From Record Shots to Real Megawatts
A sneak peek of the interior of JET fusion lab’s nuclear fusion reactor in Oxford, UK.
Credit: UKAEA / EUROfusion
First-of-a-kind fusion plants could displace fossil peakers and firm renewables (2035–2045)
For decades, fusion’s promise has been framed as “limitless, clean energy—someday.” What’s changed is not the physics, but the credibility of the engineering path. In the past two years we’ve seen repeatable scientific milestones, clearer programme roadmaps, and the first commercial offtake deals. Together they point to a pragmatic role for first-of-a-kind (FOAK) fusion plants this coming decade: displacing gas peakers, providing firm, rampable capacity to stabilise renewable-heavy grids, and reducing exposure to fossil price shocks. That—not instant baseload abundance—is the commercial entry point that matters.
Consider the scientific signal. In February 2024, the UK Atomic Energy Authority’s JET facility in Oxford closed its campaign with a world-record 69 megajoules over ~5 seconds using deuterium-tritium fuel—an experimental result that delivers high-quality operating data for next-gen machines. The record is less important than the reproducibility and the systems learning it unlocks for future plant design. In parallel, the US National Ignition Facility has repeatedly achieved ignition, culminating in multi-megajoule yields (up to 8.6 MJ reported in 2025). While inertial confinement and magnetic fusion target different applications, the message is common: the physics conditions for net-energy fusion can be produced and tuned with increasing confidence. That lends momentum to the engineering programmes now under way.
Programme reality matters more than headlines. ITER has presented an updated baseline that prioritises earlier research operations with hydrogen and deuterium, targeting deuterium–deuterium fusion operation in 2035 before stepping up to full-power phases. In practical terms, this means sooner learning on plasma scenarios, components and maintenance regimes—inputs industry needs for design-to-cost pathways. The UK’s prototype plant, Spherical Tokamak for Energy Production (STEP), is scoped to put power to the grid in the 2040s at West Burton, with a design that bakes in tritium self-sufficiency via breeding blankets—again, an engineering decision that aligns the technology with real-world fuel cycles rather than lab shots.
Private developers are stitching those ingredients into bankable projects. Helion’s power-purchase agreement with Microsoft—structured to start deliveries as early as 2028 and now backed by site construction activity in Washington State—signals buyer appetite for firm, zero-carbon capacity uncorrelated with weather and fuel prices. Even if schedules slip (they often do for FOAK), the existence of contracted anchor demand from hyperscale buyers is a watershed for financing. Commonwealth Fusion Systems (CFS) is advancing SPARC, enabled by high-temperature superconducting (HTS) magnets proven at 20-tesla—magnets that reduce machine size for a given field strength and open a route to higher power density. Assembly and commissioning updates through mid-2025 suggest steady, execution-led progress.
Where fusion slots into the power system first
The earliest commercial win is not replacing coal baseload; it’s displacing fossil peakers and providing firming for variable renewables. Fusion plants—especially compact designs targeting tens to low hundreds of megawatts—can be sited near existing substations and contracted on availability and ramping value, not just kilowatt-hours. That makes them natural complements to wind and solar portfolios and an insurance policy against gas-price volatility and geopolitical supply risk. It also addresses the grid-balancing problem that storage alone can’t economically solve across multi-day weather events.
A go-to-market arc that financiers understand
The commercialisation path will look familiar to anyone who has lived through first-plant deployment in other capital-heavy sectors:
FOAK → NOAK cost curve: The first plant will be expensive; the point is to retire technology and supply-chain risk quickly, then standardise. Developers should publish a clear transfer plan from FOAK learnings to “nth-of-a-kind” (NOAK) designs. Iteration cadence wins here.
Anchor offtake and creditworthy buyers: Expect early PPAs and availability-based contracts with data-centre operators, utilities and energy-intensive industry. The Helion–Microsoft deal is the bellwether for how these contracts might be structured—even if volume and start dates evolve.
Policy enablers as market design, not subsidy: Contracts for Difference for firm, clean capacity; accelerated interconnection for non-emitting firm resources; and clarity on fusion-specific regulation (distinct from fission) lower the cost of capital without picking technology winners.
Co-location for heat and power: Industrial sites (chemicals, steel, cement) value high-grade heat; fusion’s thermal output creates optionality for combined heat and power, improving project economics versus electricity-only siting.
Fuel cycles, supply chains, and what’s “different” about fusion
Unlike fission, fusion’s primary fuel inputs can ultimately be derived from widely available materials: deuterium from water and tritium bred from lithium inside the plant. The engineering challenge is breeding enough tritium to be self-sufficient while extracting heat efficiently and managing materials fatigue—hence the focus on breeding-blanket design and pebble-bed architectures. Those are solvable, industry-scale problems, not scientific unknowns, and they shape siting (access to lithium supply chains), operations (online tritium handling), and O&M models (component swap-out regimes).
On the hardware side, HTS magnet supply, cryogenics, power-electronics stacks, and precision manufacturing will dominate early bottlenecks. The upside: these are sectors with mature suppliers once specifications stabilise. CFS’s 20-tesla magnet demonstration shows why magnets are more than a component—they’re the lever that makes machines smaller and cheaper, and turns civil engineering into a manageable line item rather than the defining cost.
Risks to be honest about—and how to de-risk them
Schedules will slip. Multi-disciplinary integration—plasma control, materials, fuel cycle, maintenance robotics—will produce “unknown unknowns.” The credible response is programme transparency: publish phased test objectives (not just go/no-go dates), pre-contract long-lead items to protect the critical path, and structure PPAs with staged milestones so buyers see progress before megawatts flow. For policymakers, the task is to align permitting and market mechanisms so FOAK plants can connect early, prove performance, and earn returns for availability and grid services while capacity scales. ITER’s updated phasing and STEP’s staged approach model the mindset: bring forward learning, then scale.
What success looks like by the mid-2030s
A realistic “win” for fusion in 2035–2045 is a handful of FOAK/early-NOAK plants delivering tens to hundreds of megawatts each into real grids—dispatchable, zero-carbon, with contractual flexibility that lets utilities shape net load and lets corporates hedge long-term energy risk. That’s enough to start pushing gas peakers off the margin, firm up renewable portfolios, and chip away at dependency on fossil fuels. The science will keep improving; the commercial proof will come from machines connected to the busbar. The difference now is that there’s a line of sight from records in the lab to megawatts on the grid—and buyers are already signing on.
