AI-generated illustration — not a photo of any real place, product or person.
On a single Tuesday — 3 November 2026 — two announcements landed on opposite sides of the Atlantic and told the same story from opposite directions. In Yorkshire, England, Drax Group, Brookfield and Radiant announced a strategic partnership to build gigawatt-scale AI data centers at Drax Power Station, with Brookfield estimating that total potential investment could exceed £20 billion (about US$26.8 billion). In the United States, voters went to the polls and delivered the sharpest rebuke the data-center industry has ever received: 33 local ballot measures restricting data centers passed or were leading, per Stateline's count, while Texas and Georgia elected governors who had campaigned on pausing new construction. The money behind artificial intelligence has never been bigger, and the public permission it needs has never been harder to get. This article explains both sides of that collision — the biggest infrastructure bet of the AI era, and the voter revolt it ran into — using only verified facts from retrieved sources.
1. Two Tuesdays That Changed the AI Build-Out
The AI infrastructure story of November 2026 is really two stories — one about capital, one about consent — and they collided on 3 November.
The first story opened early on Tuesday morning in London. At 7:01am GMT on 3 November 2026, Drax Group published its official press release announcing a strategic partnership with Brookfield Asset Management and Radiant, Brookfield's UK-headquartered AI infrastructure platform, to jointly develop, deliver and operate gigawatt-scale data centre capacity at Drax Power Station in North Yorkshire. The release said Brookfield estimates that total investment in the development could potentially exceed £20 billion, including an initial £2.5 billion as part of the first phase. By that afternoon in London, Drax shares had risen around 8.5% — the best performance on the FTSE 250, according to AJ Bell figures reported by DC Market Insights — as investors welcomed the deal.
The second story unfolded hours later, as American voters went to the polls for the midterm elections. As reported the morning after, voters in 10 states faced 41 data-centre measures on the November 3 ballot, according to a Ballotpedia count published on election day. Stateline reported that by Wednesday, 33 local ballot measures restricting data centres had passed or were leading in counts still under way. In Texas, the state with the most planned AI data-centre capacity, Democrat Gina Hinojosa defeated Governor Greg Abbott after campaigning on a moratorium on new data-centre construction. In Georgia, Keisha Lance Bottoms won the governor's race after calling for a pause on new data centres. Investors noticed immediately: Nasdaq-100 futures fell 0.8% overnight, and shares in the AI cloud company CoreWeave dropped 7% and Oracle 4% in overnight trading, CNBC reported.
2. Who Is Building What: Drax, Brookfield, Radiant and Qai
Four names sit on the Drax deal — Drax Group, Brookfield Asset Management, Radiant, and Qai — and each brings a different piece of the puzzle.
As tfTC reported, Drax Group, Brookfield Asset Management, and Radiant signed the agreement on 3 November to jointly develop, deliver and operate a greater-than-1 GW AI data center on the Drax Power Station estate in North Yorkshire, England. Drax's contribution is the physical platform: the land (about 1,000 acres, per Construction Briefing figures reported by DC Market Insights), the power (2.6 GW of dispatchable generation), the cooling systems, and — the asset everyone agrees matters most — the pre-existing 4 GW grid connection. Drax brings the site; that is its whole side of the bargain.
The money and the operating muscle come through Brookfield and its partners. Brookfield Asset Management is one of the world's largest alternative asset managers, and its head of AI infrastructure, Sikander Rashid, called the project a major opportunity for the firm's AI infrastructure strategy — "a critical piece of sovereign AI." Radiant, Brookfield's UK-headquartered AI infrastructure platform, will lead design, customer engagement and commercialisation while Drax handles land, power integration and planning approvals; Radiant chief executive Mahdi Yahya said building and operating from the UK would give British researchers, startups and enterprises "access to frontier compute at home." Funding flows through the Brookfield-Qai AI infrastructure joint venture, formed in December 2025 with Qai, a technology company headquartered in Qatar. The deal is significant precisely because it puts sovereign-backed capital, a large dispatchable power plant and a 4 GW grid connection behind a single campus — a combination tfTC called the first of its kind in the UK.
3. The £20 Billion Number, Properly Read
The £20 billion headline is real — but it is Brookfield's estimate of potential investment if the whole campus is built, not a cheque that has already been written.
This distinction matters more than any other single fact in this article, because the headline number has been widely repeated without its conditions. DC Market Insights laid out the terms carefully: Brookfield puts total potential investment, including chips, at more than £20 billion (about US$26.8 billion, though some outlets round to US$27 billion), and that is an estimate of what could potentially be invested if the campus is constructed in full — a lifetime ceiling figure, as tfTC put it, not a check already written. The only money with anything like a commitment behind it is the first phase: up to 100 MW of data centre capacity, targeted for commercial operation from Q1 2028, backed by an initial £2.5 billion.
