The Big Beautiful Bill and the Race to Secure Power

The One Big Beautiful Bill lowers the cost of some investments while withdrawing support from others. For power and compute developers, the opportunity may be to turn today’s tax benefits, financing and permitting conditions into operating infrastructure before competing projects become harder to complete. The advantage would belong not to whoever announces the most capacity, but to whoever can deliver useful power—and retain the resulting cash flow. S1 S2 S5 S6

Different clocks, not one four-year window

Public Law 119-21, enacted on July 4, 2025, does not create a uniform four-year development holiday. It changes several existing incentives and introduces others, each with its own eligibility requirements and timing. A change after the 2025–2029 presidential term is also different from a change four years from this article’s date, around 2030. Neither is a universal statutory expiration date. S1

The distinction between starting construction and placing an asset in service is particularly important. Some benefits depend on the first event, some on the second, and some on both.

Figure 1. The incentive calendar

Sources: IRS guidance and Internal Revenue Code provisions. This is a selected federal-policy summary, not a project eligibility opinion. S2 S3 S4 S5 S6 S7 S8 S30

There is still a development window, but it is better understood as several overlapping opportunities. Some remain open. Some have already closed to new qualifying starts. Some have no scheduled sunset under current law.

What a tax benefit actually buys

The restoration of 100% bonus depreciation allows qualifying investment to be deducted immediately rather than depreciated over several years. For a compute project, the analysis must separate servers, networking, eligible electrical equipment and other components from land and ordinary building structures. A campus is not one universally deductible asset. S2 S9

Consider a deliberately simplified example: $100 million of qualifying equipment, a 21% assumed federal marginal tax rate and full immediate use of the deduction.

$100 million × 100% deduction × 21% assumed tax rate = $21 million of immediate tax shielding.

Figure 2. A deduction is not a grant. The $21 million is the immediate shield in this worked example, not the incremental value over ordinary depreciation. Source: Ephesus research dossier, §1; IRS bonus-depreciation guidance. S2

The government is not paying the entire equipment bill. Nor does the example establish that the project is worth $21 million more: deductions would otherwise have arrived over time. Their earlier availability can improve cash flow and present value, but the incremental benefit requires an actual tax schedule.

The owner matters, too. A tenant’s GPU deduction does not automatically belong to the landlord. Specified rate-regulated utility property is excluded from bonus depreciation, so a merchant generator and a regulated utility cannot simply share the same tax assumptions. S9

The law also restores an EBITDA-style calculation for the business-interest deduction limitation and current deductibility for qualifying domestic research expenditures. Those changes can improve selected financing and investment economics, but neither creates unlimited interest deductibility nor makes every engineering invoice qualifying research. S10 S11

InferenceOur inference is straightforward: tax policy can help a viable project cross a financing threshold. It cannot establish the tenant, construction schedule or power connection that makes the project viable in the first place.

The factory incentive is not a blanket data-center write-off

The new qualified-production-property allowance creates a finite opportunity for qualifying manufacturing, chemical production, agricultural production and refining buildings. The activity must meet the relevant production and substantial-transformation requirements. Office, administrative, research, software-development and engineering areas are among the exclusions in the guidance. S3 S4

A data-center developer should therefore not assume that “producing compute” makes the entire building a qualifying factory.

The more defensible connection may be the infrastructure supply chain. Better economics for qualifying physical-production facilities could encourage investment in the equipment needed to build the next generation of campuses. That is an inference, not an established net outcome. Initially, a simultaneous construction rush could instead intensify competition for labor, materials and electrical connections.

The semiconductor incentive reinforces the distinction: the increased 35% §48D credit supports eligible semiconductor and semiconductor-equipment manufacturing. It is not a 35% credit for an ordinary data center purchasing chips. S7 S8

The investment question is whether new manufacturing capacity can arrive in time to reduce a developer’s costs—not merely whether an upstream supplier has a favorable tax provision.

The energy system gets a different set of incentives

The law does not treat all power pathways equally.

Wind and solar face an accelerated termination rule that generally affects facilities starting construction after July 4, 2026 and entering service after December 31, 2027. Earlier qualifying starts may preserve access, subject to the applicable rules. The July 2026 date is already behind us; it should not be presented as an opportunity that remains open for a new protected start. S5 S6 S8

Storage is not subject to that same accelerated cutoff, including qualifying storage co-located with wind or solar. Eligible non-wind/solar generation also has a longer general credit schedule. Storage can shift energy into constrained hours, but it must first be charged: installed megawatts of batteries are not equivalent to unlimited-duration generation. S5 S6

Fuel cells are a notable exception. For qualifying property beginning construction after December 31, 2025, §48E provides a flat 30% credit and removes the normal zero-greenhouse-gas eligibility test. The rate cannot be increased by other provisions of §48E. This is not a general 30% credit for gas turbines, engines or every behind-the-meter system. S5

Carbon capture also retains meaningful support. For qualifying new facilities and equipment, the law equalizes qualifying utilization and enhanced-oil-recovery treatment with storage rates. At the 2025–2026 wage-qualified level, headline credits include $85 per tonne for point-source capture and $180 for direct-air capture, with inflation adjustments thereafter. S12

That can improve selected project economics. It does not remove capture-equipment costs, energy consumption, transport, storage, monitoring or execution risk. Similarly, a fuel-cell tax advantage does not settle its delivered electricity cost, maintenance burden or suitability for a particular campus.

