Issue 07: The Vote That Knocked Almost a Quarter off Data Center Stocks

She can keep the freeze without anyone's permission. Governors direct agencies through executive orders, and courts have largely held that the constitution lets the governor manage the executive branch. Abbott's own freeze rests on a letter to the PUCT and ERCOT; reporting on it cites no specific statute. Whatever legal footing he had, she inherits. The fastest path for her is not a new ban. It is simply to tell the same agencies that the audit is not finished. She takes office a week after the legislature convenes, about eleven weeks after the vote. The audit was due to finish only about a month before that, so unless Abbott lifts the freeze in his last weeks, there is little gap for approvals to restart.

She cannot quickly remake the regulator. The governor appoints PUCT commissioners, and Jefferies thinks her biggest influence will be on the commission's direction, though members would be replaced gradually as terms end. Appointees need a two-thirds Senate vote to be confirmed, and Republicans will hold 19 of the 31 Senate seats. Every commissioner she names will need Republican votes.

She cannot pass laws alone, but she can block them. Democrats are projected to hold 83 of 150 House seats, their first majority there in more than two decades. The Senate stays Republican, and its majority can rewrite the rules that give the new Democratic lieutenant governor her power over which bills reach the floor. So the parts of Hinojosa's plan that need a statute, such as giving local communities a vote on projects, union rules and water protections, face a Republican Senate that can stall them. Her veto runs the other way: an override needs two-thirds of both chambers, and Texas has not seen a successful one in more than four decades. Any bill that tries to override her freeze or protect the industry is unlikely to become law.

The tax break is probably going either way. Texas exempts qualifying data centers from its 6.25% state sales tax on servers, cooling, generators, software and the electricity they use. The comptroller estimates the exemption already costs at least $1.3 billion a year, rising toward $1.8 billion by the end of the decade, up from $5 million to $30 million a year in its early years. Abbott himself pledged to seek repeal, and the Republican chair of the Senate Finance Committee called the numbers "extremely concerning". This is the one change with support on both sides, so investors should treat it as likely rather than as a Hinojosa risk.

How much is that tax break worth per megawatt? Epoch AI puts the capital cost of the Abilene Stargate campus at about $31.2 billion for 824 MW, or roughly $38 million per megawatt including chips. At 6.25%, that is up to about $2.4 million of state sales tax per megawatt if everything bought were taxable, before the tax on electricity. Most of that would fall on whoever buys the chips, usually the tenant rather than the landlord. That gives tenants a reason to price Texas against other states. It does not make a Texas campus uneconomic.

Grid rules are already tougher, under a Republican law. Senate Bill 6, signed last year, makes data centers pay for grid upgrades, standardizes how they request power and requires them to join grid stability programs. Abbott's freeze applies to projects asking for 75 MW or more. The direction of travel, paying more and waiting longer, was set before this vote.

The governor controls the pause; the legislature controls anything permanent

Governor aloneFast, from her first day

• Keep Abbott's audit and approvals freeze in place
• Direct the PUCT, ERCOT and the environmental agency
• Veto bills; overrides need two-thirds of both chambers

Can delay new projects for months

Needs the legislatureDemocratic House, Republican Senate

• Repeal the sales tax exemption (support from both parties)
• Local votes on projects, union and water rules
• Confirm her PUCT picks (two-thirds of the Senate)

Tax break likely goes; the rest is contested

Out of reachWithout new law or a court fight

• Halls already energized and running
• Sites that make their own power
• Parts of Texas outside ERCOT

Existing capacity keeps earning

Ohio, Georgia and Washington

Texas was the shock, but it was not the only result that matters for data centers.

Ohio. Democrat Amy Acton beat Vivek Ramaswamy for governor. She backs a "conditional moratorium": data centers would have to pay 100% of their utility costs and build only with union labor. Her campaign added environmental standards, community benefit agreements and an end to nondisclosure agreements. The governor's main levers are appointments: five members of the Public Utilities Commission of Ohio, one member of the Power Siting Board and the board of JobsOhio, the state's development agency. Republicans keep majorities in both chambers of the legislature, but they lost the House supermajority they needed to override her vetoes without Democratic votes. Ohio had already stopped considering new data center tax exemption requests earlier this year, so as in Texas, the subsidy fight is partly over.