Read per megawatt, the phases tell their own story. Phase 1 at £2.5 billion for up to 100 MW prices at £25 million per MW — about US$33.5 million per MW, per DC Market Insights' calculation. The full campus, at more than £20 billion for more than 1 GW, is at most £20 million per MW, because the MW figure is a floor. Nothing in the agreement is a final investment decision: progression of phase one depends on planning consent, regulatory and network approvals, commercial agreements, and — the condition every analyst flagged — securing a suitable tenant. No anchor tenant has been named.
| Item | Detail | Source |
|---|
| Parties | Drax Group; Brookfield Asset Management; Radiant (Brookfield's UK AI platform); funding via the Brookfield-Qai joint venture | Drax press release; AJ Bell |
| Site | Drax Power Station estate near Selby, North Yorkshire; ~1,000 acres; 4 GW grid access; 2.6 GW dispatchable generation | Drax; Construction Briefing |
| Total potential investment | More than £20bn (~US$26.8bn), including semiconductor equipment — an estimate, not a committed price | Brookfield estimate, via AJ Bell and Resultsense |
| Phase 1 | Up to 100 MW; £2.5bn; targeting commercial operation from Q1 2028 | Drax; AJ Bell |
| Phase 2 | Ambition to ramp to 500 MW of capacity between 2028 and 2031 | Drax |
| Phase 3 | Additional capacity after 2031; campus above 1 GW in total | Drax |
| Conditions for phase 1 | Planning consent, regulatory and network approvals, an acceptable tenant, commercial agreements, a final investment decision by each party | Drax |
| Next step | Public consultation on the phase 1 planning application expected to start in November 2026 | AJ Bell; securities.io via tfTC |
| Community | Voluntary community benefits fund, separate from and additional to planning obligations | Drax |
Drax Group chief executive Will Gardiner framed the ambition plainly: "If constructed in full, this project could become one of the largest data centres in the UK," he said, adding that the power station is one of the few UK sites that can provide large-scale reliable power directly to data centre customers by using existing and underused generation capacity. Supportive words, though, are not planning permission — and the consultation that begins this month will be the first real test of whether the local community agrees.
4. Why the Power Station Matters More Than the Money
In the AI infrastructure race, the binding constraint is no longer chips or capital — it is grid access, and Drax already has 4 GW of it.
This is the insight that makes the Drax deal make sense, and tfTC stated it most bluntly: grid access, not compute hardware, is the binding constraint, and institutional capital just proved it. UK data centre developers routinely wait years in the National Grid interconnection queue. DC Market Insights reports that in parts of Britain, a new grid connection can take up to 10 years. Ofgem has said that data centres account for at least 73 GW of grid connection requests in Great Britain. Against that queue, a site that already has 4 GW of connection and 2.6 GW of dispatchable generation on site is scarce — and the price per megawatt reflects a ready-to-use asset, not just a building.
The UK government has now tried to formalise the advantage. The Budget on 28 October 2026 introduced a new Data Centre Covenant, offering grid priority to "strategic" sites, and ASPlatX reported that under it, strategically important data centres that meet tests on local investment, jobs and water use are promised faster access to power and planning — and may even build onsite generation while they wait for a connection. Drax and its partners said their initial assessment is that the project aligns with the Covenant's criteria on energy security, community benefits, environmental standards, reindustrialisation and sovereign infrastructure. DC Market Insights added a useful scale comparison: the UK government says the country needs 6 GW of AI capacity by 2030, and Drax's phases 1 and 2 alone (600 MW by 2031) would equal 10% of that target. A campus above 1 GW would roughly match all 1,025 MW of UK data centre capacity that DC Market Insights currently tracks as live, across 191 facilities. That is the scale of what is being proposed — and it explains why the queue-jumping premium exists.
5. The Biomass Question: Drax's Fuel Is Part of the Deal
There is no getting around it: the electricity at the heart of this AI campus comes from burning wood, and that is the deal's most contested fact.
Drax Power Station is, by the company's own description, the UK's largest single source of renewable electricity by output, and it runs mainly on biomass — burning wood pellets along with smaller amounts of other plant material, as ASPlatX summarised. The plant runs four biomass units of 645 MW each, 2.6 GW in total, Argus reported via DC Market Insights. Under UK rules, those wood pellets count as renewable. But the subsidy story underneath is complicated. The plant's current subsidy structure runs through March 2031, with rates halved from 2027, per tfTC — and the current arrangement ends in March 2027 before a new low-carbon dispatchable Contract for Difference (CfD), signed with the government's Low Carbon Contracts Company in November 2025, runs from April 2027 to March 2031 at a strike price of £109.90/MWh in 2012 prices for about 6 TWh of power a year, as Drax disclosed and DC Market Insights reported.
This is where the deal becomes an exit strategy. tfTC reported that JPMorgan analysts saw the announcement raising expectations that the plant keeps running after its subsidy scheme ends in March 2031, and noted that Drax has received more than £7 billion in UK taxpayer subsidies since 2012 to burn biomass for electricity, according to multiple published reports (tfTC flagged that the precise figure should be confirmed against Drax filings or government data). The data centre partnership is, in tfTC's reading, designed to replace government contract revenue with hyperscaler rents. tfTC also reported, citing securities.io, that a mechanism in Drax's CfD agreement allows the company to request that up to 500 MW be directed to power the data centre during the contract period — tying the energy pivot directly into an existing government arrangement.
