Rubio Tells NYT He Plans To Stay With Admin Until the End

Decision Tree Research CEO Greg Allen said AI data-center projects already underway are unlikely to be disrupted by growing political scrutiny and could bring tens of gigawatts of new compute capacity online over the next 18 months. Rep. Subramanyam said Democrats want to prevent data centers from imposing local burdens, including building near homes and schools or raising residents’ power costs.

Analysis

The key distinction is between political scrutiny and executable power access. Projects already financed or deep into construction may be hard to stop, but local opposition can still delay permits, transmission upgrades, or cost-allocation decisions—shifting returns from data-center developers toward utilities and grid-equipment suppliers able to deliver contracted capacity. The second-order risk is that communities and regulators push more upgrade costs onto large-load customers, weakening project economics even without outright bans.

Near term, rhetoric alone is not evidence of canceled capex; avoid shorting AI infrastructure solely on this signal. Over 1–3 months, watch local permitting outcomes, utility rate cases, interconnection queues, and whether hyperscalers disclose revised timelines or power costs. Over 6–18 months, the binding constraint may be energized megawatts rather than announced compute capacity. The tens-of-gigawatts estimate is a source's forecast, not independently verified delivery.

Contrarian angle: political scrutiny may improve the position of incumbents with existing sites, firm power, and community agreements by raising barriers to new entrants. But without project-level locations, power contracts, and cost-sharing terms, broad sector positioning has poor signal quality. Thesis weakens if permits and interconnections proceed on schedule without material cost increases; it strengthens if delays or customer-funded grid upgrades appear in filings and guidance.

Key Decisions for Investors

• Do not make a broad short of AI infrastructure on political commentary alone. Treat it as a catalyst to tighten diligence on project schedules and power procurement.
• Build a watchlist of regulated utilities and grid-equipment suppliers with verifiable, contracted data-center load or approved transmission investment; favor those over developers whose capacity is announced but not yet powered. Verify exposure in filings before initiating.
• For a conditional relative-value trade, consider long power-delivery beneficiaries versus underwrite-light data-center developers only after evidence of permitting or interconnection slippage. Falsify the thesis if project timelines and utility cost recovery remain intact.
• Track local permit decisions, utility rate cases, interconnection-queue changes, and hyperscaler disclosures over the next 1–3 months; absent concrete delays or cost pass-through, there may be no trade.