Unable to load audio. Please try again later.
Companies that compute with light are hiring like every other chip company: from a handful of employers, in a handful of cities.
Neurophos is looking for a chief engineer. By Zeki Data’s count, there are just over 20 people in Austin with the 15 years’ experience and the advanced degree the job demands. Most of the city’s hardware systems engineers work for AMD, Apple or Intel.
That is a surprising problem for a company whose chip computes with light. Neurophos has raised $110 million to build an optical processor for AI inference, and its pitch is a break with the electronics that Nvidia and its rivals sell. Yet the people it is hunting, analog designers, verification engineers, performance modellers and systems leaders, are the people every conventional chip company hunts. Light changes the physics. It does not change the labour market.
That matters because optical computing is awash with capital, and capital is usually assumed to solve hiring. OLIX, a London start-up building an optical processor, has raised $532 million in 2026 alone, nearly five times Neurophos’s round, and Marvell paid $3.25 billion for the optical-interconnect company Celestial AI. But the money does not appear to have bought OLIX what it most needs. Its technical team of 92 scores 5.9 out of 10 on our measure, against Neurophos’s 9.7 from just 31 people, and the hires it made in the UK were of lower quality than those it is now making in Austin and Toronto. London, our data shows, has few people who have built chips.
The pattern runs through Europe’s other optical-chip companies. Q.ANT, from Stuttgart, recruits almost a third of its staff directly from universities and research institutes, the highest share of the four companies we compared, and opened a US headquarters in Austin in April. Lumai, which draws on Oxford, is advertising its first US roles in San Francisco. On this evidence the research bases of Oxford and Stuttgart produce researchers more than chip builders, and the start-ups that grow out of them are going where the chip builders are.
Once they arrive they find three kinds of scarcity. Analog designers are scarce by geography: the Bay Area has 1,052 of the 3,107 we have mapped, Austin has 263, and Cirrus Logic is the only sizeable local employer. Design verification engineers are scarce by competition: Austin has 917, but Apple, Arm and AMD employ more than a third of them, and Meta and Nvidia are recruiting from the churn at Intel and Qualcomm. Performance modellers and senior systems leaders are scarce outright, with 177 modellers in Austin and those twenty or so candidates for the chief engineer’s job.
The Bay Area holds two to four times as many of the specialists Neurophos needs as Austin does
Share of mapped US specialists located in each metro (%)
Source: Zeki, Neurophos Diagnostic, November 2026
Nor can a start-up wait for people to come loose. Between 6 and 12 per cent of these specialists change employer in a year, and the typical one has been in post for four to six years. The supply that does move is created largely by turmoil at incumbents, which is why the churn at Intel and Qualcomm feeds Apple, Arm, AMD, Meta and Nvidia long before it reaches a start-up.
Neurophos has an advantage here that money cannot copy: it is already in the right city, and it has used it. Its technical staff average 21.6 years of applied experience, nearly twice OLIX’s, and 16 per cent are former founders, against 3 per cent. It is also hiring in Sunnyvale, where the analog designers are. What it lacks is scale. OLIX is reported to be aiming for customer deliveries as soon as 2027, a year before the 2028 that Neurophos has targeted for its first chips, and ten of its 20 open roles are in Austin. In August Neurophos hired a lead recruiter from OLIX, an early sign of how the contest will be fought.
Neurophos’s technical bench scores 9.7 out of 10, but it is a third the size of OLIX’s
Technical team score and headcount, Neurophos and comparator companies
Source: Zeki, Neurophos Diagnostic, November 2026
For anyone investing in or hiring for hard technology, the lesson is unfashionable. The number to watch is not the size of the round but where a company hires, and from whom. A talent cluster cannot be raised in a funding round. It is inherited by whoever chose their city early. The race to build computers that run on light will be decided partly in the laboratory and partly in two or three metros, by whoever persuades a settled engineer to leave Apple.
About this analysis: drawn from the Zeki Diagnostic Report on Neurophos, November 2026, using publicly available profile data for Neurophos, OLIX, Lumai and Q.ANT and for specialists mapped across the United States. Technical bench scores combine applied experience, relevant prior experience, qualifications and entrepreneurial background. Specialist counts are people Zeki has mapped, not complete populations, and skill categories are based on job titles. The full diagnostic also covers recent hires, local employers and open roles for each skill.
Citations: Data Center Dynamics (OLIX funding and timeline); Yahoo Finance (OLIX Series B); Data Center Dynamics (Neurophos funding); The AI Insider (Neurophos timeline); Quantum Computing Report (Q.ANT in Austin); SDxCentral (Marvell and Celestial AI).