The fuel question also carries reputational risk. DC Market Insights noted that campaign groups have opposed using Drax's wood-burning plant to power data centres, and a hyperscale tenant with strict 24/7 carbon targets will look closely at both the sourcing and the scrutiny. Drax says the plant's CfD requires sustainably sourced biomass with enhanced supply chain auditing. For now, the fuel question sits alongside the missing tenant as one of the two open items that will decide whether this project looks visionary or vulnerable in two years' time.
6. £25 Million per Megawatt: What an AI Campus Actually Costs
An AI data centre is priced less like a building and more like a semiconductor factory — and the Drax numbers prove it.
DC Market Insights put three benchmarks next to Drax's phase 1 price of £25 million per MW (about US$33.5 million), and each one teaches something different. First, the cost of just building the shell and plant: Turner & Townsend's 2026-27 construction index puts the average cost to build a data centre in Europe at US$12.76 million per MW — Drax phase 1 is about 2.6 times that. The gap only makes sense if GPUs and servers are inside the £2.5 billion, which is exactly what Brookfield's "including semiconductor equipment" language suggests. Second, a recent European AI deal with a disclosed cost split: VOLT's Rotterdam AI gigafactory, with Nebius lined up for up to 200 MW, needs about €7 billion, of which €2.5 billion goes to buildings and infrastructure and €4.5 billion to chips and equipment, Techzine reported on 12 October via DC Market Insights. That is €35 million per MW all in, with 64% of the total going to equipment. Third, DC Market Insights' own AI data centre model, which puts the full cost of a large AI facility including IT at about US$44.6 million per MW. Drax phase 1, at about US$33.5 million per MW, sits at roughly 75% of that model — their calculation, not a guess.
The pattern is consistent: in an AI data centre, the chips cost more than the building. That matters because it explains where the industry's real money is going — and it is the bridge between this story and the semiconductor business that drives GPU pricing for ordinary buyers too (see our linked price guide in section 18). The same AI accelerators that sit in a £25-million-per-megawatt campus in Yorkshire are priced off the same global supply of advanced chips that gamers and creators buy. When Brookfield says £20 billion "including semiconductor equipment," it is acknowledging that the silicon inside the halls may cost more than the halls themselves.
| Cost component | What it covers | Benchmark from a disclosed deal (source) |
|---|
| Semiconductor equipment (GPUs, servers, networking) | AI accelerators, servers, high-bandwidth networking, storage | ~64% of total: €4.5bn of €7bn at VOLT Rotterdam (Techzine, via DC Market Insights); DC Market Insights' model implies ~70% at US$44.6m/MW all-in vs US$12.76m/MW build-only |
| Building shell and electrical plant | Structures, substations, transformers, switchgear, cooling systems, shared site works | ~36% of total: €2.5bn of €7bn at VOLT Rotterdam; Europe average build cost US$12.76m/MW (Turner & Townsend 2026-27, via DC Market Insights) |
| Power access and land | Grid connection, generation capacity, the land itself — the "ready-to-use" premium | Drax's 4 GW connection valued at roughly 4x all UK live data centre capacity (1,025 MW) in DC Market Insights data; brownfield power sites command a premium over greenfield land |
| First-phase overhead | Substations, cooling plant and shared works that later phases reuse | Drax phase 1 carries 12.5% of the investment for ~10% of capacity — normal for a first phase (DC Market Insights) |
7. The $1.8 Trillion Pipeline: How Big the Global Boom Really Is
The Drax deal is a rounding error next to the global pipeline: Reuters Breakingviews reported that the AI datacentre pipeline tracked by Cushman & Wakefield now totals US$1.8 trillion.
To grasp the scale of what the AI industry is trying to build, start with the United States, where the numbers are the most extreme. Stateline reported, citing the US Data Center Map website, that data centers across the United States currently use about 49 gigawatts of electricity — and that new projects planned or under construction total nearly 350 gigawatts, roughly a quarter of the country's current electrical generation capacity. Morgan Stanley estimates a 32 GW US data center power deficit through 2028, tfTC reported, warning that electricity constraints could favour deep-pocketed tech giants and squeeze smaller players. On the demand side, the commitments are staggering: OfficeChai noted that Anthropic alone has disclosed US$895 billion in compute commitments.
The labs themselves — OpenAI, Anthropic, Google, Meta and xAI — all need vast new data centres to train and run their models, as madrobot noted, and much of that capacity is planned in exactly the states that voted this week. CNBC's reporting adds that Texas is projected to overtake Virginia as the world's largest datacentre market by 2030. The $1.8 trillion pipeline is not a prediction of what will definitely be built; it is a tracker of what is planned or in motion. But it frames the central tension of this article: the capital is committed, the demand is real, and the physical permission to build — the grid connections, the zoning approvals, the community consent — is becoming the scarcest commodity in the whole system.
8. What American Voters Did on November 3
Voters in 10 states faced 41 data-centre measures on the November 3 ballot — and the restrictive ones won overwhelmingly.
