Applied Digital just got past one gigawatt of contracted capacity at Polaris Forge 3 — unnamed hyperscaler, undisclosed PPA term, no in-service date. So, yeah, that’s a real number wrapped around three open questions. Welcome to The Data Center Daily. Today we find out whether “high investment-grade hyperscaler” is a counterparty or just a press release wearing a contract’s clothes. Also on deck: IID’s large-load tariff is up for public review in Imperial County, interconnection.fyi’s GridTracker drops its first queue data of the week, and Google’s Missouri fifteen billion finally has a map pin. The gigawatts are announced — now we see what the infrastructure can actually carry. Markets Insider writes:
15-year take-or-pay leases with the same U.S. based high investment-grade hyperscaler previously signed at Delta Forge 1, valued at approximately $7.5 billion in base-term contracted revenue, $18.2 billion if all options are exercised. 300 MW of critical IT load, purpose-built for large-scale AI training and inference workloads.
Applied Digital crossed 1 GW of contracted capacity yesterday, and the Polaris Forge 3 lease gets them to four campuses. This one is 300 MW of critical IT load, with 430 MW of grid-connected utility power behind it, 15-year take-or-pay terms, $7.5 billion in base-term revenue, and $18.2 billion if the options get exercised. All told, they’re saying $31 billion in contracted lease revenue across the four sites. The number is real. The counterparty isn’t named. “U.S. based high investment-grade hyperscaler” — same language they used for Delta Forge 1 — and we still don’t know who it is, when this goes live, or what “contracted” means in construction terms. That’s the announced-versus-signed question sitting right there in the release. A 15-year take-or-pay is a signed instrument — real paper. But Polaris Forge 3 is in “a northern state,” which is doing a lot of work for a 430 MW grid-connected site. And there’s still no in-service date. Four hundred thirty megawatts of grid-connected utility power in a northern state, no named utility, no cited interconnection queue position, and a counterparty they won’t name — I’ll believe the $31 billion when the substation is energized. From ABC17 News:
Gov. Mike Kehoe and leaders from Google unveiled a data center project Wednesday that includes a $15 billion infrastructure investment from the tech giant. Kehoe joined local officials and Google executives at the Laborers and Contractors Training Center in High Hill on Wednesday afternoon. Speakers touted the jobs the project will bring and sought to dismiss concerns about data centers' use of electricity and water resources.
Google’s Missouri announcement now has a campus location — High Hill, Montgomery County — and a named state commitment: $15 billion. That’s a real geographic pin on what had just been a total capex number from Q1 earnings. Ruth Porat’s on stage saying each full-time job creates nine others, and the state’s spokesperson is telling the room it’s air-cooled so water use is just “kitchens and bathrooms.” Before I take any of that seriously, I need to know what the interconnection ask looks like for Ameren Missouri. To be fair, air-cooled is a real design difference — it’s not nothing. But the event was doing the job-multiplier math out loud while the actual load addition to the Missouri grid never made it on stage. Fifteen billion in “infrastructure investment” — is that the data center capex, or does it also include whatever transmission and generation buildout the state agreed to subsidize to land the announcement? Those are very different commitments hiding inside the same number. Here's Fierce Network:
Google is going neocloud. The tech giant just inked a $5 billion deal with private equity firm Blackstone to build a new U.S. based data center business based on its TPU chips, with the goal of bringing 500 MW of capacity online next year. But why, exactly?
The Google-Blackstone N1 JV has been on the board since Tuesday, but the Fierce Network piece finally asks the operational question: why does a company spending up to $190 billion on its own infrastructure need a $5 billion neocloud arm? Five hundred megawatts by next year, TPU-native, but the customer targeting is still described as unclear. Unclear customer targeting on a $5 billion vehicle — come on, that’s not a mystery, that’s a pitch deck. And Jason Andersen at Moor Insights is right to size it: CoreWeave crossed 1 GW and is forecasting four more in three years, so N1’s half-gigawatt looks more like a proof of concept than a competitive entry. The TPU angle is the honest reason this structure exists at all — Google gets manufacturing-scale benefits when N1 buys chips, and N1 theoretically offers lower price points than Nvidia-based competitors. That’s real economic logic. Whether there are real customers lined up is the part the article can’t answer. Blackstone brings the balance sheet and the real estate muscle — they know how to finance a data hall. But “customers TBD” at five billion dollars is the kind of line that would get laughed out of a utility PPA negotiation. Nick Manderlink, writing in Interconnection.fyi:
The past 2 weeks have seen exceptional activity in the queues. 157 new projects totaling 36.92 GW entered the queues against 4.56 GW of withdrawals, largely driven by 25 GW of additions from Southern Company’s 2026 cluster window. The broader renewable versus firm/thermal bifurcation of the queues continues: gas led additions this period (roughly 20 GW added, +50% YoY), while renewable and battery storage queue capacity continues to contract.
Interconnection.fyi dropped its first GridTracker Insights this week — May 2 through 15 — and the headline number is 36.92 gigawatts of new queue additions against only 4.56 gigawatts of withdrawals. Nick Manderlink is co-authoring it, same guy who worked on LBNL’s Queued Up report, so this is a primary source with real queue pedigree. Twenty-five of those 36 gigawatts came from Southern Company’s queue alone in two weeks. That’s one transmission operator’s intake. Queue position is not a shovel, it’s not a permit, it’s not a signed PPA — and Applied Digital just dropped a one-gigawatt contracted-capacity headline the same morning this report lands. Those two numbers are measuring completely different things. And that’s exactly the stack worth flagging — 36 GW queued nationally in a two-week window, while Linn County, Kansas put a twelve-month moratorium on large-load interconnection back on May 11. Commissioner Hamilton’s vote looks a lot less like a local overreaction when the queue intake rate looks like this. Here's Beyond Borders News:
Under the proposed framework, IID would establish clear guidelines governing the evaluation, interconnection, and service requirements for large-load projects. The intent, according to the utility, is to create a transparent and non-discriminatory process while also safeguarding existing ratepayers from bearing costs associated with serving new and potentially high-impact energy users.
Following up on the large-load rules thread from earlier this week — IID is putting a cost-responsibility tariff out for public review. The core mechanism is simple: large-load customers foot the bill for transmission upgrades, system studies, and collateral. Existing ratepayers are explicitly held harmless, at least on paper. I want to see what “collateral requirements” actually means when the ink is dry. That’s the clause that either has teeth or doesn’t — and it’s why every other utility with a large-load tariff has been fighting over that language for two years in commission proceedings. And the timing matters. Monday we were talking about Imperial County’s water stress from the CalMatters piece. IID is in the same basin — and now they’re building the rate-design framework for the same load growth that’s already straining the aquifer side of the ledger. Two proceedings, one geography, neither resolved. The FERC Section 403 rulemaking has a June deadline on the federal side. IID is a state-chartered irrigation district — not FERC-jurisdictional on retail rates. So you’ve got a federal clock ticking and a parallel local process that can move on its own timeline. That’s not coordination, that’s two separate lanes. If The Data Center Daily helps you stay on top of the industry, take a moment to subscribe or leave a review wherever you’re listening. It really helps other people find the show.
We’ve put links to every story from today’s briefing in the show notes, so if one caught your attention, you can dig into the source material there.
That’s The Data Center Daily for this Thursday. Thanks for listening, and we’ll be back with you next time. This is a Lantern Podcast.