September 9, 2026 · 13 min
Google's €13 Billion Finland Bet
About this episode
Google pledges at least thirteen billion euros to a Finnish AI-infrastructure buildout bundled with a nuclear life-extension deal, new wind and battery storage — while Ohio, Iowa and Virginia show what the buildout looks like when that bundling doesn't happen.
- Google's AI commitment to Finland — Google
- Questionable data center forecasts are driving up Ohio power bills — Canary Media
- DOE Closes $1.9-Billion Loan to Restart Duane Arnold Nuclear Plant — POWER Magazine
- Virginia kept its climate law. Dominion sees a loophole. — Canary Media
Source links
- @curious_founder on X
- @mpukita on X
- bloomberg.com
- canarymedia.com
- cnbc.com
- freemalaysiatoday.com
- indianagazette.com
- ohiomfg.com
- techzine.eu
- utilitydive.com
- wdadradio.com
- wdadradio.com
- wdadradio.com
Concrete Compute is an AI-voiced podcast, built and run by a real person. Nothing in this episode is financial advice.
More from Brian Lampert: Quickly Quantum, the daily quantum computing briefing, and Space Stakes, the business of the new space race. Transcripts and every episode: concrete-compute.kngoworld.chatgpt.site.
Listen in your podcast app
In full
Episode transcript
Today on Concrete Compute: Google just pledged thirteen billion euros to Finland — its biggest single investment anywhere in Europe. But does the company's own math on jobs and economic impact survive contact with construction reality? Before that, in the headlines: Ohio ratepayers may be footing the bill for data centers that never get built, the Department of Energy just closed a loan to bring a shuttered Iowa nuclear plant back from the dead — one with a direct line to Google's own power needs — and Virginia's biggest utility may have found a loophole in the state's own climate law. Welcome back to Concrete Compute, your daily brief on the AI infrastructure buildout. It's Wednesday, September 9, 2026. Let's get into it.
Let's start in Ohio, where the math on data centers is starting to look shaky. Canary Media reports the state now hosts roughly one hundred operating data centers, with nearly one hundred fifty more proposed — and the state's largest utility, AEP Ohio, is forecasting demand will nearly triple over the next decade, largely on the back of those projects. Here's the problem, and you'll want to remember this one: not all of those projects are going to get built, and Ohioans are on the hook either way. The Ohio Manufacturers' Association's own testimony to state regulators, also reported by Canary Media, calculates that just two proposed data-center projects in AES Ohio territory could trigger two-point-seven-seven billion dollars in guaranteed transmission spending over forty years — while the data centers themselves only guarantee about one-point-four billion dollars back. That leaves more than one-point-three billion dollars of exposure on everyone else's power bill if the load never shows up. So who eats that difference? 'We have watched projected demand become the justification for enormous spending, record capacity costs, and increasingly urgent claims about grid reliability,' said Brad Belden of the Ohio Manufacturers' Association Energy Group. 'Reliability cannot become the magic word that ends every conversation about cost.' Now, PJM's own Asim Haque isn't buying the skepticism, calling the idea that load growth isn't real 'both naive and irresponsible.' Here's a detail I want to flag: Ohio policy researcher @mpukita noted on X that some of the newest mega-projects aren't even waiting on the grid — they're building dedicated, behind-the-meter gas generation instead, which sidesteps the ratepayer fight but raises a different one about emissions. Either way, this is the same interconnection-forecast fight that keeps showing up state to state, just with Ohio's own numbers attached.
Sticking with power that's actually getting built instead of just forecast: the Department of Energy has closed a loan of up to one-point-nine billion dollars to NextEra Energy to restart the Duane Arnold Energy Center in Iowa, POWER Magazine reports. This plant closed back in 2020 after forty-five years of operation, damaged by a derecho and judged too costly to fix at the time. Now it's getting a second life, targeted to resume commercial operation in the first quarter of 2029, pending Nuclear Regulatory Commission approval. Now, here's the direct line to our main story: NextEra already has a twenty-five-year power purchase agreement to supply this plant's electricity to Alphabet's Google. POWER Magazine reports the restart is projected to support around one thousand five hundred construction jobs and more than four hundred fifty permanent operating jobs. Six hundred fifteen megawatts of baseload power, back from the dead, financed by federal dollars and anchored by a hyperscaler's demand — that's becoming the template you'll keep hearing about. Worth noting: the loan is described as 'up to' one-point-nine billion, meaning actual disbursement depends on the plant hitting restart milestones — this isn't the whole check written on day one.
Now to Virginia, where the buildout is testing a law the state just doubled down on. Canary Media reports — and this one's single-source, so we haven't independently confirmed it yet — that lawmakers this year didn't just preserve the twenty-twenty Virginia Clean Economy Act, they strengthened it, expanding requirements for battery storage, community solar and efficiency. But Dominion Energy, the state's largest utility, is now proposing one of the biggest natural gas plants in the country in rural Cumberland County — three gigawatts of capacity, more than any operating gas plant in America except a four-point-three gigawatt complex in Florida, according to Canary Media. The law says new fossil plants aren't allowed, and existing ones must close within two decades — except regulators can waive that if they decide grid reliability is at stake. Dominion already used that escape clause once, for a smaller plant in Chesterfield County. If it works again here, on a plant three times that size, you have to ask: is Virginia's climate law still enforceable, or is the reliability waiver just how these things actually get built? The company says the plant won't come online before twenty thirty-three at the earliest, and both state permits could stretch into twenty twenty-eight.