The numbers come from a Ballotpedia count published on election day and Stateline's reporting the day after, both compiled by madrobot. Ohio had 21 of the 41 measures — more than any other state — followed by Illinois with seven, Maine with four and Kansas with three. A "yes" vote on 28 of the 41 would limit data centres, through bans, moratoriums, zoning rules or public votes on big projects; the other 13 would help them get built. Stateline, citing the tracker Data Centers on the Ballot, reported that by Wednesday, 33 local measures restricting data centres had passed or were leading in counts still under way. Only one measure — in North Dakota — went the way of the industry's supporters. Most of the week's results were still unofficial at the time of reporting.
This was not a sudden development. Ballotpedia found that every data-centre measure decided earlier in 2026 had also gone for tighter limits, including a ban in Monterey Park, California, that won 88.3% of the vote. Fortune noted that local officials had proposed more than 120 moratoriums across 38 states by June, per OfficeChai's summary, and Fast Company reported that datacenter measures were on ballots in 13 states by election week. The direction of travel has been consistent all year.
| Measure | Figure | Source |
|---|
| Data-centre measures on the Nov 3 ballot | 41 measures in 10 states | Ballotpedia (election-day count), via madrobot |
| States with the most measures | Ohio 21, Illinois 7, Maine 4, Kansas 3 | Ballotpedia, via madrobot |
| Measures that would limit data centres | 28 of 41 (bans, moratoriums, zoning rules, public votes) | Ballotpedia, via madrobot |
| Measures that would help build them | 13 of 41 | Ballotpedia, via madrobot |
| Restrictive measures that passed or were leading | 33 | Stateline, citing Data Centers on the Ballot, via madrobot |
| Measures that went the industry's way | 1 (North Dakota) | Stateline, via madrobot |
| Earlier 2026 measures | All went for tighter limits, including Monterey Park, CA (88.3%) | Ballotpedia, via madrobot |
| Moratoriums proposed nationwide (by June 2026) | More than 120 across 38 states | Fortune, via OfficeChai |
9. Ohio: The Epicenter of the Revolt
No state carried more of the fight than Ohio — 21 ballot measures, the most in the country, and a clean sweep against the industry around Columbus.
Around Columbus, voters went against data centres in Ashville, Pataskala, Grove City, Sunbury and Granville, according to unofficial counts reported by NBC4 and compiled by madrobot. The mechanics varied by town: Pataskala is set to change its charter to ban data centres with a monthly demand of 25 megawatts or more; Pataskala, Sunbury and Granville all amended their charters to ban datacenters of 25 megawatts or more, per OfficeChai's summary; Grove City passed a referendum requirement; and in Ashville, voters appear to have overturned the village's agreement with the data-centre developer EdgeConneX — though construction there is already under way and the vote will not stop it, madrobot noted. In Urbana, Ohio, a datacenter ban passed with about 77% of the vote.
Stateline's pre-election reporting, republished by Planetizen, explained why Ohio became the epicenter: the state agreed to sales tax exemptions for data centers in the 2010s, and now has 154 operational facilities and 84 more in the works — more than all but three other states. "For so many years, Ohio rolled out the red carpet to become a tech hub," said Nick Owens, an attorney with Conserve Ohio. "Unfortunately, what we've become is a tech dump." Notably, Owens is a lifelong Republican and 2024 Trump delegate — Stateline's analysts concluded the opposition has come from Republican and Democratic voters alike, a rare genuinely cross-party movement.
10. Texas: The Governor Who Won on a Moratorium
In the state with the most planned AI data-centre capacity in America, the candidate who promised a moratorium on new construction just became governor.
Democrat Gina Hinojosa defeated Governor Greg Abbott, with the Associated Press calling the race at 11:17pm Eastern time on Tuesday, Yahoo Finance reported. With about 93.5% of the expected vote counted, she had 52.8% to Abbott's 45.7%, according to NBC News, and Abbott conceded. Hinojosa campaigned on a moratorium on new data centres, stricter rules on their water and power use, and an end to tax breaks for the industry. Abbott, who once said he wanted Texas to be the "epicenter of AI development," had announced a moratorium of his own before the election — which Hinojosa dismissed as an "election year pause." (Outlets differ slightly on the exact timing of Abbott's announcement; Stateline's pre-election reporting placed it in the month before the vote. All sources agree it came before election day.)
The Texas result reverberated because of the state's sheer weight in the AI build-out. CNBC's reporting, summarised by madrobot, notes that Texas is projected to overtake Virginia as the world's largest datacentre market by 2030, and Wells Fargo puts 27% of the gigawatts of planned US data-centre capacity in Texas alone. The state also flipped a Senate seat: Democrat James Talarico defeated Republican Ken Paxton, OfficeChai reported — Paxton was the Texas candidate that Elon Musk spent roughly $212 million backing, according to OfficeChai's tally. In the same election cycle, Democrats also flipped Senate seats in North Carolina and Maine, where Troy Jackson defeated Susan Collins, and Democrats seized control of both chambers of Congress. Data centres were far from the only factor in those races — the Democratic wave also reflected broader discontent with the Trump administration, which has aggressively pushed the AI buildout, OfficeChai noted.
11. Georgia and Ohio: Two More Pauses at the Top
Texas was not alone — Georgia elected a governor who called for a statewide pause, and Ohio elected one who promised tougher conditions on every new project.
