I joined Astera Institute this month as a Resident to create and lead 1cFE, an initiative to explore two questions: can fusion reach sub-1¢/kWh Levelized Cost of Electricity (LCOE), and, if so, what must be true to achieve it within ten years.
Fusion today sits between hype and skepticism. 1cFE exists to add rigorous analysis that helps the field converge toward truth.
Why sub-1¢/kWh matters: it sets a hard, aggressive target below today’s best new-build costs. Meeting it would redefine what civilization can choose to build and do, because energy sets the constraints. Our frontier backcasting approach avoids energy incrementalism and the replacement trap. We work backwards from the sub-1¢/kWh end goal to map corridors.
It is common to overestimate what is possible in one year and underestimate what is possible in ten. That gap will widen in the decade ahead as AI accelerates progress across science, engineering, and manufacturing. A decade is short enough to drive focus and long enough for breakthroughs. Now is the time for boldness.
We will publish our models, data, code, and assumptions in public, including corrections and negative results so anyone can check, reuse, and build on the work. Outputs will include a transparent technoeconomic model, cost-first corridors, and AI benchmarks for science and engineering that speed evaluation and iteration.
What 1cFE will do
The 1cFE initiative at Astera aims to explore whether fusion could reach sub-1¢/kWh LCOE and, if so, what would have to be true to achieve that within ten years. Fusion at 5¢/kWh or even 10¢/kWh LCOE would still be meaningful. It would clean up parts of the grid and displace fossil power. But it would not expand what our civilization can build. The question we are asking is whether fusion can be within our lifetimes a transformative energy source, not just a replacement. Sub-cent electricity would change the feasible set for industry, computing, water, transportation and materials.
Our method is cost-first. We will map corridors to sub-cent power that tie physics, engineering, manufacturing, and finance into one model, and publish the analyses and artifacts. We are studying acceleration. We will pilot new ways of doing science, engineering, and deployment using AI and the emerging next generation of CAE tools. We will map critical paths across design, build, test, and scale. We will interrogate bottlenecks in supply chains, licensing, and workforce formation, and state which are likely to yield soon and which are likely to persist.
Outputs
Our outputs will be public and reusable. 1cFE will publish technoeconomic analyses, reusable tools, and clear evaluations of models, forecasts, and claims. All code and data will ship under permissive licenses so others can review, reuse, and repurpose the work. Corrections and negative results will remain in the record.
In Q1 2026 the initiative will release our first versioned TEA with sensitivity analysis, a small set of reference designs spanning fuels and confinement schemes, and a preliminary supply-chain model. In Q2 2026 we will publish updated cost corridors with external reviews and a report benchmarking new AI tools and techniques across science, first-of-a-kind engineering, and manufacturing and deployment, including datasets, tasks, metrics, and reproducible code.
Goals
We believe clear, ambitious, and quantitative goals help if they are achievable. 1cFE’s role is to investigate the plausibility of sub-1¢/kW fusion power.
We are agnostic to confinement approaches and fuel. We do not back a central plan or a single true path. We keep the outcome front and center and evaluate approaches on how they move cost, time, and risk.
We will run two modes in parallel. First, backcast from sub-cent LCOE to the technical, industrial, and policy constraints it implies. Second, test new science, tools, and processes that accelerate development and lower costs.
1cFE serves the subset of the fusion sector that chooses to pursue sub-cent LCOE. We surface the highest-leverage problems and fastest corridors, and point to where AI tools can accelerate them.
We will publish reusable frameworks, data, and code. We will log corrections and negative results quickly to save others time and money. The aim is to cut uncertainty, focus resources where they matter for sub-cent fusion, and raise the pace and rigor of the work. If sub-cent proves out, we will show how. If it does not, we will show why.
Fusion at sub-cent cost would not just power civilization, it would upgrade it. Achieving it would mark a turning point in humanity’s control over energy and our capacity to build the future we choose.
LCOE is the north star
LCOE is the primary figure of merit in electricity generation, combining all relevant factors: capacity factor, capital cost, operations and maintenance, fuel cycle, decommissioning, and financing. Its counterpart, LACE, measures the value of that electricity to the grid rather than its cost, and together they form the LACE/LCOE ratio, which captures both economic competitiveness and the value of dispatchability across different energy sources. Throughout this post, electricity costs are expressed in both ¢/kWh and $/MWh, which are used interchangeably. 1¢/kWh = 10$/MWh.
Other factors, such as carbon intensity, enter the equation directly if there is a carbon price or indirectly if customers are willing to pay a premium for zero-carbon electricity. Investment tax credits and production incentives modify the realized LCOE but do not change the physical cost basis that determines whether a technology can compete on fundamentals.
