Proxima Fusion Just Raised €411M. The Stellarator Pool Is 66.

On July 7, 2026, Munich-based Proxima Fusion closed a €411 million Series A led by XTX Ventures with Google, RWE, and the EU EIC Fund participating. The round pushes valuation to €2.4 billion and explicitly funds "aggressive hiring" across Munich, Zurich, and Oxford. The problem: the identifiable global stellarator talent pool is smaller than most Series A engineering orgs.

How big is the global stellarator engineer pool, really?

In Refolk's index of professional profiles, 66 people worldwide carry stellarator-specific expertise, and Proxima Fusion already employs 6 of them. That is roughly 9.1% of the addressable planet, before a single euro of the new round hits payroll.

The wider "fusion plasma engineer" pool sits at 39 globally, concentrated at MIT PSFC (3) and PPPL (2), with Greater Boston as the top region. Together, the two overlapping populations that Proxima needs to raid total something like 100 people, distributed across roughly a dozen institutions and four countries.

9.1%

If Proxima triples its stellarator-specific headcount to spend Alpha's €2 billion capex, the round implies roughly €10M of capital per specialist hire. That inverts the usual sourcing math: a €150k retained-search fee to close one senior magnet engineer is a rounding error, yet most fusion startups still budget recruiting like they run a SaaS company.

Why LinkedIn keyword search fails for stellarator hiring

LinkedIn misses this pool because "stellarator" is a research-paper word, not a resume word, and most of the qualified people sit inside government labs whose profiles never say the thing you are searching for. The IPP Greifswald technician who has physically bolted W7-X coils for eight years lists "Max Planck Institute" and "mechanical engineer" on LinkedIn, not "quasi-isodynamic optimization."

Three specific failure modes:

• Institution obfuscation. IPP staff show up as "Max Planck Gesellschaft" or "Max-Planck-Institut fuer Plasmaphysik" with no project tag. W7-X contributors are invisible to a "Wendelstein" text search.
• Title compression. In Refolk's stellarator sample, the most common title is "Mechanical Engineer" (3), not "plasma physicist." Recruiters filtering on PhD physics titles miss the actual bottleneck.
• Language and diacritics. "Technischer Ingenieur" at IPP, "ingénieur superviseur bobines" at CEA Grenoble. English-only Boolean strings skip a meaningful slice of a 66-person universe.

This is the exact gap Refolk closes: you describe the person in plain English ("mechanical engineers who worked on Wendelstein 7-X coils or contributed to STELLOPT") and get a ranked shortlist that pulls from GitHub, arXiv, ORCID, and LinkedIn in one query, not five.

The real bottleneck: magnet mechanical engineers, not plasma physicists

The scarce resource for Proxima is mechanical and materials engineers who can build twisted non-planar HTS coils, not plasma theorists. Wendelstein 7-X essentially closed the physics question the QI-HTS concept rests on; the remaining ten years of work is a manufacturing and structural engineering problem.

Refolk's stellarator sample bears this out. The top title is "Mechanical Engineer" (3). Coil supports, cryostat penetrations, quench detection, joint resistance, and helium plumbing are the day job. A recruiter who keeps sourcing PhD plasma theorists is fighting the last war, and losing the war they are actually in.

What the target profile actually looks like

The archetype is Francesco Sciortino himself: MIT PSFC to IPP to founder. Reproduce that shape:

1. HTS magnet build experience at Commonwealth Fusion Systems, Tokamak Energy, CERN, or MIT PSFC's SPARC program.
2. Cryogenic mechanical design from ITER supplier chains (ASG Superconductors, Bilfinger Noell, Babcock Noell).
3. Direct W7-X exposure via IPP staff, contractors (MAN Energy Solutions built the W7-X vessel), or PhD work at TU München, TU Berlin, or Greifswald.
4. Stellarator optimization code contributions to STELLOPT or SIMSOPT on GitHub.

Where the 66 people actually live

Two towns hold roughly 15% of the world's identifiable stellarator specialists: Greifswald (4) and Munich (2), both on a single Deutsche Bahn ICE corridor. Princeton (2) and Grenoble (2) round out the top four city clusters.

• Greifswald, Germany. IPP's W7-X site. The single largest alumni pond; every serious W7-X operations paper on arXiv is a sourcing index in disguise.
• Munich, Germany. IPP's second campus in Garching, TU München plasma physics, and Proxima's own HQ. Local poaching happens on foot.
• Princeton, New Jersey. PPPL, home of the canceled NCSX stellarator program. Thea Energy sits here too, using planar coils but hiring the same mechanical talent.
• Grenoble, France. The hidden node. CEA/CNRS superconducting magnet cluster plus Renaissance Fusion. Two of 66 profiles, or 3% of the pool. Meaningful when the universe is this small.
• Cambridge, Massachusetts. MIT PSFC (2 stellarator specialists, 3 fusion plasma engineers). Feeds Commonwealth Fusion Systems and Type One Energy.

A single ICE train from Munich to Greifswald passes through 15% of the world's addressable stellarator engineers.