In Georgia, Keisha Lance Bottoms defeated Republican Rick Jackson by 54.2% to 45.8%, with 98.2% of the expected vote counted, on NBC News's tally. She made a statewide pause part of her pitch on energy bills, saying in October, as quoted by WTOC and relayed by madrobot: "Georgia is currently No. 1 in the country for data centers. We need to pause. We need consumers like you. We need experts in this field to sit around the table and give us guidance as to how the state of Georgia should move forward." Wells Fargo's figures put 7% of planned US data-centre capacity in Georgia.
In Ohio, Democrat Amy Acton beat Vivek Ramaswamy by 52% to 44.7%, according to NBC News. Both candidates had promised tougher conditions on new data centres, the Ohio Capital Journal reported via madrobot — which means that in Ohio, at least, the industry lost no matter which candidate won. The Ohio governor's race mattered for an additional reason: the governor appoints the five members of the Public Utilities Commission of Ohio, which sets electricity rates and the terms under which data centres connect to the grid. With both major candidates committed to tougher conditions, the regulatory direction is set regardless of party.
12. The Polls Behind the Backlash
The election results did not come out of nowhere — every major poll taken before November 3 showed the same deep public scepticism.
The most striking figure came from the University of Massachusetts Amherst. In an October poll directed by political science professor Tatishe Nteta, only 11% of Americans said they would support the construction of an AI data center in their local community, while 65% opposed it. "The people are speaking, and they're speaking in one voice in opposition to this infrastructure," Nteta told Stateline. "It is not something that people desire in any shape, form or fashion." A Washington Post poll, cited in a Fast Company report summarised by OfficeChai, found that two-thirds of Americans oppose building new AI datacenters, and the same report noted an October poll finding that 72% of Americans do not want a data centre near them.
On election day itself, the polls confirmed that the issue moved votes. An NBC exit poll found that over 50% of voters said AI and data centres were a "crucial or major factor" in their vote, and CNBC's voter poll put the figure at 56%. These are extraordinary numbers for an infrastructure issue — they mean that for a majority of voters, a question about where servers get built ranked alongside the economy, healthcare and the usual top-tier concerns in deciding their ballot. The CNBC figure is particularly telling because it came from a business-news outlet's own polling, not from an advocacy group.
13. Your Electricity Bill: Why Household Power Costs Became the Deciding Issue
Strip away everything else, and the voter revolt comes down to one fear: that AI data centers will make ordinary people's electricity bills go up.
It is the thread that connects the ballot measures in Ohio villages to the governor's races in Texas and Georgia. Data centers are among the largest single electricity consumers ever connected to the grid — a single hyperscale facility can draw as much power as a small city — and voters have concluded, in large numbers, that they will end up paying for the grid upgrades, the new generation and the strained supply that follows. Some forecasts show the rapid buildout straining the electrical grid and driving up energy costs, Stateline reported. Whether those forecasts prove right in every market is almost beside the political point: the fear itself moved the votes.
The polling backs the centrality of the bill. When an NBC exit poll finds over half of voters calling AI and data centres a crucial or major factor in their vote, and CNBC's own polling agrees at 56%, the simplest reading is that voters experienced the issue through their monthly bills and their local news about grid strain — not through abstract debates about artificial intelligence. The electricity bill is the one place where a data centre two towns away becomes a personal financial fact. Every candidate who won on this issue, from Hinojosa in Texas to Bottoms in Georgia to Acton in Ohio, made some version of the household-cost argument. And every industry counter-offer — Amazon Web Services boss Matt Garman warned in October that local moratoriums could cost America the AI race and offered $1 billion to win towns over, madrobot reported — has so far failed to change the basic public calculation: the money flows to the companies, the risk flows to the ratepayers.
14. Who Pays for the Power: The Money Fight
The policy question underneath the revolt is brutally simple — should data centres pay the full cost of the electricity and grid infrastructure they use? — and both parties are now competing to say yes.
In Ohio, the issue became the centerpiece of a Senate campaign. Sherrod Brown, the Democratic former senator challenging Republican incumbent Jon Husted, made data centres a central theme of his campaign, calling for them to pay the full cost of their power so household bills do not rise, and attacking Husted as the "face of data centers," OfficeChai reported. The National Republican Senatorial Committee warned in a memo that data centers were "the anchor hanging around Husted's neck," the Washington Post reported via Planetizen. Husted, who had championed tax breaks for data centers during his time as lieutenant governor, responded by drafting a bill that would require state utility regulators to consider making data centers pay to cover the costs of their own electrical infrastructure needs. Senate Democrats blocked the bill last month, calling it "toothless" political messaging, Planetizen reported. Brown won the race — and with it, the argument that the industry must pay its own way gained its most prominent national advocate.
The federal response has been more theatrical than legislative. President Trump has championed the AI buildout, but he has also told AI companies they must "make their own electricity" and pay communities a "dividend," madrobot reported — an attempt to square support for the industry with the voter anger. Back in September, Trump called the pushback against data centres "all a hoax" and said data centres "make people wealthy," a line that sat awkwardly next to his own party's warnings about the issue. Meanwhile, a federal moratorium bill proposed by Bernie Sanders and Alexandria Ocasio-Cortez remains a long shot even with Democrats holding both chambers of Congress, OfficeChai noted. The real action, for now, is at the state and local level, where governors, utility commissions and zoning boards hold the actual levers.