It is instructive to compare LCOE to wholesale electricity prices while understanding differences. According to FERC in 2024 the average U.S. wholesale price was about 38$/MWh, ranging from roughly 28$/MWh in SPP North to 60$/MWh at Mid-Columbia. These values reflect the marginal cost of power from existing plants, many built decades ago, long since depreciated, and operating on sunk capital. LCOE describes what it would cost to replace those assets now.
Lazard’s 2025 LCOE+ report, which estimates unsubsidized new-build costs, places most technologies above current wholesale prices because it includes contemporary capital and financing assumptions.
source: Lazard’s 2025 LCOE+ analysis
AI acceleration
AI is already compressing the loops that constrain fusion progress, and the pace is increasing. It is the single biggest reason to revisit timelines for fusion economics.
DeepMind and EPFL demonstrated learned plasma control on TCV, showing reinforcement learning can shape and stabilize tokamak plasmas. A Princeton team reported AI control that reduces the likelihood of disruptive tearing instabilities on DIII-D. LLNL is using AI agents to accelerate LIF target design exploration. CFS and DeepMind announced a program to apply AI across SPARC and ARC development. For a broader survey see CATF’s Survey of Artificial Intelligence and High Performance Computing Applications to Fusion Commercialization. These advances are why AI is central to 1cFE’s frontier backcasting approach.
This alignment between fusion and AI is now a national focus. The U.S. Commission on the Scaling of Fusion Energy’s Fusion Forward report (2025) highlights predictive fusion science, AI-driven design, and rapid digital engineering as core to achieving commercial viability on a “competition-relevant timeline.” 1cFE applies the same frontier-backcasting lens to quantify what this acceleration could mean in practice.
We are exploring three areas for acceleration: AI for science, AI for first-of-a-kind engineering, and AI for manufacturing and commercial deployment. We evaluate emerging tools and techniques, benchmark how much they could reduce time, cost, and uncertainty on representative tasks, and publish reproducible methods and results. With large-scale compute through Astera’s affiliate Voltage Park, we can run these benchmarks at meaningful scale and share datasets, scripts, and baselines.
Why abundant energy matters
At the limit, everything humanity values depends on abundant, inexpensive energy. Cheap, reliable energy lowers the cost of food, shelter, transport, water, and compute. It reduces conflict, expands opportunity, lifts living standards and ultimately unlocks things not even imagined today. Abundant energy is a measure of what a civilization can choose to build and do. For more on why abundance is the master lever for progress I recommend Vaclav Smil’s Energy and Civilization: A History, Eli Dourado and Austin Vernon’s Energy Superabundance and Ian McKay’s The future is made of energy.
AI makes this more immediate. Cognitive labor is being shifted from running on human brains powered by food to data centers powered by electricity. This makes dependable low-cost electricity increasingly important.
The Kardashev scale is a metric for civilizational development based on energy use. A Type I civilization harnesses all the power incident on its planet from its star: about 1.7 × 10¹⁷ W for Earth. Humanity’s current primary energy use averages roughly 2 × 10¹³ W, placing us about four orders of magnitude below Type I.
Why fusion
Most usable energy across the universe is fusion with extra steps. The best fusion reactors we know are stars that use gravitational confinement. The Sun sits about 150 million km away and delivers on the order of 1 kW/m2 meter at Earth’s surface on a clear day.
The reduction in the cost of solar PV in the last 25 years has been remarkable: module costs have fallen 96% from $6.20/W in 2000 to $0.26/W in 2024. Solar plus storage now sets the competitiveness target for firm clean power. Utility-scale solar is among the lowest-cost new-build electricity generation.
So why is terrestrial fusion worth pursuing? First, it could plausibly become the lowest LCOE source if key design and manufacturing levers move enough. Second, it offers system advantages that solar lacks: compact land use, siting close to large loads, firm output at high capacity factor, and access to high-temperature heat for industry. Third, there is an upper limit to how much energy can be harnessed terrestrially from solar. Finally, fusion is also the ultimate propulsion source for space, with energy density roughly a million times that of chemical fuels, enabling fast, frequent interplanetary travel once mastered on Earth.
State of fusion
Generating electricity from terrestrial fusion will be a landmark achievement for humanity. The last decade has brought real and accelerating public and private progress toward it.
The two main figures of merit for fusion performance are the triple product and gain. The triple product n \cdot T \cdot \tau_E combines plasma density n, temperature T, and energy-confinement time \tau_E.
Gain Q measures how much fusion energy is produced compared with the energy supplied.
- Q_{\text{sci}} = \frac{E_{\text{fusion}}}{E_{\text{input}}} is the scientific gain, comparing fusion output to the energy delivered to the plasma or target; breakeven occurs at Q_{\text{sci}}=1.
- Q_{\text{eng}} = \frac{E_{\text{fusion}}}{E_{\text{wall}}} includes the driver efficiency from wall plug to plasma.