The sourcing channels that actually work

Reconstruct the pool from primary research artifacts, not job-board keyword filters. Six channels return more real leads than LinkedIn ever will:

1. arXiv physics.plasm-ph co-authorship graphs. Pull every W7-X operations paper from 2022 onward. The second, third, and last authors are usually the engineers who did the work; the first author is the PI who will not move.
2. ORCID pulls filtered on "W7-X" or "quasi-isodynamic." ORCID exposes affiliations across time, so you catch the postdoc who left IPP in 2023 and is now a contractor.
3. IEEE Transactions on Applied Superconductivity author lists. The HTS magnet community publishes here. Cross-reference with ITER supplier employee rosters.
4. SOFT and IAEA FEC attendee rosters. The Symposium on Fusion Technology and the IAEA Fusion Energy Conference publish speaker lists. Every serious engineer in this field attends one of them every two years.
5. GitHub PrincetonUniversity/STELLOPT and hiddenSymmetries/SIMSOPT. The two open-source stellarator optimizers. Every non-trivial contributor is a lead. Commit history is a shadow org chart.
6. IPP Greifswald PhD thesis registry. Public. Names, dates, advisors, thesis titles. Cross-match against current LinkedIn employer field and you have a defection map.

The STELLOPT and SIMSOPT tell

STELLOPT and SIMSOPT are the two open-source stellarator optimization codes; if a person has merged a non-trivial PR to either, they are on the short list by definition. There are maybe 40 contributors between the two repos, and the overlap with the 66-person Refolk index is probably 15 to 20 people. That is your first sourcing pass, done in an afternoon.

Who else is hunting the same 66 people

Proxima is not the only bidder. At least four named competitors are drawing from the same well, and Google now sits on cap tables on both sides of the Atlantic.

Sifted reports Google's investment in Proxima is its first European fusion bet; Google previously backed Commonwealth Fusion Systems and signed a future PPA with them. The same HTS magnet engineer can now be recruited by Google-affiliated portfolio companies from both Munich and Cambridge, Massachusetts. Compensation compression between the two continents is the near-term recruiting risk. Expect the Boston bar to walk to Bavaria within twelve months.

What a €411M raise means for compensation math

Proxima's €411M plus a matching €400M from the Bavarian government brings Alpha's committed funding to €800M, with €1.2bn more targeted from a federal German tender later in 2026. Against a stellarator-specific headcount that probably needs to triple, per-hire capital efficiency dwarfs any plausible recruiting spend.

Concretely:

• A €200k base senior magnet engineer, fully loaded at €300k, costs 0.0125% of the €2.4bn valuation per year.
• A €150k retained search to close that person is 0.006% of the round.
• One month of Alpha construction slippage from an unfilled coil-lead role easily exceeds either number.

The mistake most fusion startups make is treating this like a SaaS hiring plan with quarterly headcount adds and standardized comp bands. The 66-person pool does not respond to standardized comp bands. It responds to targeted, patient, senior-led outreach against a named list, which is where a tool like Refolk earns its keep: it turns "find me every W7-X coil engineer who has published since 2022 and is not currently at Proxima" into a query you run in ninety seconds instead of a two-week internal research project.

€10M

The 90-day sourcing plan for Proxima's talent team

Treat the 66-person pool as a named list, not a funnel. In week one, build the full inventory; by week twelve, you should have talked to every person on it.

1. Weeks 1 to 2. Build the master list. Pull STELLOPT and SIMSOPT contributor graphs. Scrape IPP Greifswald PhD thesis registry back to 2015. Pull every W7-X arXiv paper from 2022 to 2026 and extract non-PI co-authors.
2. Weeks 3 to 4. Enrich. Match ORCID, GitHub, and LinkedIn identities. Tag current employer, city, and last publication date. Flag anyone whose employer has changed in the last 18 months.
3. Weeks 5 to 8. Prioritize by movement signal: recent conference talk, recent paper without institutional affiliation, GitHub activity from a new email domain, LinkedIn "open to work" tag on a public repo profile.
4. Weeks 9 to 12. Founder-led outreach. Francesco Sciortino or a technical cofounder sends the first message on every top-quartile lead. Recruiter follow-up only after that door is opened.

The pool is 66 people. Treating it as an inbound funnel is malpractice. Treating it as a named account list, worked personally, is how Proxima spends €411M without stalling on staffing.

FAQ

How does Refolk count the "66 stellarator specialists worldwide" figure?

Refolk's index tags professional profiles by keyword expertise pulled from public sources including LinkedIn, GitHub, arXiv, ORCID, and personal sites. The 66 figure counts individuals whose profile carries a stellarator-specific tag ("stellarator," "quasi-isodynamic," W7-X project involvement, or STELLOPT/SIMSOPT contribution). It is a lower bound on the true pool since not every qualified engineer is indexed, but it is the addressable universe a recruiter can actually contact today.

Why not just hire fusion generalists and train them into stellarator work?

You can, and Proxima will have to. But Alpha's early-2030s deadline does not allow a two-year onboarding curve for the coil-design lead or the cryostat integration engineer. The 66-person specialist pool is the seed team you build the trainable generalists around. Miss on the seed team and the whole plan slides right.

What's the single best channel for finding W7-X alumni who have left IPP Greifswald?

The IPP PhD thesis registry cross-referenced with current LinkedIn employer fields. IPP posts thesis titles, defense dates, and advisor names publicly; matching those names against current employers surfaces every defector to industry, plus their exact topical expertise from the thesis title. A tool like Refolk collapses that cross-reference into a single natural-language query.

Is Grenoble really worth the trouble for two profiles?

Yes. When the total addressable pool is 66, two people is 3% of it, and the Grenoble cluster around CEA and Renaissance Fusion carries deep superconducting-magnet expertise that maps directly onto Proxima's HTS coil problem. French-language sourcing (technical CVs, CNRS internal pages, école doctorale registries) is a small investment for a meaningful slice of a scarce pool.