15. Wall Street Reads the Results
The market's verdict came in two acts: relief that the AI trade was still alive on election day, then a sharp overnight repricing once the data-centre results sank in.
Act one played out on Tuesday itself. As fxstreet's market note recorded, technology stocks found their footing on election day after days of being punished over AI spending and financing questions: the Nasdaq gained 0.8%, the S&P rose 0.4% (closing at 7,568), semiconductors jumped 2.5% and memory names rose 2.8%. Then came the earnings: AMD jumped about 9% in after-hours trading as strong data-center sales and an upbeat forecast gave investors fresh evidence that AI demand remains alive and well, while Arista Networks (ANET) rallied 8% after revenue surged nearly 59% and its outlook topped expectations. The whiplash, fxstreet wrote, was enough to make your head spin — one day investors loved the AI trade, the next day they did not, and the day after they could not get enough of it.
Act two came overnight, as the election results — and specifically the data-centre results — were digested. Nasdaq-100 futures fell 0.8% overnight, madrobot reported, and in overnight trading CoreWeave dropped about 7%, Nebius 4% and Oracle 4%, according to CNBC's reporting summarised by OfficeChai — on fears that state-level restrictions could slow buildouts in key markets. fxstreet noted that futures turned lower as the results came in, though the selling eased somewhat overnight with futures ending mixed. The Nasdaq-100 had already slipped into correction territory at the end of October amid concerns about slowing growth in datacenter spending, OfficeChai added. Apple, meanwhile, fell 1.9% on a weak margin forecast that the company blamed on memory-chip costs — a reminder that the AI boom's supply chain squeezes even its biggest winners.
16. Elon Musk and the Space Escape Hatch
Elon Musk's response to the election summed up the industry's mood in a single sentence: "Data centers in space looking pretty appealing right now."
As OfficeChai reported, Musk stayed quiet through most of election night — even as Ken Paxton, the Texas Senate candidate he spent roughly $212 million backing (OfficeChai notes that tallies of his total midterm spending vary by outlet, from roughly $150 million to more than $200 million), lost to Democrat James Talarico. When the outcome was largely clear, Musk posted the space line on X on November 4. It was a joke, but it was also a thesis: if building on Earth means negotiating with thousands of zoning boards, utility commissions and angry voters, why not build where none of them exist?
The thesis is not entirely new. SpaceX has asked regulators for permission to launch up to a million satellites as orbital datacenters, OfficeChai reported — an idea that skeptics, including OpenAI's Sam Altman, have called ridiculous for now. The physics are genuinely daunting: launching compute into orbit, powering it, cooling it in vacuum and beaming results back down is enormously expensive compared with a ground-based facility, and no one has demonstrated it at anything approaching hyperscale. But the joke landed because it named the real constraint. The industry's problem is no longer money or chips — it is permission. Space, as OfficeChai drily noted, may be the one place without a local zoning board.
17. Why the Debate Is Genuinely Two-Sided
It would be easy to frame this as a simple story of greedy companies versus angry citizens — but the honest version is that both sides have real arguments, and both have real weaknesses.
The case for building is not trivial. The UK government argues, in Drax's press release, that countries able to host and power data centres will be better placed to protect their sovereignty and capture the benefits of the digital economy. Emma Reynolds, the Chief Secretary to the Treasury, said the Drax investment would help create good jobs in Yorkshire and deliver growth across the country. In the United States, the industry points to construction jobs, increased property tax revenue in communities desperate for money, and the national-competitiveness argument: artificial intelligence is key to the country's economic future, and the compute has to live somewhere. Josh Vlasto, co-leader of the tech-backed Super PAC Leading the Future, told Stateline that voters want the jobs and the revenue but want the risks mitigated.
The case against is equally substantive, and it just won an election. Voters fear higher electricity bills, strained grids, water consumption, noise, pollution, the loss of farmland, and closed-door deals cut with non-disclosure agreements. The UMass poll (65% oppose, 11% support) and the Washington Post poll (two-thirds oppose) are not the numbers of a fringe movement. And the industry's own behaviour has fed the fire: Stateline's reporting documented private "behind-the-meter" gas plants, tax-break packages negotiated in secret, and a Super PAC whose associated operations were linked by investigations to AI bots posing as journalists and dark-money payments to influencers. When the other side's argument is "trust us," and its recent record includes bot-written news sites, trust is in short supply.
