how reliably announced supply becomes operating infrastructure.