The big Ohio asset is the PORTS-Pike Technology Campus. OpenAI said it had secured 8 GW of computing capacity there; SoftBank and its energy arm SB Energy are developing it with the US Department of Energy, Nvidia is backstopping it, and the total cost could top $500 billion. Capacity is due online in stages from the year after next. Morgan Stanley said Acton's win creates "a significant risk of data center project delays" in Ohio, including a 10 GW project involving Nvidia that is the state's largest. Nvidia is also the backstop at PORTS-Pike. Either way, Ohio is a smaller slice of the country than Texas: Wells Fargo puts it at 5% of planned gigawatts, with Georgia at 7% and Pennsylvania at 5%. Local voters pushed the same way: of 41 local data center measures in 10 states, voters sided against data centers in every one with results except Harwood, North Dakota.

Ohio's risk is concentrated in one giant campus

8 GWOpenAI capacity at PORTS-PikeMorgan Stanley cites 10 GW

$500B+Possible total costNvidia is backstopping it

5%Ohio share of planned US gigawattsTexas holds 27%

The wider tally. Goldman Sachs expects Democratic state and local wins to bring temporary halts that could touch about 45% of planned US capacity, but sees them as likely "short-lived". UBS estimates that around 30% of expected US data center capacity additions in the year after next could be at risk from construction pauses. Jefferies counts about 13 GW of announced Texas projects as immediately at risk. These are analyst estimates of exposure, not policy. Notice that all three are about the pipeline two or more years out, not about capacity running today.

Washington. Democrats took both chambers of Congress. Expect noise: Representative Greg Casar said "AI company CEOs should be called to testify under oath," and Representative Robert Garcia's message was "Lawyer up." Dan Ives of Yorkville Ives & Co. expects "some major hurdles for AI and tech legislation" over the next two years. Hearings, subpoenas and budget riders are real tools. But any new law still needs the signature of a Republican president, and an override needs two-thirds of both chambers. Federal agencies keep their current leadership. And most of Texas's grid, the ERCOT region, sits largely outside federal wholesale power regulation. For the physical build-out, Washington is a headline risk. The binding constraints are in state capitals.

The exposure map

Here is how the listed names line up, split by what they own. Moves are closing prices from Yahoo Finance and match the day's closing table. The last column measures over two days, from the close before election day, for a reason we explain below.

Landlords and AI clouds

Power and equipment

Sources for the exposure columns: Hut 8 and Stock Titan for Beacon Point; Cipher's site list, Black Pearl bond filing and earnings call; Galaxy; IREN; Core Scientific; TeraWulf; Applied Digital; Oracle via Epoch AI; Vistra; suppliers.

Texas-heavy landlords and contractors fell hardest; builders outside Texas fell least

Hut 8−23.2%

Sterling−21.5%

Modine−21.4%

Cipher−20.0%

Galaxy−17.0%

Comfort Systems−15.1%

Vistra−13.8%

Vertiv−12.9%

IREN−11.2%

Core Scientific−10.4%

Applied Digital−6.7%

TeraWulf−6.6%

S&P 5001.6%

Close to close, election day to the day after the vote.

Four things stand out.

Hut 8 is the purest Texas bet. Its Beacon Point campus near Corpus Christi holds 1,000 of the 1,330 MW it has under construction and about $19.6 billion of its $26.6 billion in contracted lease value. Phase 1 is funded with $4.25 billion of non-recourse project bonds, and its first data hall is due in about a year. Under construction does not mean fully approved, though: Hut 8 says its grid classification at Beacon Point is still subject to Abbott's audit, with results expected before year end, and that the vote does not change the outlook for what it calls a late-stage project. Phase 2, another 352 MW of IT capacity worth about $9.8 billion in base rent, is still in civil design, with its first hall due about a year and a half out. Its other campus, River Bend, is in Louisiana. Hut 8 also reported results that same morning, with revenue down on a year earlier, so not all of its fall was politics.

