Tuesday was a big day for fusion money and an even bigger day for the reason fusion money exists. Type One Energy announced a $200 million Series B led by Breakthrough Energy Ventures and Clutterbuck Capital, with Siemens Energy's venture arm joining as a new backer — pushing the stellarator developer past $400 million raised to date. The same week, Xcimer Energy locked in up to $30.5 million more from the Department of Energy's Milestone-Based Fusion Development Program, bringing its total federal non-dilutive funding to $43.5 million. And on the exact same day, Google and Constellation Energy announced a 3,590-megawatt power deal built almost entirely on uprating existing nuclear reactors — because Google needs electrons in 2028, not 2036.
That juxtaposition is the whole story. Fusion is finally attracting the kind of capital a civilizational technology deserves. It is also, by every public timeline its own developers have given, a decade or more away from putting a watt on the grid. Meanwhile the grid itself is choking on demand that's arriving now. Both things are true, and a publication obsessed with energy abundance has to hold them separately instead of letting the fusion headlines paper over the interconnection crisis sitting underneath them.
The Fusion Capital Stack Is Finally Real
For years fusion funding was a parlor trick — big headline numbers from a handful of labs, little evidence of a repeatable financing model. That's changed. The DOE's Milestone-Based Fusion Development Program just disbursed a second tranche of $42 million across eight private developers — Commonwealth Fusion, Helion, Tokamak Energy, TAE Technologies, Type One Energy, Realta Fusion, Xcimer Energy, and Focused Energy each pulled $3 million to $8 million. Crucially, the awards are tied to technical gates — first plasma, tritium self-sufficiency, magnet endurance — not calendar dates. DOE Office of Science director Asmeret Asefaw Berhe has framed the program's purpose plainly: derisk the first-of-a-kind cost estimates that make or break private financing of a commercial plant, according to the program's own reporting. Total federal commitment through 2029 is expected to top $1.2 billion, with industry cost-share more than doubling that.
Layer private capital on top of that public scaffolding and the numbers get genuinely large. Dealroom has tracked fusion startups raising a reported $4.0 billion in 2026 — the firm's own tally, which should be read as a market-tracker's aggregate rather than an audited figure, but it's directionally consistent with what's showing up in individual rounds: Helion at a reported $16 billion valuation, Commonwealth Fusion at $6 billion, Pacific Fusion at $5.4 billion. Type One's new $200 million is the third-biggest fusion Series B of the last five years by PitchBook's count, per Reuters. This is what energy maximalism looks like when it finds a willing capital market — serious institutional money, backed by Bill Gates' Breakthrough Energy Ventures and an industrial giant in Siemens, betting on the technology that could eventually make the Kardashev ladder look less like science fiction.
Nobody's Hiding the Timelines, and That Matters
Here's where I want to be the data-driven evangelist rather than the breathless one. Type One's CEO Christofer Mowry told Reuters the company is "not an experimental science project" and is aiming to launch its commercial plant, Infinity Two, within a decade. Xcimer is more specific still: its commercial program is targeting electricity on the grid in the mid-2030s, and this summer DOE approved the preconceptual design milestones for its Athena power-plant architecture after Xcimer brought online Phoenix, described as the world's largest privately owned laser system.
Those are honest, ambitious, and still a long way off. A decade is nothing on a civilizational clock — it's everything on a PJM capacity auction clock. The Milestone Program's gate-based structure is the right design precisely because it refuses to pretend otherwise: you get paid when the magnet holds, not when the calendar says you should be done. That discipline is good news for fusion's credibility. It's also a blunt reminder that fusion cannot be this decade's answer to the thing actually straining the grid right now.
The Thing Actually Straining the Grid Right Now
While fusion developers raise money against 2030s milestones, the interconnection system is drowning in today's demand. Lawrence Berkeley National Laboratory's Queued Up report counted roughly 8,200 active U.S. interconnection projects at the end of 2025, totaling 2,061 gigawatts of proposed generation and storage — more than the 1,312 gigawatts currently installed on the entire U.S. grid. That is not a rounding error. That is a queue bigger than the grid it's trying to join. FERC is attacking the backlog from two directions: Order 2023 pushes transmission providers toward a first-ready, first-served cluster-study process with real financial penalties for non-serious projects, and in June 2026 the commission opened six Section 206 show-cause proceedings forcing grid operators to explain how they'll speed up large-load connections — the data centers and manufacturing plants driving demand growth in the first place, per CoBank's analysis of the filings.
On the demand side, the Department of Energy projects total U.S. electricity demand growing roughly 15-20% over the next decade, driven by AI, data centers, and electrification — a broad DOE planning estimate tied to the 2050 net-zero target. A separate figure from EPRI, cited in the same DOE post, estimates data centers specifically could climb from about 4% of U.S. electricity load in 2023 to as much as 9% by 2030 — according to DOE's own summary of that research. Those are two different agencies measuring two different things with two different methodologies — one a whole-economy demand forecast, the other a sector-specific share-of-load estimate — and conflating them is exactly the kind of sloppy math that gets this publication in trouble. What they agree on is direction: up, and fast, inside a grid whose connection process was built for a slower era.
What's Actually Filling the Gap: Existing Reactors, Not New Physics
This is why the Google-Constellation deal is the more immediately important story of the week, even next to fusion's funding haul. Google contracted for 3,590 megawatts of power from Constellation in PJM — 890 megawatts of new nuclear capacity from uprated reactors under a 20-year power purchase agreement, plus 2,700 megawatts from the existing fleet under a 15-year supply deal, Reuters reported. Constellation is putting more than $4.3 billion into upgrading 11 reactor units across Illinois, Pennsylvania, and New Jersey, with the first uprated plant expected to deliver power by 2028 — per the companies' joint statement, confirmed by Bloomberg. Constellation shares jumped more than 13% on the news, a market telling you plainly which kind of nuclear bet it trusts to cash flow on schedule.
Note the honest accounting here, because it's the kind of distinction this publication insists on: the 3,590 MW figure is a sum of contracted capacity across two separate deals, not a measure of continuous output, and the "additional" 890 MW is future capacity from reactor modernization, not new plants breaking ground — a nuance energynews.pro flagged that's easy to blur in the headline numbers. It's also not the first such deal — Constellation restarted Three Mile Island for Microsoft and Google separately contracted to restart NextEra's Iowa plant, Reuters noted — which tells you the hyperscalers have already concluded that squeezing more out of the existing fission fleet is the only lever that moves fast enough to matter this decade.
Two Clocks, One Mission
None of this is an argument against fusion. It's an argument for running two clocks at once. The fusion capital stack — DOE's gate-based milestones, Type One's $200 million round, Xcimer's laser program, the broader $4 billion Dealroom is tracking across the sector — is exactly the kind of long-horizon bet a civilization serious about energy abundance should be making. A decade from now, if even two or three of these architectures clear their technical gates, the addressable upside dwarfs anything a reactor uprate can deliver. But betting on 2035 doesn't excuse neglecting 2026, and right now the binding constraint on American electricity growth isn't physics — it's a 2,061-gigawatt queue trying to squeeze through a grid built for a smaller country.
Watch three things over the next several months: whether FERC's six Section 206 compliance filings, due from PJM, MISO, SPP, and the other named grid operators, actually shorten large-load connection timelines or just produce more paperwork; whether Xcimer's Athena design clears its next DOE milestone gate on schedule; and whether Constellation's first uprated reactor hits that 2028 delivery date it just promised Google. Fusion will get there. The grid doesn't have until fusion gets there.