Our main story today: the promise-versus-delivery gap in a thirteen-billion-euro pledge. Google says it's putting at least thirteen billion euros — about fifteen-point-one billion dollars — into AI infrastructure in Finland over the next two years, and the real question isn't whether that number is big, it's whether the company's own math on what it delivers survives construction. Google's own blog, corroborated by CNBC and Bloomberg, frames this as the company's largest single investment anywhere in Europe. This isn't Google's first stop in Finland — it's had a data center running in Hamina for fifteen years — but this expansion pushes into three new sites: Kajaani, Muhos and Vaala, timed to a construction window in twenty twenty-seven and twenty twenty-eight. Now, what makes this different from a plain 'we're building a data center' press release is what Google bundled alongside it. According to Google's own announcement, the package includes a twenty-two-year life-extension deal for the Loviisa nuclear plant, new onshore wind, and a ninety-four-megawatt battery — all explicitly framed as protection for local ratepayers, the kind of protection you just heard Ohio and Virginia listeners very much do not have guaranteed elsewhere in this buildout. That's a real point in Google's favor: pairing new load with new generation, rather than just showing up and asking the grid to absorb it. And Google didn't stop at megawatts — it also released its own commissioned estimates of what this means for Finland's economy: thirty-seven thousand jobs and three-point-six billion euros a year in GDP, according to Google's blog. Those are the numbers I want you to sit with for a second, because this is exactly the kind of figure that shows up in every one of these announcements, and it's exactly the kind of figure nobody outside the company has audited yet. Remember Duane Arnold from a few minutes ago? Same company, same demand, different continent — NextEra's restarting six hundred fifteen megawatts in Iowa on a twenty-five-year contract with Google. Finland is the same appetite, wearing a nuclear-plus-wind-plus-battery costume instead of a restarted reactor. Both are Google trying to lock down firm power ahead of a buildout that keeps outrunning the grid's ability to plan for it. One more caveat before we get to the read: Google describes the thirteen billion euros as 'at least' that amount, meaning the final scope could grow — and the reporting doesn't break out how much of that is data-center hardware versus the nuclear, wind and battery investments bundled into the headline figure. That's a distinction with real money attached, and it's one nobody's answered yet.
Here's my read. The energy piece of this deal is the part I'd actually call a genuine win, in plain terms: Google isn't just showing up and asking Finland's grid to absorb a few gigawatts of new load, it's paying to keep an existing nuclear plant running for another twenty-two years, adding wind, and adding storage. That's the beneficiary-pays model I keep wanting every hyperscaler to hit — the company creating the new demand is also paying for the new generation, instead of quietly hoping the transmission bill lands on somebody else's monthly statement. If more of this buildout looked like that, half of Ohio's fight and most of Virginia's loophole debate wouldn't exist. But then there's the thirty-seven thousand jobs and three-point-six billion euros in annual GDP, and this is where I want you to slow down with me, because these are Google's own commissioned projections, released alongside its own investment announcement, not numbers an independent Finnish agency has signed off on. And buried in that same skepticism is a detail worth saying out loud: of those thirty-seven thousand jobs, only around seven thousand are the ongoing, operational kind — running the data centers, maintaining the grid, working at the plant — once the construction phase wraps. The rest are transient build-phase jobs and induced local spending, which is real money while the cranes are up, but it isn't the same as a community getting a permanent, seven-day-a-week payroll. That's not a knock on Google specifically — every hyperscaler's economic-impact study runs this same play, bundling construction and operational jobs into one headline number, and you hear 'thirty-seven thousand jobs' and picture thirty-seven thousand permanent paychecks. My standard here, and I apply it the same way to Amazon, Microsoft, Meta, and everybody else spending tens of billions a year on this buildout: if you can afford the concrete, you can afford to publish the two numbers separately — construction jobs and permanent jobs — instead of making a listener, or a Finnish regulator, do the subtraction themselves. Here's what would make this credible going forward: a signed interconnection filing for the new sites, a published breakdown of construction versus permanent roles from a source other than Google's own communications team, and confirmation that the ninety-four-megawatt battery and new wind actually get built alongside the data centers rather than trailing behind them by a few years. Now, the open question nobody's answered yet: will Finland's own regulators or utilities publish independent verification once construction actually starts, the way Ohio's ratepayer advocates are now forcing utilities to defend their demand forecasts in front of the state's own commission? Until that happens, thirteen billion euros is a plan, not a receipt. And notice the company itself hedges its own number, describing it as 'at least' thirteen billion euros — which tells you even Google isn't entirely sure where the final scope lands. Time for the Hype Check. I'm putting this one at a six out of ten on substance. The nuclear life-extension deal, the wind, and the battery are real, signed commitments that put Google's money where the grid's mouth is — that's the strongest part of this announcement, and it's genuinely better than how a lot of this buildout gets financed elsewhere. But the jobs and GDP numbers are doing a lot of the announcement's emotional lifting, and they're Google's own homework, graded by Google, released the same day as the investment news. Construction hasn't started. The sites in Kajaani, Muhos and Vaala are still lines on a map. Come back to this one in twenty twenty-eight and see how many of those thirty-seven thousand jobs are still on a payroll — that's the number that actually tells you whether this was infrastructure or just a very well-produced blog post.
The next time you see a hyperscaler announcement with a giant jobs number attached, ask the question this episode kept asking: how many of those jobs are still there once the ribbon-cutting photo op is over. If that's the kind of scrutiny you want in your feed, go ahead and follow Concrete Compute wherever you're listening. This has been Concrete Compute, an AI-voiced podcast, created and built by a real human using today's cutting-edge technology. Nothing you heard on this show is financial advice. I'm Brian Lampert, and I'll catch you all tomorrow — take care!
I also host Space Stakes: the business of the new space race, every day. What actually flew, what the contract is really worth, and who has customers. Find it wherever you get your podcasts.