Cipher's leased capacity is safer than its fall suggests. Its two big leased campuses, Barber Lake and Black Pearl, are listed as energized. The Black Pearl bond documents describe a newly built substation that is "fully approved" for 300 MW with "no pending approvals", leased to Amazon with Amazon guaranteeing the rent. A freeze on approvals has nothing to grab there. Cipher's exposure is its pipeline, led by the 1 GW Colchis site, which has only a conditional classification in Batch Zero. Management says it has complied with the audit and expects a final classification once it ends. Colchis is targeted for about two years out. That is exactly the bucket a long freeze delays.

Galaxy owns something the freeze makes rarer. Helios already delivers 133 MW to CoreWeave, with Phase II fully financed and under construction. Helios II has 830 MW of approved capacity not yet leased. If new approvals stop, already-approved Texas power becomes scarcer, and Galaxy is negotiating to lease exactly that. The market sold it 17% anyway.

Power producers got hit for the right reason, in the wrong year. Vistra has about 45% of its fleet in Texas. Fewer data centers means less demand growth and lower power prices than investors had hoped for. But Vistra had hedged, meaning sold forward at fixed prices, about 94% of its expected generation for next year, and about 72% for the year after. KeyBanc downgraded the stock, saying the result creates "significant policy uncertainty" for companies selling power to data centers.

There is one more pattern in the table. Many of the landlords had jumped on election day itself, before any result was known: on Yahoo Finance closing prices, Hut 8 rose 15.9%, Cipher 18.4% and Galaxy 15.4%. Measured over the two days, most of the miners-turned-landlords are roughly flat, and some are up. The contractors and power producers are not.

Landlords mostly gave back an election-day jump; contractors and power producers really fell

Day after the voteTwo days, from the close before election day

−25−20−15−10−55%0

−23.2%−11.0%

Hut 8Day after the vote−23.2%Two days, from the close before election day−11.0%

−20.0%−5.2%

CipherDay after the vote−20.0%Two days, from the close before election day−5.2%

−17.0%−4.2%

GalaxyDay after the vote−17.0%Two days, from the close before election day−4.2%

−11.2%−1.4%

IRENDay after the vote−11.2%Two days, from the close before election day−1.4%

−10.4%2.0%

Core Sci.Day after the vote−10.4%Two days, from the close before election day2.0%

−21.4%−20.1%

ModineDay after the vote−21.4%Two days, from the close before election day−20.1%

−21.5%−15.4%

SterlingDay after the vote−21.5%Two days, from the close before election day−15.4%

−13.8%−10.6%

VistraDay after the vote−13.8%Two days, from the close before election day−10.6%

Hut 8CipherGalaxyIRENCore Sci.ModineSterlingVistra

That changes the story. The headline "landlords down 20%" was partly a crowded trade reversing. The lasting repricing landed on the companies whose next two years of orders depend on new Texas projects starting: Modine, Sterling and the Texas power producers.

Did the sell-off overshoot? A worked model

To test the moves, separate a delay from a ban.

A data center is worth the rent it will collect, discounted back to today. Discounting means a dollar received later is worth less than a dollar now, because you could have invested it in the meantime. If a project is pushed back a year and nothing else changes, every rent payment arrives a year later. At a 10% discount rate, each payment is then worth 1 ÷ 1.10, or about 91% of what it was. A one-year delay costs about 9% of the project's value. Two years costs about 17%. Three years, about 25%. A ban is different: the value goes to zero, unless the land and the power hookup can be sold to someone else.

Watch it play outA two-year delay costs a project about 17% of its value; a ban costs all of it

Call the value of a project's future rent 100.

Delay it one year at a 10% discount rate and it is worth about 91.

Two years: about 83. Three years: about 75.