Some building incentives move in the opposite direction. The §179D energy-efficient-commercial-building deduction is unavailable for property whose construction begins after June 30, 2026. The §45L new-home credit and §30C charging/refueling credit have separate termination triggers tied to acquisition or service dates. S13

The relevant comparison is the complete, after-tax power system—not the largest advertised credit.

The risk: accelerating compute investment faster than power delivery

Our central hypothesis is a timing mismatch. Investment incentives may improve the economics of computing equipment while the physical systems needed to energize it remain slow to deliver. The law’s uneven treatment of generation could amplify that mismatch in some regions, although it does not establish that this will happen nationally.

The interconnection evidence explains why this deserves attention.

Berkeley Lab’s 2026 Queued Up release, using data through the end of 2025, identifies 549 GW with draft or executed interconnection agreements that had not reached commercial operation. For regions with available timing data, projects completed in 2025 had median request-to-operation times exceeding five years. In the historical 2000–2020 request cohort, only 13% of capacity was operating by the end of 2025. S14

Figure 3. An agreement is not an operating megawatt. The measures use different denominators and describe generation/storage queues, not data-center load requests. The historical 13% result is not a current-project completion probability. Source: Lawrence Berkeley National Laboratory and GridTracker, Queued Up, 2026 edition. S14

These figures do not prove that the law caused the bottleneck. They show why a tax benefit and a signed agreement should not be confused with deliverable capacity.

Demand needs equal discipline. Berkeley Lab’s older 2024 data-center study estimated 176 TWh of U.S. electricity use in 2023 and a 325–580 TWh scenario range for 2028. That is not a fresh 2026 measurement, a confidence interval or a binding commitment from customers. Annual energy use also differs from the dependable power a site needs at a particular hour. S15

The thesis requires both constrained supply and enough creditworthy demand to pay for it. Large pipeline announcements alone establish neither.

A policy reversal would reach operating supply with a lag

Suppose policy changes around 2029–2030 and makes some new power projects less attractive. The immediate effect would most naturally appear in investment decisions and new starts—not necessarily in a sudden disappearance of operating electricity supply.

Existing assets, projects already under construction and proposed projects are different categories.

Figure 4. One possible transmission mechanism—not a forecast or fixed construction schedule

Source: Ephesus conditional scenario analysis. No dates or numerical capacity losses are assigned to the stages.

The potential opportunity sits in that lag. A developer that reaches operation before competing starts become less attractive may own something increasingly difficult to replicate. But an owner still waiting for turbines, a substation, permits or project financing may be harmed by exactly the same scarcity.

Consider two hypothetical developers. One has an energized campus, enforceable power rights and a credible tenant. The other has land, a partially completed shell and an unresolved connection. A later tightening of development conditions could strengthen the first project’s position while delaying or undermining the second.

Power scarcity is not automatically good for developers. It may help owners of usable power while hurting developers still trying to obtain it.

Nor does a change of administration necessarily reduce total power development. A future government could restrict one technology while accelerating renewables, storage or transmission. Alternatively, budgets, local opposition and infrastructure-cost allocation could restrain several pathways at once. Those are competing scenarios, not predictions about a particular election outcome.

Federal incentives do not eliminate local policy risk

Some policy uncertainty is already visible within the current cycle.

Texas directed a comprehensive data-center audit in August 2026. A June directive separately called for protecting residential customers from infrastructure costs associated with data-center expansion. These are evidence of scrutiny and policy direction—not proof that every announced future legislative measure has already become law. S16 S17

For a developer, connection costs, water arrangements and community acceptance may determine whether an otherwise tax-efficient campus is economically buildable. Federal incentives are only one layer of that calculation.

Permitting policy also has different durability from tax legislation. The July 2025 executive order on data-center infrastructure is separate from the tax statute and itself revoked a January 2025 predecessor. Executive priorities should not be treated as permanent statutory rights. S18

The law’s NEPA opt-in route allows a sponsor to pay 125% of anticipated review costs for accelerated environmental-review timelines. Faster review is not guaranteed project approval and does not erase other permits. S19

The inference is that project quality includes more than a favorable tax model. It includes the ability to carry an investable design through the relevant regulatory, physical and community constraints.

