Midterms Stall Federal Policy; Data Center Risk Shifts to States

Three states — Texas, Georgia and Ohio — have proposed data center moratoriums that could delay or restrict new projects, according to Paul Eitelman, chief investment strategist at Russell Investments, in an outlook published November 4, 2026.

The proposed pauses arrive after a midterm election that produced a divided government, an outcome that makes major federal policy changes unlikely. That combination matters more than the vote count itself. With Washington less able to act on permitting, power procurement and tax treatment for compute campuses, state capitals become the decisive venue for AI infrastructure.

Russell Investments has operated since 1936 and oversees US$331 billion in assets under management as of December 31, 2024. The firm serves clients in 30 countries from offices in 17 cities, headquartered in Seattle, Washington, and describes itself as a global investment solutions partner to institutional investors, financial intermediaries and individual investors. Eitelman's assessment carries a timestamp of 10:00 a.m. Eastern Time on November 4, 2026, the original publication date of the note.

The note reads less as a forecast of the next election than as a rerating of risk inside the current one. Eitelman's framing separates what the vote did to the market, which he argues was little, from what the vote did to the process of building, which he argues was considerable. For AI infrastructure, the second channel is the one with revenue attached.

What did the midterm results actually change?

The asset-allocation view, essentially nothing. Russell Investments sees no change to its broad constructive market outlook, and it rests that stance on two supports: "resilient global growth and earnings." What the election adds is a second layer of risk, which the firm flags as increased "political and execution risks" for AI infrastructure into 2027.

Eitelman's political reasoning compresses into one line: a divided government makes major federal policy changes unlikely. The market consequences run in a single direction. Federal inaction removes Washington's near-term action from the AI buildout, and with it removes federal action as a stabilizing variable. Approval of campuses then depends on a patchwork of state legislatures, county boards and utility commissions, each with its own timetable.

The risk did not vanish. It relocated.

Why do state-level moratoriums carry more weight now?

The note is explicit about the mechanism. Proposed moratoriums in Texas, Georgia and Ohio "could delay or restrict new data center projects, raising costs and execution risks for the AI infrastructure supply chain."

A moratorium is a pause on new permitting or construction in a jurisdiction. In general terms, a data center campus advances only when site entitlements, an electric interconnection agreement and a credible path to new generation align. Interrupt any one of those steps and the capital plan behind it slips, along with the equipment orders and contractor schedules attached to it. The note offers no offsetting federal lever to counter state-level pauses, which is why it links the two effects directly.

For the supply chain, the shift from policy risk to execution risk has a specific shape. Project timelines move to the right, financing gets repriced, and delivery dates for large electrical and cooling packages get pushed. Vendors that built backlog on assumptions of continuous interconnection queues carry schedule risk instead, and the projects whose economics assumed a specific power-in-service date lose their basis.

Which industries are most exposed?

Russell Investments highlights six segments as the ones potentially affected if the state proposals become law and delay projects:

• Construction — site preparation, foundations and the building shell, the earliest line item on any campus.
• HVAC (heating, ventilation and air conditioning) — the cooling load that dense compute halls impose on a facility.
• Electrical equipment — switchgear, transformers and cabling that feed and protect installed racks.
• Industrial machinery — the generation equipment and balance-of-plant systems behind new capacity.
• Utilities — the power and water providers whose capital plans assume sustained load growth.
• AI hardware — accelerators and servers, whose demand exists only once a finished facility can host them.

The list is deliberately broader than the obvious one. Investors who trade AI hardware often assume only the chip and server vendors carry policy exposure. Russell Investments argues the exposure starts upstream, in the companies that build the sites and supply the power, and only ends downstream at the racks.

Exposure also runs in sequence rather than in parallel. Utilities set the load forecast, industrial machinery and electrical equipment follow it, construction and HVAC turn a permit into a building, and AI hardware fills the racks that justify the capital expenditure. A delay at the first link propagates through the rest, which is why the firm lists suppliers and asset owners together instead of treating AI hardware as the only exposed category.

What does this mean for AI infrastructure into 2027?

Execution risk, in plain terms, is the risk that a project gets approved, financed and contracted but arrives late or at a higher cost. It differs from policy risk, which is the risk that a rule changes. The midterm shifts AI infrastructure exposure from the second category toward the first, which is why Eitelman ties his political observation directly to three state proposals rather than to a broader macro argument.

The firm leaves its broad market outlook unchanged and moves the political and execution risks on AI infrastructure into 2027. For anyone modeling the buildout, that is the operative detail. The calendar is no longer one national vote followed by a single federal decision; it is a sequence of state legislative sessions, regulatory filings and utility dockets in Texas, Georgia and Ohio.

Three things follow from that sequence:

• Whether the Texas, Georgia and Ohio proposals advance in their respective statehouses, and how far they travel.
• The scope of each pause, including start date, duration and treatment of permits already issued, none of which the note specifies.
• Order books and backlog across the six exposed sectors, where schedule slippage surfaces before it surfaces in earnings.

Eitelman and Russell Investments have not changed the market call. They have changed the risk map. For AI infrastructure, the 2027 buildout schedule now depends less on the composition of Congress than on how three state capitals handle a permitting question.