A ban takes it to zero, unless the land and grid connection can be sold.

Before counting debt, a 23% fall in a day is more than a two-year delay of everything Hut 8 owns would justify.

Replay

Two more pieces turn that into a stock price. First, what share of the company's value is actually delayed. Second, debt. If a company is funded half with debt and half with equity, a 9% fall in the value of the business is roughly an 18% fall for the shareholders, because the lenders still get paid in full. Many of these developers fund each campus with project bonds, so their equity is leveraged in exactly this way.

How much a delay should take off the shares

Pick how much of the company is delayed, for how long, and how it is funded

Share of value delayed

25%50%100%

Length of delay

1 year2 years3 years

Funding

No debtHalf debt

Fall in share value2.3%4.5%4.3%8.7%6.2%12.4%4.5%9.1%8.7%17.4%12.4%24.9%9.1%18.2%17.4%34.7%24.9%49.7%

2.3%49.7%

Debt equals the equity

Now run it the other way. Take each stock's fall and ask: if this were only a two-year delay, with no debt, how much of the whole company would have to be delayed to justify it? The answer is the fall divided by 17.4%.

For four of these stocks, even a two-year delay of the whole company would not justify the fall

Hut 8134%

Sterling124%

Modine123%

Cipher115%

Galaxy98%

Comfort Systems87%

Vistra80%

Vertiv74%

IREN65%

Core Scientific60%

GE Vernova54%

Share of company value that would need a two-year delay, at a 10% discount rate and no debt, to explain the fall the day after the vote. Above 100% means the market priced cancellations, not delays.

Read it in two halves.

For the suppliers, the falls price in more than Texas can deliver. Sterling, Modine and Comfort Systems sell across the country. Suppose Texas really is 27% of their future data center work and all of it slips two years. That is 27% × 17.4%, or about a 4.7% hit, even before counting their factory and chip plant work. Use Goldman's 45% for all the states that may restrict building and you get about 7.8%. They fell 15% to 21%. Comfort Systems lost about $6.8 billion of market value in a day, equal to 38% of its entire $17.8 billion backlog of signed revenue. What the market really marked down was the growth multiple, the premium paid for years of fast growth ahead. Comfort was up more than 30% this year before the drop. Multiples can stay lower for a long time, so "overshot" does not mean "bounces next week". But on cash flow, the suppliers with national order books took the least justified hit.

For the leveraged landlords, the model needs debt. Hut 8, Cipher and Galaxy fund campuses with project bonds, so their equity moves more than their business. With debt equal to equity, a 23% equity fall is consistent with about a two-year delay of two-thirds of the business. That is aggressive but not absurd for Hut 8, whose value is concentrated in one Texas campus. For Cipher, whose leased capacity is already energized, it is harder to justify.

Hut 8: what Phase 2 is worth

Hut 8 is the cleanest test, because the exposure is one campus and the numbers are public. Beacon Point Phase 2 carries about $655 million a year of average net operating income, the rent left after running costs, over a 15-year base term. Building cost for Phase 2 is not yet disclosed. As a stand-in we use the $4.25 billion of bonds raised for the identically sized Phase 1.

• Value of 15 years of $655 million, discounted at 8%: about $5.61 billion.
• Less the building cost of about $4.25 billion: about $1.36 billion of value created, measured at the start of operations.
• Cost of a one-year delay to Phase 2: about $0.10 billion. Two years: about $0.19 billion.
• Losing Phase 2 entirely: about $1.36 billion.

Now the market. Hut 8's 120 million shares lost $14.81 each the day after the vote, about $1.78 billion. Over two days, from the close before the election-day jump, the loss is about $0.73 billion.