The strongest counterargument: supply can respond

An early-asset advantage is not inevitable. The same incentives that pull development forward may encourage enough new investment to reduce scarcity later.

GE Vernova’s July 2026 results set out annual global gas-turbine output targets of 20 GW in the third quarter of 2026, 24 GW in 2028 and 30 GW in 2030. Those are company targets—not evidence that the corresponding U.S. power plants have been delivered or commissioned—but they illustrate a potential supply response. S20

Grid financing is moving as well. DOE reports a July 2026 loan closing for AEP Texas of up to $3.26 billion, supporting more than 2,800 miles of rebuilt, reconductored or new transmission. That is neither 2,800 miles of entirely new lines nor proof the works are complete. It is evidence against assuming that transmission supply will remain static. S21

There is also a cost counterargument to the legislation itself. Rhodium’s July 2025 analysis modeled industrial energy expenditures in 2035 at $7–11 billion per year above its baseline. This was a modeled counterfactual, not a measured result or an updated 2026 forecast. It reinforces the need to distinguish cheaper investment from cheaper electricity. S22

Falling realized demand, better equipment availability or more flexible consumption could all weaken the scarcity thesis. A rush to qualify can also lock a developer into expensive hardware or unfavorable contracts. Waiting may forfeit a benefit but improve other parts of the investment.

The correct comparison is risk-adjusted project value under both schedules—not the value of the tax incentive in isolation.

Who actually captures the advantage?

Even when a project benefits, the listed company’s common shareholders may not receive all of the gain.

A landlord may not own the tenant’s deductible equipment. A tax credit may sit inside a joint venture. Project debt, preferred capital, partner allocations and the parent’s economic interest can stand between a nominal project benefit and shareholder cash flow.

Contract terms can reverse the headline conclusion. Higher power prices may help an unhedged generator while hurting a landlord that promised fixed-cost power. A genuine cost pass-through can protect the landlord’s direct margin, while still pressuring the tenant’s economics. A long lease can stabilize income but postpone participation in higher market rents until expansion or renewal.

These are analytical possibilities, not findings about the undisclosed contracts of any particular company.

Figure 5. Follow the benefit to its owner

Qualifying asset and tax owner → usable deduction or credit → project cash flow after costs and financing → partner allocations and parent ownership → residual value for shareholders.

Conceptual flow only. Each step can reduce, defer or redirect the benefit.

Tax benefits themselves also interact. Energy investment credits generally reduce depreciable basis by half the credit. In a simplified $100 million example, an assumed 30% investment tax credit leaves $85 million of depreciable basis. At an assumed 21% tax rate and full bonus depreciation, the deduction shields $17.85 million; the combined nominal federal benefit is $47.85 million, not $51 million. S23

That calculation assumes eligibility and full use; it excludes timing, state tax, fees, limitations and recapture. It is not the incremental benefit of the 2025 law. Prohibited-foreign-entity restrictions and domestic-content bonuses are also separate tests, while a transferable credit’s face value is not necessarily its net cash proceeds. S24 S25 S26

For valuation, keep tax savings, financing effects, operating margins and speculative scarcity value separate. A mechanical deduction calculation is not permission to increase a terminal multiple.

Beyond compute: a broader development map

The law’s development effects extend beyond a data-center fence.

The Opportunity Zone framework is made permanent through recurring designation cycles, with new zones scheduled to take effect in January 2027. Affordable-housing credit allocations increase by 12% beginning in 2026. The New Markets Tax Credit receives a permanent $5 billion annual qualified-investment allocation—not $5 billion of annual grants. S27 S28 S29

These provisions may matter to qualifying rural development, workforce housing and community investment. They do not automatically subsidize a compute campus or solve its connection problem.

For site selection, our hypothesis is that a usable existing industrial connection or a credible expansion right can matter more than cheap land alone. But retained rights, contamination, refurbishment costs, water and demand must be tested individually. Behind-the-meter generation can likewise replace some grid dependencies with equipment, fuel-delivery, air-permitting and maintenance dependencies.

The most attractive location is therefore not necessarily the one advertising the largest incentive package. It is the one where the full project can be delivered at an attractive risk-adjusted return.

Three outcomes—not one prediction

The following scenarios are framed from the perspective of an owner of an energized, power-backed asset, not the development sector as a whole.

These are conditional analytical scenarios. No numerical probabilities are assigned, and none implies a particular company’s stock will rise.

Evidence that would strengthen the thesis includes completed generation, energized substations, enforceable power and fuel rights, closed financing and customer commitments backed by credible credit. Those milestones must then be compared with the supply actually competing for the same tenants and delivery dates.

Evidence that would weaken it includes repeated reductions in firm demand, better equipment delivery times, improving regional supply conditions and contracts or regulation that pass most benefits away from the owner.

