The Energy Department has a roadmap to fusion by the mid-2030s. Read the part that commits no money.

The most honest line in the Energy Department's new fusion roadmap is not a milestone. It is a disclaimer. Tucked into a document that promises commercial fusion power on the American grid by the mid-2030s is a sentence admitting that the roadmap commits the department to no specific funding, and that its timelines depend on public-private partnerships and Congressional appropriations that do not yet exist. I have spent a long time reading fusion announcements, and I keep a private archive of every 'fusion by [year]' claim alongside the year it actually turned out to be. By the standards of that archive, this is a careful, serious, unusually candid document. It is also, read closely, an aspiration with a calendar attached — and the calendar is the part to be skeptical of.

Let me be clear about what I am not saying, because the field gets miscovered in both directions. I am not saying fusion is fake, or perpetually fifty years away as a matter of physical law. The science has moved, genuinely and recently, and I will get to why. I am saying that a roadmap is a statement of intent, and intent is the cheapest thing in this entire enterprise. The expensive things — the materials that survive a neutron flux, the tritium you have to breed because the world's supply would otherwise fit in a few suitcases, the engineering of a machine that does all this reliably for thirty years — are exactly the things a roadmap can schedule but cannot accelerate by being written down.

What the document actually says

The Fusion Science and Technology Roadmap, released in finalized form on June 9, organizes the national effort around a slogan: Build, Innovate, Grow. Build the infrastructure — test stands, materials facilities — to close the gaps between a physics experiment and a power plant. Innovate through research, high-performance computing and AI. Grow the ecosystem through public-private partnerships, supply chains and a workforce. The actions are sorted into near-term (the next two to three years), mid-term (three to five) and long-term (five to ten). Dr. Darío Gil, the under secretary for science, framed the moment in the announcement: fusion, he said, 'has entered a new era defined by extraordinary scientific progress and public-private momentum.'

The momentum is real and worth stating plainly. The department says the roadmap drew on more than 800 scientists and engineers, over 70 universities, more than 10 national laboratories and over 15 private companies, against a backdrop of more than $10 billion in private fusion investment. That private number is the genuinely new thing in the field this decade. When I started covering fusion, it was almost entirely a government science program measured in physics results. Now there is a parallel industry of venture-backed companies promising not results but products, and a survey of those companies finds a majority believing a commercially viable plant will exist by 2035, with a few saying before 2030. The roadmap is, in part, the government trying to get into formation with that private optimism.

A roadmap is a statement of intent, and intent is the cheapest thing in this entire enterprise.

Demonstrated versus announced

Here is the distinction I always reach for, because it is the entire story: the difference between what has been demonstrated and what has been announced. What has been demonstrated is remarkable. In late 2022, the National Ignition Facility achieved ignition — a fusion reaction that released more energy than the laser light delivered to the fuel capsule — and has since repeated and exceeded it. Tokamaks around the world have held high-performance plasmas for longer and longer durations. These are real, replicable, hard-won results, and anyone who waves them away has not understood how difficult they were.

But ignition in a single controlled pulse and a power plant are different problems separated by orders of magnitude, and most of those orders are engineering, not physics. NIF's gain is measured against the energy in the laser beams, not the vastly larger energy drawn from the wall to fire them. A plant has to produce net electricity, not net fuel-capsule energy, and it has to do so continuously, economically, and for decades, while its inner walls are bombarded by neutrons that make materials brittle and radioactive. The roadmap's own Build pillar — the test stands for structural materials, the work on plasma-facing components — is an admission of how much of this remains unsolved. You do not build a national program to study materials damage if the materials problem is behind you.

This is why the funding disclaimer matters so much. A roadmap that schedules the hardest engineering of the century into a five-to-ten-year window, and then notes that it commits no money and depends on appropriations Congress has not made, is telling you where the real uncertainty lives. It is not in the slogan. It is in the gap between the milestone chart and the budget line that would pay for it. The Government Accountability Office made a version of this point in its own review, warning that misalignment between public and private research priorities was itself a risk to commercialization — that everyone agreeing fusion is close is not the same as everyone working on the same hard problem.

On what timescale

So let me ask the question I always ask out loud: on what timescale? The private companies carrying much of this optimism are real and serious. Commonwealth Fusion Systems is building SPARC in Massachusetts and has announced a first commercial machine, ARC, in Virginia. Helion has signed a power-purchase agreement with Microsoft for a 50-megawatt plant in Washington State. These are not vaporware; they are funded, staffed, pouring concrete. But a power-purchase agreement is a contract, not a kilowatt. SPARC has not yet run. ARC has not been built. The agreements are denominated in the late 2020s, which in fusion's history is the horizon that has a way of staying just over the next hill.

And this is where my archive earns its keep. The mid-2030s target is not new; it is the latest position of a goalpost that has been moving my entire adult life. The honest reading of this roadmap is that it neither confirms nor refutes that target — it simply restates it and lists the work required to find out, while carefully declining to promise the resources. That is not a criticism of the people who wrote it. It may be the most scientifically responsible thing the document does. The roadmap is candid about its own contingency; the danger is only that the candor gets stripped out in the retelling, and 'a plan that could deliver fusion by the mid-2030s if Congress funds a decade of unsolved engineering' collapses into 'fusion by the mid-2030s.'

Why this is worth getting right

There is a reason I am this insistent on holding the two truths at once. Fusion's biggest enemy has never been physics; it has been the erosion of patience that follows each round of overpromising. Every time a date slips, a little of the public goodwill that long-horizon science depends on drains away, and fusion is the longest horizon we have. The way you protect a field that genuinely deserves decades of investment is by refusing to sell it as if it were arriving on a quarterly schedule. A roadmap that says 'here is the work, here is the order, and here is the money we have not yet been given' is, paradoxically, the kind of honesty that keeps the funding coming — because it does not set a deadline that, when missed, becomes the headline.

So read the roadmap. It is a good document, written by serious people who know exactly how hard this is. Admire the Build-Innovate-Grow logic, take the private investment seriously, take the NIF results seriously. And then read the disclaimer again, the one about funding and appropriations and partnerships that do not yet exist, and understand that it is not boilerplate. It is the most accurate sentence in the file. Fusion is real, and important, and coming — and it will take longer than the calendar on the cover. Both things are true. The roadmap, to its credit, almost says so.