Hut 8 lost more than Beacon Point Phase 2 is worth, and far more than a delay would cost

1-yr delay$0.10B

2-yr delay$0.19B

Two-day fall$0.73B

Phase 2 value$1.36B

Fall day after$1.78B

$0.0$0.5$1.0$1.5$2.0B

Measured from election day, the market took off more than Phase 2 is worth on our numbers, as if it would never be built and then some. Measured over two days, it priced roughly even odds of losing it. Delay alone, even two years of it, explains a small fraction of either number. For Hut 8, then, the sell-off is a bet that Phase 2 gets cancelled or loses its tenant, not that it slips. The real question is whether the freeze reaches a campus that already has 1,000 MW of utility power arranged and construction under way, but whose grid classification still waits on the audit. If the audit clears it and the new order covers only fresh approvals, the market overshot. If it also blocks final grid energization for halls already being built, it did not.

Vistra: the hedge buys time

Vistra lost about $6.5 billion of market value in a day, close to a full year of next year's adjusted EBITDA, which its finance chief said was tracking toward the bottom of a $7.4 billion to $7.8 billion range. EBITDA is earnings before interest, tax, depreciation and amortization, a rough measure of operating cash. A Texas freeze cannot touch next year much, since 94% of output is already sold forward. The risk is the year after, when only 72% is hedged, and later years when Texas demand growth was supposed to lift power prices. After the fall, the stock traded at about 12 times forward earnings, against a five-year average of 15 times.

Vistra's Texas exposure is large, but next year's output is already sold

45%Fleet in Texas19,858 of 43,641 MW

94%Next year hedgedSold forward at fixed prices

72%Year after hedgedWhere a freeze would bite first

Our read on Vistra: the fall is closer to fair than the suppliers' falls. Its upside case leaned on Texas data center demand, and that case now waits at least a year. But a forward multiple three points below its own five-year average already prices a lot of that in.

Second-order winners

A freeze on approvals does not reduce demand for computing power. It reduces supply of new places to put it. That shifts value toward three groups.

Owners of energized capacity, in Texas and elsewhere. BNEF's summer forecast had only 1.2 GW of new ERCOT data center capacity coming online over roughly a year. Its estimate of what waiting costs is striking: about $1.76 billion per gigawatt per month in lost AI computing revenue, against colocation rent of up to $175 million per gigawatt per month in Dallas. When a tenant loses ten times more from waiting than it pays in rent, a hall with power today can charge more. That favours Core Scientific's Denton site, Cipher's energized Barber Lake and Black Pearl, and Galaxy's approved but unleased 830 MW at Helios. All three fell 10% to 20% the day after the vote.

Sites in other states. HSBC's strategists put it plainly: one outcome is "data centers simply move from Texas to other states." Michael Webber of the University of Texas said "the threat of a moratorium is enough to get the developers to consider other states." Applied Digital is building in North Dakota, where Harwood voters rejected an extra review of large projects 325 to 105. TeraWulf sold its Texas joint venture stake and builds in New York and Kentucky, including a roughly $19 billion lease with Anthropic. IREN has a 1,600 MW Oklahoma campus behind its Texas sites. Applied Digital and TeraWulf each fell about 7%, in sympathy rather than for any change in their own approvals.

The cloud giants. Microsoft rose 6.7% and Amazon 6.1% the day after the vote. Investors are betting that slower construction will rein in their capital spending and make existing facilities worth more. Amazon alone expects about $220 billion of capital spending this year. If it builds the same capacity in other states, its suppliers lose little. If it simply builds less, they lose more.

A Texas freeze moves value toward capacity that already has power

Likely helpedScarcer supply, higher rents

• Energized Texas halls: Core Scientific Denton, Cipher's leased sites
• Approved but unleased power: Galaxy Helios II
• Out-of-state builders: Applied Digital, TeraWulf

Sold 7% to 20% anyway

vs

Truly exposedNeeds new Texas approvals

• Unapproved Texas pipelines: Cipher Colchis, Hut 8 Phase 2
• Texas power producers from the year after next
• Contractors whose next orders start in Texas

Risk is real, from the year after next

What would prove this right or wrong

Most of this turns on a handful of documents and numbers that will arrive over the next few months. Watch for them in this order.