- Q_{\text{plant}} = \frac{P_{\text{net,grid}}}{P_{\text{internal}}} compares net electric power exported to the power consumed internally, including all recirculating loads. A practical power plant must achieve Q_{\text{plant}} >> 1.
Triple product and gain matter because they set driver power, component lifetime, and availability, which flow directly into LCOE through capital, O&M, and net output.
The below charts progress in the triple product over time, with contours of constant Q_{\text{sci}}. It is a scientific benchmark for tracking physics and machine performance, not an economic one.
Source: Wurzel & Hsu (2025)
In December 2022, Lawrence Livermore’s NIF achieved scientific breakeven, producing 3.15 MJ of fusion energy from 2.05 MJ on target. This was the first time a controlled experiment generated more fusion energy than delivered to the fuel. In 2025 NIF produced 8.6 MJ of fusion energy while delivering 2.08 MJ of laser energy to the target, a target gain (Q) of 4.13. However, given the ~0.5 percent wall-plug efficiency of NIF’s laser drivers, Q_{\text{eng}} remains far below one.
No private company has yet demonstrated Q_{\text{sci}} > 1 but many appear close. Commonwealth Fusion Systems expects SPARC to achieve that milestone around 2027, and others have set similar timelines for first breakeven experiments.
The private fusion sector continues to expand rapidly in number, diversity and scale. The Fusion Industry Association’s 2025 report recorded $2.64 billion in new investment last year and $9.8 billion cumulative across more than 50 companies. The number of active firms, confinement approaches, and investment have all increased rapidly over the past five years.
Source: 2025 Fusion Industry Report
In the same survey, five companies anticipated operating a commercially viable pilot plant before 2030, and 35 reported plans to do so by 2035.
Early commercial engagement is also emerging. Google signed a 200 MW fusion power purchase agreement with CFS for the ARC plant planned in Virginia in the early 2030s. Eni announced a $1 billion-plus offtake agreement with CFS. Microsoft agreed to buy power from Helion’s first plant, targeting 50 MW by 2028. These contracts signal intent. They do not disclose price. The question remains whether plants can run at low cost with a high capacity factor.
This is phenomenal progress: new scientific records, credible private programs, growing capital, and initial customer interest. However, none of it guarantees viable, let alone transformative economics. The end state question is how experimental advances translate into plant-level costs, reliability, and capacity factor, which is why Levelized cost of electricity (LCOE) is so important.
Fusion’s LCOE
Fusion is frequently described as clean, limitless, and virtually free. Those words need to be quantified. Milestones in science and engineering are necessary, but deployment ultimately depends on cost per kWh, a value-led rather than research-led approach. On the demand side, fusion at about 10¢/kWh LCOE could be competitive in high-price markets, and at 5¢/kWh the addressable market becomes very large.
Like fission, fusion fuel is cheap; the plant and its operation drive cost. Deuterium-tritium (D-T) is the most mature fusion fuel cycle and is likely to power early pilots and first commercial plants.
Commercial D-T plants will face real operating costs and constraints: breeding and handling tritium, neutron-driven swelling and embrittlement that set replacement schedules and outages, activation and waste handling, and regulatory compliance. With no continuously operating fusion plants yet, the LCOE from fusion is uncertain; it will be set by how these realities play out in practice.
For D-T fusion in which energy must be harnessed thermally, the Rankine cycle steam block sets a floor on LCOE. Using the DOE NETL baseline for coal reference plants, Account 8 Steam Turbine and Accessories is ~500$/kW for subcritical and supercritical units. At 7 percent WACC, 30-year life, and 90 percent capacity factor, the steam block alone contributes roughly 0.5¢/kWh. Steam turbines are mature technology. Fusion concepts using aneutronic fuels and direct electromagnetic capture could relax this floor but these are significantly less mature than D-T with thermal conversion.
Each fusion pathway carries distinct technical risks that shape its economics. Driver efficiency, blanket lifetime, recirculating power, and material damage all affect cost, uptime, and scalability. Contemporary LCOE estimates and targets vary across fuels, architectures, and maturity. Here are recent figures spanning several approaches:
Tokamaks burning D-T are the most mature path. An Energy Policy 2023 study projects early tokamak designs at 15¢/kWh or higher, driven by frequent replacement of vessel components and low-efficiency power cycles.
Four less-mature, advanced fusion approaches (Plasma-Jet MIF, Stabilized Liner Compressor, Staged Z-Pinch, Flow-stabilized Z-Pinch) were evaluated as part of ARPA-E’s ALPHA program. The analysis estimated an average LCOE of 5.11¢/kWh, a low of 3.97¢/kWh, and a high of 6.68¢/kWh.
A 2020 model for inertial fusion found, under optimistic gain and shot-energy assumptions, designs with LCOE as low as about 2.5¢/kWh.
