Ten thousand jobs was the number on the banner. "Fusion has the potential to provide virtually limitless clean energy while supporting more than 10,000 jobs across the UK by 2030," Michael Shanks, the UK's Minister for Energy, said at Mansion House in London on 14 September, as Britain and the United States signed two agreements at the Global Fusion Policy Summit.
One was a supercomputing link-up. The other was a joint statement on regulation.
Neither was a physics result.
Last issue this column traced a fusion race fragmenting into national camps, each chasing its own machine, its own money, its own rules. This week the same governments signed a document that extends cooperation explicitly into regulation, industry and workforce — the plumbing of a business, not the physics of a reactor.
Read the fine print and the shift is unambiguous. The joint statement, co-chaired by Shanks and US Under Secretary for Science Darío Gil, commits signatories to "proportionate, risk-informed regulatory approaches," to regulator-to-regulator exchanges of knowledge, and to seats in international standards bodies.
It commits them to treating fusion's safety profile as distinct from fission's, with radiological safety standards — not fission standards — as the more appropriate starting point.
That is a substantive decision. It is also, notably, silent on tritium, on fuel cycles and on cost.
The supercomputing agreement is the concrete half. The UK Atomic Energy Authority and Princeton Plasma Physics Laboratory signed a Joint Declaration of Intent to federate UKAEA's SUNRISE supercomputer with PPPL's STELLAR-AI platform.
SUNRISE is backed by £45 million of government money and sits inside the AI Growth Zone at Culham. STELLAR-AI is operated with support from Princeton University.
The two labs run similar compact spherical tokamaks — MAST Upgrade in Oxfordshire, NSTX-U in New Jersey — which makes shared AI training unusually practical. Models trained on one machine struggle to predict across machines; two datasets plus two computers is a better bet.
"By federating the SUNRISE and STELLAR-AI computer platforms, we can train AI on data from two leading fusion facilities," said PPPL chief scientist Jonathan Menard, "and accelerate the path to a compact fusion power plant."
Shantenu Jha, who heads computational sciences at PPPL, put the ambition more sharply: "We will turn a collection of supercomputers into a single engine for fusion discovery."
Fine. But federated supercomputers do not breed tritium, and they do not tell an investor what a kilowatt-hour will cost.
Fine. But federated supercomputers do not breed tritium, and they do not tell an investor what a kilowatt-hour will cost.
The fuel question is where the pacts stop. The UK is putting £180 million into the Lithium Breeding Tritium Innovation programme, described in its own announcements as a "globally unique facility" to address "one of the key technical challenges facing future fusion power plants."
That phrasing is the tell. Tritium is not a supply-chain detail; it is the fuel, and a fusion plant that cannot breed its own tritium burns through a material that barely exists commercially. A £180 million research facility is an admission that the answer is not in hand.
Cost is quieter still. The government's headline market figure is £3 trillion to £12 trillion globally, an estimate spanning two orders of magnitude, which is another way of saying nobody knows.
The actual industrial money announced alongside the summit was £2.63 million for a University of Birmingham and Electric Power Research Institute collaboration on reactor shielding materials, and up to £3 million from Japan's Kyoto Fusioneering into Project ALBION for materials testing on STEP.
Set those against a 10,000-job claim for 2030 and the arithmetic of the headline becomes visible. Jobs are a policy target. They are not an energy forecast.
The wider pattern says the same thing. At the same summit, Korea's Ministry of Science and ICT sent First Vice Minister Koo Hyuk-chae, who met Shanks to review joint development of high-temperature superconducting magnets for STEP, the UK's 2040 demonstration plant.
Korea's stated goal is electricity demonstration in the 2030s. Its ministry was explicit that cooperation is expanding beyond R&D into regulation, standards, supply chains and industrial ecosystems — because, it said, commercialisation requires a safety and regulatory framework that research programmes never built.
Nine governments took part: the UK and US as co-hosts, plus Canada, France, Germany, Italy, Japan, Korea and Finland.
And there is Tennessee. The UK has signed a Letter of Intent with the state covering R&D, skills and workforce development, regulation, and access to supply chains — the same four items, in the same order, as the London statement.
Supply chains, not plasmas. Workforces, not confinement times.
This is what industrial policy looks like when a technology is pre-commercial: governments cannot yet buy the product, so they build the conditions under which someone else might. Regulatory clarity is the cheapest lever available.
The template already exists on the fission side. Under the Atlantic Partnership for Advanced Nuclear Energy signed last September, the US and UK agreed to fast-track reactor design checks for designs already rigorously reviewed in the other country, avoiding duplicated work. Fusion is now being handed the same treatment.
It is a reasonable bet, and there is evidence it works. The UK was the first country to establish a dedicated fusion regulatory framework, and it has since published a Regulation Landscape, an Early Engagement Process run by the Environment Agency and the Health and Safety Executive, and a draft Fusion National Policy Statement. In March 2026 it became the first country to commit to a dedicated market framework for fusion.
Predictable rules lower investor risk. Lower risk lowers the cost of capital. That is the whole mechanism.
But harmonisation is also soft in the places that matter. The joint statement aligns expectations "where appropriate," supports greater alignment "over time," and pursues "technology-appropriate harmonisation." Those hedges are honest — different countries have different legal systems — and they are also where deals go to die.
Meanwhile the things that decide whether fusion is an industry remain outside the pacts. Where the tritium comes from. What a plant costs per megawatt. Whether a licensed design can be financed.
The bet, stated plainly: by synchronising rules now, governments can manufacture a licensable fusion industry before the physics and economics are settled, and be first in line when they are.
If it works, the 10,000 jobs arrive as a consequence. If it doesn't, the number stays what it is today — a banner.