CFO.

The lab's burn profile had become the dominant feature of its operating reality. Compute commitments dwarfed payroll, model-training spend ran in tranches that did not resemble any conventional opex line, and the structuring of GPU leases sat across treasury, capital markets and operational planning in ways that the existing finance lead — promoted internally — could no longer hold alone. The CEO wanted a CFO who could think with equal fluency about a credit facility against compute and about a research-led organisation's payroll and equity model.

The non-negotiables were specific. Demonstrable bank-side or capital-markets experience structuring large, unconventional balance-sheet exposures, paired with the temperament to operate inside a small senior leadership group at the pace of a research-led company. The comp band was structured to reflect both the seniority of the seat and the asymmetry of the equity opportunity. Geography was preferred near the lab's main hub, with willingness to consider candidates on either side of the Atlantic. The CEO was explicit that they did not want a traditional SaaS CFO, however senior.

The bench for CFOs who can hold a balance sheet of this shape is small. Most senior finance leaders at tech companies have been trained inside a SaaS or marketplace shape, where opex scales with headcount and revenue. The compute-led financial profile of a foundation lab does not resemble that — it more closely resembles infrastructure financing crossed with frontier R&D, and the leaders who have lived inside that combination tend to come from project-finance backgrounds, telecom or energy capital teams, or the small group of CFOs who have already sat inside an AI lab through a similar transition.

Our read was that the shortlist would draw from a wider perimeter than a conventional CFO search. Search firms without practitioner depth in AI tend to surface a familiar list of SaaS CFOs and miss the point of the seat. We built the shortlist across capital-markets-trained operators, infrastructure-finance leaders who had moved into operating roles, and the small number of CFOs who had recently sat inside a comparable lab. We assessed each on the specific structuring work they had personally led — the deals they had closed, the unconventional exposures they had held, and the operating cadence they had run alongside a technical CEO.

The shortlist composition spanned a deliberately wider band of backgrounds than a conventional CFO search. We held the bar on imagination and structuring fluency rather than on direct sector tenure, and we screened heavily for operating temperament inside a research-led founding team.

• Capital-markets-trained operator with prior CFO experience.
• Infrastructure-finance leader who had moved into a CFO seat.
• Former bank-side structuring lead with operating credentials.
• Senior finance leader from an AI lab at a similar inflection.

The hire came from the capital-markets-trained corner of the shortlist — a CFO whose prior seat had involved structuring a balance-sheet-heavy infrastructure rollout, and who had spent the preceding year as an adviser inside the broader AI sector. What made them right was less the pedigree of either chapter than the integration of the two: they could walk the CEO through a compute-financing structure on a whiteboard, and in the next conversation talk to the head of research about the cadence of model spend without flinching. The board chair, who had been the most sceptical of an unconventional profile, was decisive in the final week.

The close moved at the pace of the brief. There were no competing offers; the candidate had been holding two adviser conversations that did not progress to formal processes. The structuring of equity and the framing of the seat's remit took most of the final two weeks. Brief to offer ran eleven weeks, with the offer accepted inside the second of two reference cycles.

“The engagement sharpened a view we had been forming for finance seats inside foundation labs: the work is not a senior version of SaaS finance, and the assessment should not borrow from that template. Candidates who were strong inside the SaaS CFO grammar were structurally underprepared for a balance sheet shaped by compute commitments, and reading them through that grammar would have produced a confident, wrong shortlist. The hire was the candidate whose work history did not, at first glance, match the brief — but whose structuring craft and operating temperament did, in the specific shape this company required.”

— Craig Oliver