One private fusion company, Helion, has publicly stated a long-term aim of 1¢/kWh
Sub-cent fusion power
Sub-cent power is an aggressive target. Perhaps it is not possible but a clear line focuses design choices and makes architectural decisions and tradeoffs visible. An instructive analogy is orbital launch: $/kWh LCOE in fusion is to $/kg to orbit for the space industry. Prior to SpaceX, $/kg to LEO was around $10,000. Setting an order-of-magnitude reduction as the goal revealed an unavoidable reality: reusability had to be mastered. SpaceX achieved this with Falcon Heavy and is on track to bring cost down another order of magnitude towards $100/kg with Starship. We will use the ¢1/kWh target in fusion to expose which levers are unavoidable and how far they must move and in what order.
What does the envelope for ¢1/kWh look like? With 30 year plant life, 7 percent WACC, and a 95 percent capacity factor, CAPEX of $1,000/kW contributes ¢0.97/kWh before any O&M or replacements. That implies viable sub-cent corridors look like $600-800/kW with very low O&M and durable components paired with very high capacity factor.
How does this compare to other more mature electricity generation sources? EIA’s AEO 2025 lists overnight capital for combined-cycle gas at 868-921$/kW for large H-class plants, advanced nuclear (two AP1000 units) 7,861$/kW, and ultra-supercritical coal without capture 4,103$/kW. Natural gas has low capex but fuel dominates. Fission has high capex and low fuel cost. Coal is mid to high capex with meaningful fuel costs.
Why Astera
Astera’s mission aligns with this work: “The future, faster.” It exists to accelerate science with open tools and public goods, which is what 1cFE seeks to do. As a lifelong sci-fi reader who is impatient with the status quo, I want to help bring forward an abundant future. Our outputs are designed to steer some fraction of ambitious R&D, capital, and policy toward the fastest corridors to sub-cent fusion.
“Astera is a private foundation on a mission to steer science and technology toward an abundant future for all. We believe the coming years will bring an era of unprecedented scientific and technological advancement as exponential progress in AI converges with central advances in other fields to dramatically accelerate innovation. This inflection point provides an unparalleled opportunity to fundamentally rethink the institutions, systems, and tools that drive scientific progress. Astera is uniquely positioned to shape the course of rapid scientific and technological innovation: we are willing to deploy our $2.5B endowment to take on experimental projects that might fail, and we support projects and people that other organizations won’t or can’t.” more astera.org
Astera’s head of strategic investments, Eli Dourado, has written about how we actually get to an abundant future with technoeconomics, entrepreneurship and sociopolitics as the pillars. 1cFE will contribute by making the economics legible and reusable, so others can act on them.
Astera believes that AI will accelerate science and engineering. That is part of the “why now” for this initiative. We intend to examine how AI can shorten design, control, and manufacturing loops, and we will publish the benchmarks that matter. Through Astera’s affiliate Voltage Park, we have access to large-scale compute that can support this work.
Finally, I share Astera’s stance on scientific publishing. Too much of humanity’s distilled knowledge is still locked behind paywalls or slow processes. 1cFE will publish models, data, and code in the open, invite critique, and keep corrections and negative results in the record.
Why me
I grew up in a remote part of Botswana where power was often unreliable. Weeks without electricity were normal. Diesel and gasoline deliveries were vital for vehicles and generators. Solar powered bore-hole pumps were subject to the clouds. The importance of energy was never abstract.
At Williams F1 I learned to turn physics into machines under hard limits. F1 is one of the fastest and most unforgiving prototyping environments in peacetime, a benchmark for the upper limits of engineering speed. The real rate limiters were smart people with keyboards and mice at workstations and synchronizing the organizational context and making decisions. New AI and CAE tools can push this frontier out. It is not surprising that one of the leading companies developing these capabilities, PhysicsX, has its origins in F1.
I named my last company Marain after the language in Iain M. Banks’ Culture novels, a nod to post-scarcity. Marain’s goal was to model and optimize large-scale electric and autonomous fleets to deliver extremely low-cost mobility. We built on open science and open source, including work like the BEAM project at Lawrence Berkeley National Lab.
I am not a plasma physicist. I am a physics-trained builder. My role at 1cFE is to build the scaffolding for cost-first fusion: clear models that tie physics to economics, a team that ships, and an open record that invites critique and collaboration.
Get involved
Join the team by reviewing open roles at 1cf.energy/join. If someone comes to mind, please share the page with them.
For collaborations, send us a note at 1cf.energy/contact.
If you’re curious and want to follow along, subscribe at 1cf.energy for email updates as we release new models, analyses, and updates.
Sub-cent fusion is an audacious hypothesis, but testing it in the open is how progress begins.
Upwards to Kardashev Type I