1. The wording of Hinojosa's first order. "Approvals" covers new permits and new grid connections. If the order also halts final energization for projects with signed utility agreements, Hut 8's Beacon Point and IREN's Sweetwater move into the danger zone, and the landlord sell-off was not overdone. If it names only new approvals, the energized and under-construction campuses are clear.
2. The audit report from the PUCT and ERCOT. Does Batch Zero restart with the projects that complied, as Cipher expects for Colchis, and does Beacon Point get the final classification Hut 8 expects before year end? Or does it stay shut? A restart on schedule would cut the delay case for the pipeline to months.
3. The first PUCT nominations and how the Senate treats them. Nominees confirmed with Republican votes would point to a negotiated regime of higher costs and clearer rules. Stalled nominations would mean a longer standoff.
4. The tax exemption bill. A repeal that protects existing certificates would mostly hit new projects. A repeal that strips existing ones would hit every Texas landlord's tenants.
5. The next round of hyperscaler capital spending guidance. If total spending holds and new campuses are announced in other states, the suppliers' sell-off reverses. If guidance falls, the market was right. Our capex tracker updates as each company reports.
6. New Texas leases. A tenant signing for Galaxy's Helios II or more of IREN's Sweetwater during the freeze would be the clearest sign that approved Texas power commands a premium.
7. Vistra's results this week, then NRG's. Listen for whether Texas data center customers are pausing contract talks.

We are wrong if the order reaches projects already under construction, if courts or the Senate fail to check it, and if Georgia and Ohio copy it closely enough that Goldman's 45% becomes real rather than "short-lived". We are also wrong if the cloud giants respond by building less rather than building elsewhere.

Caveats

Two of the biggest fallers had company news the same day: Hut 8 reported results that morning, and Modine's data center gross margin had fallen to 19.7% from 24.4% in results the evening before. The election-day jump in the landlord stocks has no single clear cause in the news flow. The model uses round discount rates, and our Hut 8 Phase 2 cost is a proxy, not a disclosed figure. On its call, Hut 8 also said the Phase 2 lease term was extended from 15 to 17 years; on 17 years of the same income, Phase 2 is worth about $1.72 billion, still below the $1.78 billion fall the day after the vote. And the backdrop was already soft: Baird data cited by Schwab's Liz Ann Sonders showed data center project cancellations in the latest quarter that were almost larger than the two previous full years combined. Some of the selling may be about that, with the vote as the trigger.

Our read

The market treated a freeze on new approvals as if it were a ban on data centers in Texas. It is not. What Texas can deliver quickly is more of what it already had: no new grid connections or permits until the legislature rewrites the rules, a tax break that was on its way out anyway, and higher grid costs that a Republican law had already set in motion. That hits the early-stage gigawatts due from the year after next. It does not touch the halls that are already energized.

So we would split the sell-off into three parts:

• Overdone: suppliers with national order books. Comfort Systems, Vertiv and GE Vernova fell 9% to 15%, when the model says even a two-year Texas delay is worth a mid-single-digit hit. Their multiples may stay lower for a while, but this is where the gap between price and cash flow is widest.
• Overdone: landlords with energized or approved Texas capacity. That means Cipher's leased campuses, Core Scientific's Denton and Galaxy's Helios. Measured over two days, most of these barely fell, and a freeze makes what they own scarcer.
• About fair: unapproved Texas pipelines and the Texas power trade. That means Hut 8's Phase 2, Cipher's Colchis, IREN's Sweetwater beyond what is being built, and Vistra and NRG from the year after next. Hut 8 is the name to watch most closely. Its price now implies that Phase 2 may never be built, and the wording of one executive order will decide whether that is too pessimistic.

Go deeper

• How the stock market reacted to the midterms: the whole-market move, the rotation into the cloud giants and what past midterms say about the next 12 months.
• AI data center power demand: why grid connections, not chips, set the pace of the build-out.
• Hyperscaler capex tracker: the spending guidance that decides whether a Texas freeze moves capacity or cancels it.