The 1cFE initiative has been completed. Producing net electricity continuously from fusion, at any cost, would be one of humanity’s crowning technical achievements. If that electricity remained expensive, its wider economic impact would be limited.
My rough assessment is that fusion at 15¢/kWh could become a substantial part of the energy mix in some markets over time. At 10¢/kWh, it could compete for new baseload capacity in most markets. At 5¢/kWh, I would expect it to become the default new build for firm baseload and rapidly gain a large share of generation. These are generation costs at the plant gate, assuming reliable plants that can be financed and built at scale. Local alternatives and construction speed would also determine adoption.
Approaching 1¢/kWh, I would expect fusion to supply most new and replacement firm capacity. The resulting cost reduction could represent a substantial civilizational upgrade at a time when more of the economy is electrifying.
With support from Astera Institute, we chose one cent per kilowatt-hour as a deliberately extreme target. It forced us to question assumed constraints and keep asking what would need to be true. We also asked whether it could happen within the next decade. Electricity infrastructure is replaced slowly, and new plants commit capital for decades. To materially affect those decisions, fusion would need to become very cheap and be deployed quickly.
Our conclusion is that one-cent fusion power is unlikely, but not impossible. Of the routes we studied, the two we judge most likely to reach it both use lower-neutron fuels: D-He3 with direct energy capture, and p-B11 in a simple magnetic mirror. The corridor synthesis shows how both cross the target in speculative cases requiring large plants, favorable financing and unproven plasma performance. Success would require favorable assumptions to hold, extraordinary scientific and technical work, substantial capital, considerable luck and perhaps significant incentives for early customers.
Reaching that cost within a decade seems vanishingly unlikely. Rapid improvements in scientific and engineering tools offer hope of much faster progress. Early research results from increasingly capable frontier models suggest that fusion, like other fields, could advance in ways that are difficult to anticipate. During this initiative, we could not estimate that acceleration with confidence.
What we produced
We built 1costingFE, an open-source cost model covering four fuels and 17 confinement concepts, with power balances and sensitivity analysis. Its cost-accounting framework builds on earlier work, including pyFECONS, developed by Simon Woodruff and colleagues. We are grateful to ARPA-E for funding the fusion technoeconomic analysis work behind pyFECONS, which was an immense help to 1cFE. ARPA-E personnel have told us they have found the updated tool useful.
Our automated research and costing pipeline examined 38 fusion concepts. The Concept Explorer brings together their sources, assumptions and outputs. Seven concepts had no published quantitative design point in the material reviewed. Their costs depended on assumed inputs and could not be treated as well-supported estimates of those designs.
We explored six corridor premises for how fusion might reach one cent per kilowatt-hour. Each study examined selected designs or scenarios within a corridor, so its results do not settle the prospects for the whole corridor.
What we learned
The costs outside the fusion core can already exceed one cent. In our free-core analysis, the modeled electricity cost of a 1 GWe D-T plant remained about $29/MWh in 2024 dollars, or 2.9 cents/kWh, after the fusion core’s cost was removed. That case assumed 85% availability, a 7% cost of capital, a 30-year life and six years of construction. Larger plants, cheaper finance and other favorable assumptions lowered the result. Reaching one cent would require a reduction in these other costs as well as a sufficiently cheap core.
Fuel choice changes the whole plant. It affects shielding, tritium systems, maintenance, staffing and energy conversion. Our models found lower plant costs for some advanced-fuel cases, assuming they could achieve the required plasma performance. The savings also depended on estimates of the buildings and equipment those plants would need, for which reference data are limited. Our direct-conversion analysis showed why converter efficiency alone does not determine the outcome. The fraction of energy available as charged particles limits what the converter can recover, and the remaining thermal equipment and recovery hardware still cost money.
Availability and financing strongly affected the modeled costs. Availability was the most sensitive input in 29 of the 38 concepts we screened. Interest rate was among the five most sensitive inputs in 33. Lower availability spreads capital and fixed operating costs over fewer kilowatt-hours. Higher financing costs increase the revenue needed to recover construction spending. These rankings apply to the selected design points, normalized to 1 GWe.
Learning requires production that someone can finance and buy. Our work on material multipliers and learning curves examined how far component costs might fall and what could drive those reductions. If early plants cost more than customers will pay, someone must cover the difference before production can continue. A low projected cost for a mature plant therefore needs an account of how the industry gets there. Small devices may offer faster iteration and access to early markets, but the experience must transfer to later plants, and the physics must work at the proposed scale.
We used AI extensively throughout the residency, particularly for agentic research, as its capabilities improved. We also found substantial limitations and errors, including inputs combined from inconsistent design points and library defaults that strongly affected costs. Checking sources and model inputs remained essential.
What follows
The residency reinforced my belief that some of us in the field need to strive for revolutionary outcomes, not just evolutionary improvements, and work backward from audacious goals. The work also built my conviction that, if ultra-low-cost fusion is possible, proton-boron-11 is the most promising path and remains comparatively underexplored and underfunded.
Tal and I are building Borealis Fusion to pursue that path and the cheapest firm baseload power. Other routes may succeed, but we believe this gives us the best chance of reaching $0.01/kWh and having the greatest impact. If Borealis succeeds, it will be in part because of the conviction we built using the tools and findings from this initiative.
The models let others test how new evidence on component life, procurement costs or financing changes the estimates. Demonstrating the required plasma performance remains a separate task.
Our published work is available at 1cf.energy, with source code in our GitHub repos. I hope others will use the models, correct the assumptions and test designs we did not study.
Thank you to my exceptional colleagues on this initiative: Tal Rubin, Mallory Snowden and Reid Westwood. Thank you to Astera for funding the work, and to the researchers, engineers and fellow residents who shared their time and challenged our thinking.
Damien
Conflict of interest
Damien Scott and Tal Rubin are co-founders of Borealis Fusion, a company developing proton-boron-11 fusion. Astera Institute is an investor in Borealis Fusion.
Originally published at 1cf.energy.