Technology

Spotify Founder Daniel Ek Backs Fusion Startup Europa Fusion, Led by Former ML Researcher

Martin HollowayPublished 28m ago3 min readBased on 1 source
Reading level
Spotify Founder Daniel Ek Backs Fusion Startup Europa Fusion, Led by Former ML Researcher
Photo by Lukasz Kobus/European Commission / CC BY 4.0

Spotify founder Daniel Ek is backing Europa Fusion, a UK-based fusion power startup, according to TechCrunch, which reported the backing on November 20, 2026. Bloomberg first reported Ek's support for the company.

Europa Fusion was founded by Nantas Nardelli, who started the company in February. Nardelli previously worked at Google DeepMind, Facebook AI Research and Microsoft Research Cambridge. Most recently he was head of machine learning at Pacific Fusion, a company based in Fremont, California.

Europa Fusion intends to pursue a pulsed-power approach to fusion. Pulsed-power reactors fire very large, precisely timed surges of electrical current, short and intense bursts of power, to compress fuel targets until the atoms inside fuse and release energy. The approach differs from magnetic confinement, which holds hot plasma in magnetic fields, and from laser-driven inertial designs, the routes most often associated with recent fusion headlines. The company's technical bet therefore sits on pulse timing and target compression rather than on steady-state plasma confinement.

According to a source familiar with his thinking, Nardelli favors a simulation- and data-heavy approach over hardware-focused testing. That places machine learning models, high-fidelity simulation and experimental data at the center of the development cycle, with physical test campaigns serving to validate what the models predict.

Pacific Fusion, where Nardelli most recently worked, broke ground on a demonstration facility in New Mexico.

In my view, the most useful way to read this funding is through the question of where the hard constraints in pulsed-power fusion actually sit. Fuel-target compression depends on pulse shaping, timing jitter (small variations in when each pulse fires) and the response of materials under extreme conditions. Each is expensive to probe with hardware alone. A team that leans on simulation first is betting its models can narrow the design space before each costly test. Whether that bet pays off will depend on how closely simulated physics tracks what the machines do, a question the first experiments will answer more directly than any announcement.

The broader context here is the growing overlap between machine learning talent and energy hardware. Fusion has long been a field of plasma physicists, pulsed-power engineers and materials scientists. Founders with backgrounds in large-scale ML research bring different tools to the same physical problems. Investors who back that combination are wagering that methods which transformed protein structure prediction and climate modeling can also shorten fusion development cycles. That wager is not yet tested in this sector.

For the engineering community, the immediate story is a founder with a documented path from frontier ML labs into a fusion company, and a startup choosing a pulsed-power architecture with a computation-first development strategy. If the approach works, it could make the design iteration loop faster and cheaper. If it does not, the failure will show where simulation fidelity stops being sufficient, which would be useful to a field that still measures progress in experiments per year.