When Will Fusion Power the Grid? The Real Dates Behind 2028 and 2035

No fusion plant sends power to a grid today. The earliest promised dates are Helion in 2028, aiming for 50 MW for Microsoft, and Commonwealth Fusion Systems (CFS) with its 400 MW ARC plant in the early 2030s. ITER, the giant science machine in France, will not even burn real fusion fuel until 2039. So the honest answer is: a few pilot plants first, grid-scale power later.

What is the earliest date fusion could put power on a grid?

2028, if Helion hits its target. Helion signed the world's first fusion power purchase agreement with Microsoft, aiming to deliver electricity in 2028 and reach 50 MW or greater after a one-year ramp. Constellation serves as the power marketer and manages transmission for the project.

A power purchase agreement is a signed commitment to buy the electricity once it flows, not a research grant or a press-release pledge. That is far sooner than the rest of the field, and plenty of experts think the date will slip. But it is a real contract with a real buyer, not a slide deck.

Sources: Helion | DCD

Which fusion projects have real customers already?

Three names have signed offtake deals. Call this the offtake tell: a signed power purchase agreement is a stronger near-term signal than any press-release date, because a buyer put money behind it.

• Helion and Microsoft. The first fusion PPA ever, targeting 50 MW by 2028.
• CFS and Google. Google agreed to buy half of CFS's first ARC plant, about 200 MW, expected on the grid in the early 2030s.
• CFS and Eni. The Italian energy firm signed a $1 billion-plus offtake deal for power from the same 400 MW ARC plant in Virginia.

Notice who the buyers are. Big tech and big energy, both chasing carbon-free power for data centers and industry. That demand is the same force driving today's power crunch.

Sources: CFS + Google | CFS + Eni

When will CFS actually deliver grid power?

CFS runs a two-step plan. First, prove the physics with a machine called SPARC, a tokamak, or donut-shaped reactor, built at the company's campus in Devens, Massachusetts. CFS says SPARC will hit net energy, more energy out than it takes to run, a milestone called Q greater than 1, in 2027.

Peer-reviewed papers in the Journal of Plasma Physics predict SPARC will clear that bar with margin. SPARC itself is not a power plant. Its job is to settle the science, then hand the technology to its successor, ARC, the machine that sells power.

Second, build that plant. CFS calls ARC the world's first grid-scale fusion power plant, in Chesterfield County near Richmond, Virginia, at a site called the Fall Line Fusion Power Station. It is designed to put about 400 MW of net electricity on the grid in the early 2030s. CFS says it is already talking to customers who want that power.

The technology bet under all of this is the magnet. In September 2021, CFS and MIT tested a high-temperature superconducting (HTS) magnet that reached 20 tesla, the strongest field of its kind ever created at that scale. Stronger magnets mean a smaller, cheaper machine. That is the near-term, investable layer of fusion: magnets and the companies that make them, not the grid date itself.

Sources: CFS SPARC | CFS ARC | MIT Energy Initiative

Why is ITER's date 2039, not 2035?

The title above uses 2035 because that was ITER's old target. The new schedule is later. ITER is the giant international science project in France, built to prove a burning plasma at scale. Its goal is Q of at least 10, meaning 500 MW of fusion power out for 50 MW of heating power in.

Under the latest baseline, ITER will reach full magnetic energy in 2036 and begin deuterium-tritium operation, the real fusion fuel, in 2039. That is a three-year slip on the magnets and a four-year slip on the fuel, measured against the 2016 plan, which had targeted DT operation in 2035.

One number puts the whole field in context. The tokamak fusion energy record, set by the JET experiment, is Q of 0.67, about two-thirds of the energy used to heat its plasma. No fusion machine has yet produced more energy than the whole facility used to run it. ITER is a research reactor, not a power plant. It will not send electricity to a grid.

Sources: ITER FAQs

What dates should a builder or investor actually watch?

Separate the demo date from the grid date. The demo is when a machine first proves net energy. The grid date is when a plant sells power. They can be years apart, and the grid date is the one that slips.

1. 2027: CFS SPARC targets net energy (Q greater than 1). A pass here validates the HTS magnet approach.
2. 2028: Helion targets first power to Microsoft. The first grid date to test the offtake tell.
3. Early 2030s: CFS ARC targets 400 MW to the grid in Virginia, with Google and Eni as buyers.
4. 2036 and 2039: ITER full magnetic energy, then deuterium-tritium fusion. Science milestones, not commercial power.

The materials science underneath all of this, from magnet compounds to reactor walls, is moving fast too, partly because AI is speeding up materials discovery.

Sources: CFS SPARC | Helion | ITER FAQs

If you want help sorting which deep-tech dates are real and which are vibes, that is the kind of thing I do at nomadtechnologist.com.

---

Not legal, financial, or tax advice.