244 Gigawatts at Risk? Why Wall Street Is Betting the Politics Will Pass
244 Gigawatts at Risk? Why Wall Street Is Betting the Politics Will Pass
More than half of America’s data center pipeline faces election risk. Bernstein’s argument is that the economics will win in the end. Here’s how I read it, and where I’d be more careful.
| AI on the Ballot · Part 4 of 5 · James Chiao | Founder, iTAC · President, TSSA · Smart City Strategist · November 3, 2026 (US Election Day) |
This piece discusses published sell-side research. It is not investment advice.
On Monday, the day before the vote, Bernstein put out a note with a number that got a lot of attention. Of the roughly 433 gigawatts of data center projects in the US development pipeline, about 244 GW, around 56%, sits in places where the election could make things harder. Bernstein weighted candidates’ stated positions by their chances of winning and found that more than half the pipeline is exposed.
That’s a striking figure. Bernstein’s conclusion was even more striking: the risk is overstated, and investors should treat the sell-off in AI infrastructure stocks as a chance to buy.
I think the reasoning is mostly right. I also think it’s worth separating the parts that are solid from the parts that depend on how well physical systems hold up.
The headline number exaggerates the risk. The fine print understates it.
Why the politics may not hold
Bernstein’s key point is fiscal. In the states most exposed, data center construction is now a large part of the economy, roughly 8% of state-level GDP by its estimate. That means tax receipts, construction jobs and local infrastructure funding. A governor who campaigned on a moratorium still has to balance a budget in January.
We’re already seeing signs of this. Texas Governor Greg Abbott went from promoting data centers to ordering reviews of interconnection requests and supporting a pause on incentives. Illinois Governor JB Pritzker has called for new rules too. Notice that both are asking for a pause or new rules, not an outright ban. New York’s moratorium, the toughest step any state has taken, runs for up to a year. These look like negotiations, not a shutdown.
Underneath all of it, the demand case is intact. As Part 3 of this series explained, the parties are mostly arguing over who pays for the power, not whether the buildout continues.
Bernstein’s picks, grouped by what they sell
The note’s logic is that developers will absorb higher grid and compliance costs to keep projects moving, and that cost flows to the companies selling equipment and services.
|
Segment |
Names in the note |
Why the thesis applies |
|
Grid build & electrical equipment |
Quanta Services, Eaton, Forgent, Legence |
Transmission, substations, switchgear; transformer backlogs run for years |
|
Cooling & building systems |
Vertiv, nVent, Johnson Controls, Trane |
Denser AI racks need liquid cooling; retrofits avoid new siting fights |
|
Data center REITs |
Equinix, Digital Realty |
Long leases with investment-grade tenants; metro footprints are hard to replicate |
|
Power generation (incl. Europe) |
GE Vernova, Siemens, Siemens Energy, Schneider Electric |
Gas turbines, grid electronics, energy management; firm power is the bottleneck |
The common thread: companies with multi-year order backlogs are the most insulated from a permitting delay in any one county. If a project slips six months in Virginia, the transformer just ships to the next project in line.
Where I’d be more careful
Three things the bullish case tends to underplay:
1. Reliability risk is real. The July event in Northern Virginia, when 3,800 MW of data center load dropped off the grid, is the kind of thing that turns a debate about bills into a debate about safety. New ride-through rules will add costs and delays, though they also create more demand for the equipment makers listed above.
2. Local opposition is deeper than one election cycle. When 50% of Americans say they’d oppose a data center near their home, that won’t disappear when a new governor takes office. It will show up at zoning hearings for years.
3. The REIT case depends on land approvals. Equipment makers can redirect shipments elsewhere. A REIT can’t move a site. That’s why I’d rate their political exposure as moderate, not low.
What this means for Taiwan
Taiwanese investors and companies have a direct stake here. Many of the cooling, power supply and server suppliers in the AI buildout are Taiwanese, either as competitors or as upstream suppliers to the companies above. If Bernstein is right that US developers will “absorb higher costs,” that cost increase moves up the supply chain, and Taiwan’s thermal management and power electronics companies are part of that chain.
Jim’s TakeThe headline number exaggerates the risk. The fine print understates it. The buildout will continue because both the fiscal pull and the physical demand are real. But it will continue under new conditions: tech companies paying for their own power, data centers required to ride through grid faults, and residents who have learned to show up at zoning meetings. For anyone allocating capital, my rule would be to favor backlog over sentiment. Companies whose earnings depend on equipment that’s already ordered will get through this better than those that depend on permits not yet issued. |
Discussion question: Which Taiwanese suppliers are best positioned to benefit if US data centers are forced to fund their own power and upgrade cooling in existing sites?
James Chiao (焦國安) is the Founder of iTAC and President of the Taiwan Smart Solutions Association (TSSA), a 340-member ecosystem spanning 27 smart solution domains. He advises international development institutions and has worked on smart city and digital transformation projects across Taiwan, Europe, Southeast Asia, the US and China.
For partnership inquiries: info@smartcitiesgroup.net
#AIInfrastructure #DataCenters #GridEquipment #Investing #AI基礎建設 #電力設備 #資料中心 #美國期中選舉