The window is to deliver—not merely to begin

The potentially durable advantage is not announcing a campus during a favorable political period. It is converting favorable conditions into infrastructure that works, serves creditworthy demand and produces cash flow the owner can retain.

A temporary investment advantage can become a long-lived operating advantage. It can also be competed away, absorbed by construction costs or stranded in an unfinished project.

For investors, the decisive questions are therefore practical: What is actually energized? What remains to be financed and built? Who pays when power becomes expensive? Who owns the benefit?

The race is not simply to secure a tax break. It is to secure deliverable power before the economics of the next project change.

---

Research boundaries and disclosure

This article is based on the September 4, 2026 Ephesus research dossier, with primary-source checks for the cited headline provisions and evidence. It is a federal-policy and infrastructure framework, not an exhaustive state-incentive review, project tax opinion or company valuation. It does not establish eligibility for a specific NUAI, IREN or peer asset.

The original research left the latest construction-safe-harbor litigation and appeal/stay status unresolved. This article does not treat either a physical-work-only rule or a restored five-percent safe harbor as settled project advice. Current guidance, litigation, asset classification and ownership must be checked before a project-level eligibility claim. The construction examples, ownership analysis and scenarios are inference; numerical worked examples are assumptions; supplier targets are company guidance.

Ephesus Research provides research and education, not personalized investment, tax or legal advice. Actual outcomes may differ materially from the scenarios discussed.

Sources

S1. Public Law 119-21 — enrolled H.R. 1 — U.S. Congress. July 4, 2025.

S2. Guidance on additional first-year depreciation — Internal Revenue Service. January 14, 2026.

S3. Guidance on qualified production property — Internal Revenue Service. February 20, 2026.

S4. Internal Revenue Bulletin 2026-11, including Notice 2026-16 — Internal Revenue Service. March 9, 2026.

S5. 26 U.S.C. §48E — Clean electricity investment credit — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.

S6. 26 U.S.C. §45Y — Clean electricity production credit — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.

S7. Advanced Manufacturing Investment Credit — Internal Revenue Service. Guidance consulted September 4, 2026.

S8. Instructions for Form 3468 — Investment Credit — Internal Revenue Service. 2025 tax-year instructions.

S9. 26 U.S.C. §168 — Accelerated cost recovery system — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.

S10. Updated FAQs on business-interest expense limitations — Internal Revenue Service. December 23, 2025.

S11. Instructions for Form 6765 — Research activities — Internal Revenue Service. December 2025 revision.

S12. 26 U.S.C. §45Q — Carbon oxide sequestration credit — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.

S13. FAQs on modified and terminated energy and building provisions — Internal Revenue Service. Guidance consulted September 4, 2026.

S14. Queued Up: 2026 Edition — Lawrence Berkeley National Laboratory and GridTracker. June 2026; data through end-2025.

S15. Berkeley Lab Report Evaluates Increase in Electricity Demand from Data Centers — Lawrence Berkeley National Laboratory. January 15, 2025; discusses 2024 study.

S16. Governor Abbott Directs Comprehensive Data Center Audit — Office of the Texas Governor. August 3, 2026.

S17. Governor Abbott Directs PUC and ERCOT to Shield Texans from Data Center Infrastructure Costs — Office of the Texas Governor. June 10, 2026.

S18. Accelerating Federal Permitting of Data Center Infrastructure — The White House, Executive Order 14318. July 23, 2025.

S19. 42 U.S.C. §4336f — Project sponsor opt-in fees — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.

S20. GE Vernova Releases Second Quarter 2026 Financial Results — GE Vernova. July 22, 2026.

S21. AEP Texas — Energy Dominance Financing project — U.S. Department of Energy. Loan closed July 2026; page consulted September 4, 2026.

S22. What Passage of the One Big Beautiful Bill Means for U.S. Energy and the Economy — Rhodium Group. July 11, 2025.

S23. 26 U.S.C. §50 — Other special rules — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.

S24. Guidance on material assistance from prohibited foreign entities — Internal Revenue Service. February 12, 2026.

S25. Domestic content bonus credit — Internal Revenue Service. Guidance in the September 4, 2026 Ephesus research dossier.

S26. 26 U.S.C. §6418 — Transfer of certain credits — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text in the September 4, 2026 Ephesus research dossier.

S27. Guidance for state nominations of Qualified Opportunity Zones — Internal Revenue Service. April 6, 2026.

S28. 26 U.S.C. §42 — Low-income housing credit — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.

S29. 26 U.S.C. §45D — New markets tax credit — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.

S30. 26 U.S.C. §45V — Credit for production of clean hydrogen — U.S. Code, reproduced by Cornell Legal Information Institute. Codified text consulted September 4, 2026.