| Issue | Supporters say | Opponents say |
|---|
| Jobs and growth | Hundreds of jobs protected and created; reindustrialisation; sovereign AI capability (Drax; Emma Reynolds; CBI) | Most jobs are temporary construction work; permanent headcount is small for the scale of the site (critics cited by Stateline) |
| Electricity bills | New large customers can fund grid investment; AWS offered $1bn to win towns over (madrobot) | Households will pay for grid upgrades the centres require; centres must pay the full cost of their power (Sherrod Brown; Jon Husted's bill) |
| Environment | Drax's biomass counts as renewable under UK rules, with enhanced supply-chain auditing (Drax) | Wood burning faces climate scrutiny; water use, noise, farmland loss and "behind-the-meter" gas plants (Stateline; ASPlatX) |
| Local control | Strategic projects need faster planning; the UK's Data Centre Covenant accelerates time to power (Drax; ASPlatX) | Deals are cut behind closed doors with NDAs; voters demanded public votes and won them (Stateline; Ballotpedia results) |
| National competitiveness | AI is key to economic competitiveness; the compute must live somewhere (Leading the Future; Trump administration) | Competitiveness does not require any specific town to host a gigawatt campus; 120+ moratoriums proposed in 38 states (Fortune) |
18. The Chip Engine Underneath It All
None of this — not the £20 billion, not the $1.8 trillion pipeline, not the election — makes sense without the silicon at the bottom of the stack.
Every number in this story traces back to AI accelerators. Brookfield's £20 billion estimate includes semiconductor equipment because the chips are the single largest cost in an AI data centre — around 64% of the total in the one European deal with a disclosed split, and roughly 70% implied by DC Market Insights' cost model. AMD's 9% after-hours jump on Tuesday came specifically on strong data-center sales and an upbeat forecast, fxstreet reported — the market's clearest signal that AI chip demand is still accelerating. And Apple's 1.9% fall on a weak margin forecast blamed on memory-chip costs shows the other side of the same coin: the AI boom is bidding up the price of advanced silicon for everyone, including companies that are not building data centres at all.
This is where the data-centre story meets the reader's own wallet. The same GPU market that prices AI accelerators for hyperscalers sets the prices that gamers, creators and PC builders pay — our October 2026 graphics card price guide tracks how AI demand has reshaped what a graphics card costs, and the dynamic has only intensified as the build-out accelerated. On the processor side, the same competition for advanced chip supply is playing out between the CPU giants — our Intel Nova Lake vs AMD Zen 6 comparison covers the next generation of chips that will power both AI servers and ordinary desktops. The data centre and the desktop are no longer separate markets; they are two ends of the same silicon supply chain.
19. What Comes Next: Five Dates That Decide Everything
The story is not over — it is barely past the opening announcements. Here are the five milestones that will decide what actually gets built.
November 2026 — the Drax public consultation. The partners expect the phase 1 planning consultation to begin this month, with the formal planning process to follow. It is the first real test of local consent in Yorkshire — arriving just as American voters showed what happens when communities feel unheard.
8 January 2027 — the UK's National Policy Statement for Data Centres. Consultation on the draft policy statement closes on this date, setting how large projects are judged in planning. Combined with the Data Centre Covenant, it will define the rules for every UK project that follows Drax.
1 April 2027 — Drax's new CfD begins. The low-carbon dispatchable Contract for Difference starts, running to March 2031 at £109.90/MWh in 2012 prices for about 6 TWh a year, per Drax's disclosures. How it interacts with the data centre's power demand — including the reported mechanism allowing up to 500 MW to be directed to the campus — will shape the project's economics.
Before Q1 2028 — the anchor tenant. Phase 1 needs a named tenant, network approvals and final investment decisions to hit its Q1 2028 target. tfTC called the "anchor tenant required before construction begins" clause the most important sentence in the announcement. Who signs, and when, determines whether this becomes the UK's largest data centre or an announcement that aged poorly.
2027 and beyond — the American regulatory response. The new governors take office with pause promises to keep, utility commissions face pressure to make data centres pay full infrastructure costs, and the Sanders-Ocasio-Cortez federal moratorium bill — a long shot even with Democratic control of both chambers, per OfficeChai — will test how far the backlash travels. The 45 projects already disrupted in the second half of 2026, per the Data Center Watch tracker, are only the beginning.
20. Final Verdict
The AI data center boom has the money, the chips and the demand — but it no longer has the automatic right to build, and that changes everything.
Take the two stories of 3 November together and the verdict writes itself. On one side: the largest AI infrastructure bet ever announced in Britain, backed by sovereign capital, sited on one of the few plots of land in Europe with gigawatts of power already attached, aimed at a global pipeline worth $1.8 trillion. On the other: 33 local measures passed or leading against data centres, three governors elected on pause-or-moratorium platforms, polls showing two-thirds of Americans opposed to new AI data centers, and markets repricing the whole trade overnight. Neither side is bluffing. The capital is real — £2.5 billion is committed to Drax phase 1, subject to approvals and a tenant. The votes are real too — 52.8% in Texas, 54.2% in Georgia, 52% in Ohio, and 88.3% for a ban in Monterey Park.
The honest conclusion is that both sides are half right, and the next phase belongs to whoever closes the gap. The industry is right that AI compute is strategic infrastructure that has to live somewhere. The voters are right that the current terms — tax breaks, secret deals, socialised grid costs, rising bills — are unacceptable, and they have proven, from Ohio villages to the Texas governor's mansion, that opposing data centres is a winning position. The projects that get built will be the ones that rewrite the bargain: full-cost power pricing, transparent negotiations, genuine community benefits, and power sources that survive scrutiny. The permission is the product now. Whoever earns it, builds. Whoever does not, does not.
