Altman Holdings’ fusion startup Helion has seen its valuation soar to $15.5 billion, with an aggressive goal of powering Microsoft data centers by 2028.

Sam Altman owns about one-third of the fusion startup Helion, which has completed a $465 million Series G funding round, valuing it at $15.5 billion. Helion is racing at full speed to build its first power plant, Orion, aiming for 2028 based on an industry-first fusion power purchase agreement with Microsoft.
(Background: OpenAI is negotiating a large-scale power purchase agreement with fusion startup Helion, Altman's energy alchemy)
(Additional context: AI is an energy-consuming monster! Sam Altman believes "fusion power" will bring a breakthrough in energy)

The electricity consumption of global AI data centers is skyrocketing at a rate of several tens of percentage points per year, and traditional power grids are simply not prepared. Fusion startup Helion announced on Thursday that it has completed a $465 million Series G funding round, with a valuation reaching $15.5 billion, nearly tripling the level from 2025.

This round was led by Thrive Capital, with new investors including Alta Park Capital, Anti Fund, BoxGroup… and others; existing shareholders Capricorn Technology Impact Funds, Lightspeed Venture Partners, and a university donation fund also continued to follow.

The most prominent name behind this funding is Sam Altman. He revealed during the Musk v.s Altman court hearing this year that he owns about one-third of Helion; but he also resigned from Helion’s board in March this year, trying to clarify the boundaries of his interests between Helion and OpenAI.

The timing of this funding coincides with Helion’s full-speed effort to build its first power plant, Orion. Helion has signed the industry’s first fusion power purchase agreement (PPA) with Microsoft, which means: the buyer (here, Microsoft) pre-commits to purchasing a certain amount of electricity from a future power plant, effectively locking in energy supply through pre-ordering.

According to the agreement, Helion commits to supplying at least 50 MW of commercial fusion electricity to the grid as early as 2028, with the electricity supplied to a Microsoft data center located in central Washington State. This schedule is the most aggressive public commitment currently in the entire fusion industry.

Why are capitalists rushing to fund a technology that will only be commercially viable in the next decade?

Training a top-tier large language model consumes roughly the same amount of electricity as a small city over several months. The electricity used for inference may multiply that several times.

The supply speed of traditional thermal power plants and renewable energy cannot keep up with the expansion of data centers; fusion, in theory, offers an almost perfect solution: fuel extracted from seawater, deuterium, is nearly inexhaustible; unlike nuclear fission reactors, it leaves behind high-level radioactive waste that needs storage for tens of thousands of years; and it can provide stable power around the clock, unaffected by weather.

The problem is, fusion has been an elusive "promise" for seventy years—"commercial use in twenty years" has been a mirage.

Helion’s approach bets on a completely different engineering route from its peers. Most fusion startups operate on the logic of using magnets or lasers to confine or compress ultra-high-temperature plasma, causing light atomic nuclei to collide and release enormous heat, which then boils water to drive steam turbines—essentially the same as coal-fired power, just with fusion as the heat source.

Helion skips the steam turbine step. It uses magnets to compress fuel, allowing fusion reactions to occur inside the plasma. When the plasma heats and expands, it pushes against magnetic fields, and Helion directly converts this thrust into electricity. This principle is somewhat like an electric vehicle’s regenerative braking, where the motor reverses, converting kinetic energy back into electrical energy to recharge the battery—directly transforming mechanical motion into electricity, rather than first converting to heat, then boiling water, then driving turbines. Theoretically, this route can significantly reduce energy loss during conversion.

However, some fusion physicists are skeptical of this approach, partly because Helion rarely publishes peer-reviewed papers, making it difficult for external researchers to verify the physical basis of its claims. CEO David Kirtley’s response is straightforward: "We don’t want to talk about fusion; we want to build it."

The entire industry is racing for funding: but commercial operation is still a matter for the next decade.

Helion is not the only beneficiary. Fusion has become a new darling of venture capital in recent months. Just last week, Focused Energy and Thea Energy announced new funding rounds, with $240 million and $100 million respectively; in February, Inertia Energy emerged from stealth with a $450 million Series A; a month earlier, Type One Energy was raising $250 million for its Series B. In just a few months, the entire track has absorbed over a billion dollars in funding.

However, beneath this funding frenzy lies an unspoken fact: most of these companies claim that their first commercial-scale power plants will only be operational "by the mid-next decade," around 2035. Helion’s 2028 target is the most aggressive in the industry, yet it has no peer-reviewed publications available for external validation of its technology feasibility.

In other words, capital is buying an option: in the AI energy arms race, if someone actually manages to build fusion power, you want to ensure you have a seat at that table.