Frequently Asked Questions
Is the £20 billion Drax figure a real committed investment?
No — it is Brookfield's estimate of potential investment if the full campus is built. The only figure with a commitment behind it is the £2.5 billion for phase 1 (up to 100 MW, targeting Q1 2028), and even that is conditional on planning consent, regulatory approvals, and a named anchor tenant. None has been named. Treat £20 billion as a ceiling, not a cheque.
What is Drax Power Station, and why does its grid connection matter?
Drax is the UK's largest single source of renewable electricity by output, running mainly on biomass, with 2.6 GW of dispatchable generation and a pre-existing 4 GW grid connection. That connection is the real asset: in parts of Britain, a new grid connection can take up to 10 years, and Ofgem says data centres account for at least 73 GW of connection requests. A site that already has the connection skips the single biggest bottleneck in European AI infrastructure, which is why Brookfield is willing to pay a premium for it.
When will the Drax AI data centre actually open?
Phase 1 targets commercial operation from Q1 2028 — but nothing is guaranteed. Before construction can begin, the project needs planning consent, regulatory and network approvals, commercial agreements, a final investment decision by each party, and a signed anchor tenant. The public consultation on the phase 1 planning application is expected to start in November 2026. Phase 2 aims for 500 MW between 2028 and 2031, with more than 1 GW after 2031.
Did American voters really vote against AI data centers in the midterms?
Yes — decisively, at the local level. Ballotpedia counted 41 data-centre measures on the November 3 ballot across 10 states; 28 would limit them and 13 would help build them. Stateline reported that 33 restrictive measures had passed or were leading the day after, with only one — in North Dakota — going the industry's way. Around Columbus, Ohio, voters rejected data centres in Ashville, Pataskala, Grove City, Sunbury and Granville, per NBC4.
What did the new governors of Texas and Georgia actually promise?
Both campaigned on pausing new data-centre construction — not banning it outright. Texas governor-elect Gina Hinojosa (52.8% vs 45.7% over Greg Abbott, per NBC News) promised a moratorium on new construction, stricter water and power rules, and an end to tax breaks for the industry. Georgia governor-elect Keisha Lance Bottoms (54.2% vs 45.8% over Rick Jackson) called for a statewide pause, saying Georgia — which she described as No. 1 in the country for data centers — needs to stop and consult experts. Ohio's Amy Acton (52% vs 44.7%) promised tougher conditions on new projects.
Will data centers make my electricity bill go up?
That is the central fear driving the backlash, and the honest answer is: it depends on who pays for the grid upgrades. Data centres are enormous electricity consumers, and if the transmission and generation upgrades they require are spread across all ratepayers, household bills rise. That is why Sherrod Brown campaigned on making data centres pay the full cost of their power, why Jon Husted drafted a bill to make them cover their own infrastructure costs, and why President Trump said AI companies must "make their own electricity." The fight is about cost allocation, not just construction.
Why do AI data centers cost so much more than normal ones?
Because the chips inside cost more than the building. In the one recent European deal with a disclosed split — VOLT's Rotterdam AI gigafactory — about 64% of the roughly €7 billion total goes to chips and equipment, with only 36% to buildings and infrastructure. DC Market Insights' cost model puts a full AI facility with IT at about US$44.6 million per MW, versus US$12.76 million per MW for construction alone. An AI data centre is priced like a semiconductor factory that happens to have walls.
Who is funding the Drax project?
Capital flows through the Brookfield-Qai AI infrastructure joint venture, formed in December 2025. Brookfield Asset Management brings the capital and operating capability; Qai, a technology company headquartered in Qatar (identified by DC Market Insights as the AI subsidiary of the Qatar Investment Authority), is the co-investor. Radiant, Brookfield's UK-headquartered AI infrastructure platform, will design, build and operate the facility, while Drax contributes the land, power, grid connection and planning work. Note the full £20 billion estimate is larger than the joint venture's reported US$20 billion size — so the campus would need additional capital or debt as it grows, per DC Market Insights.
How did Wall Street react to the election results?
With a sharp overnight repricing of data-centre-exposed stocks. On election day itself, tech rallied — the Nasdaq gained 0.8% and the S&P 0.4%, with AMD jumping about 9% after hours on strong data-center sales. But as the results came in, Nasdaq-100 futures fell 0.8% overnight, and CoreWeave dropped about 7%, Nebius 4% and Oracle 4%, CNBC reported, on fears that state-level restrictions could slow buildouts. The market's takeaway, per fxstreet: delays can shift the timing of AI spending, but they do not eliminate the demand.
What happens next?
Five milestones decide it. The Drax public consultation expected in November 2026; the UK's data-centre policy statement closing 8 January 2027; Drax's new CfD starting 1 April 2027; the signing of an anchor tenant before the Q1 2028 target; and, in the US, how new governors and utility commissions turn pause promises into rules. The federal moratorium bill remains a long shot. The industry's problem is no longer money or chips — it is permission.
Sources
Comments
Post a Comment