How Work Actually Gets Done
PsiQuantum has not shipped a product. It has not put a 20-qubit device online for external users. It has raised more than $1.3 billion in committed government and private capital to build a single machine: a million-qubit, fault-tolerant photonic quantum computer, assembled from wafers manufactured at GlobalFoundries' Fab 8 in Malta, New York, and integrated into cryogenic cabinets at sites in Brisbane and Chicago.
The rhythm doesn't follow a sprint cadence. It follows a physics problem that has resisted solution for two decades. Five time zones carry the work: Palo Alto headquarters, a Milpitas fabrication interface, a Chicago deployment site coming online, a UK research outpost at Daresbury Laboratory, and the Queensland construction project at Moreton Bay Central. Each site owns a layer of the stack — algorithm design in California, wafer-level manufacturing at Fab 8, cryogenic integration in the Midwest, prototype cabinet testing in the UK, and full-system assembly in Australia. The fusion-based architecture dictates the sequence, not quarterly planning rituals.
Founders Jeremy O'Brien, Terry Rudolph, Peter Shadbolt, and Mark Thompson built the company on a bet that semiconductor manufacturing could solve the scaling wall that stopped every other photonic approach. That bet sets the tempo. When GlobalFoundries processes a wafer run for the Omega chipset, the feedback loop runs at foundry speed: months per cycle, not weeks. When DARPA's Quantum Benchmarking Initiative demands evidence of error thresholds, the timeline is contractual. A nearly A$1 billion Australian package, a $125 million DARPA award, and a CHIPS Act letter of intent for up to $100 million create hard milestones that cannot slip without losing sovereign funding.
The founding team makes the technical calls. O'Brien, as Executive Chairman, has described the pre-founding period as twenty years of university research to "figure out a path whereby we could leverage the semiconductor industry." That history means questions (whether a fusion measurement threshold of 11.98% against erasure, according to the Nature paper, is sufficient, whether fixed routing eliminates enough configuration error) stay with the physicists who derived the numbers. Engineers at the Milpitas site, where recent board postings show roles like Principal Power Systems Architect ($239–281k, Zero G Talent's data shows) and Staff Hardware Design Engineer ($201–237k, Zero G Talent reported), execute against specifications that originated in Bristol and Imperial College labs.
The operational day splits between two modes. Foundry-facing teams align to GlobalFoundries' lot schedules: design rule checks, tape-out windows, yield reviews. Algorithm and architecture groups follow Construct software platform releases and error-correction simulations that the company's founders have said would take 10^100 years on a conventional GPU but minutes on the target machine. There is no intermediate product to ship. The first customer-facing utility arrives when the million-qubit system powers on in Queensland or Chicago.
The machine that could change the world will be housed in a room that looks like a data center crossed with an ice cream factory. Inside will be some 100 stainless-steel cabinets, each about six feet tall and connected to a supply of liquid helium that keeps them only a few degrees above absolute zero.
That image, from MIT Technology Review's July 2026 profile, captures the work environment: cryogenic infrastructure meeting data-center density, built on a supply chain that "looks like a souped-up and high-precision version of the existing supply chain for silicon photonic chips." The technicians who assemble those cabinets, the physicists who validate the fusion thresholds, and the software engineers who map algorithms to the fixed routing topology all operate on the same critical path. A delay in wafer yield propagates to cabinet integration. A revision in the error model rewrites the compiler.
Weekly rhythms exist, including design reviews, foundry syncs, and government program check-ins, but they serve a single milestone: the utility-scale machine. Scott Aaronson has noted that PsiQuantum "focused squarely on a commercial goal—a computer with one million qubits" while competitors published incremental chip results. That focus means internal metrics track logical qubit yield, not publication count. The hiring data reflects this: fifty-odd salaried roles on the Zero G Talent board cluster in hardware, characterization, and infrastructure, with a median band of $186k, Zero G Talent found. The titles signal ownership (Principal, Staff, Manager), not feature delivery.
The pace is set by physics and foundry physics. The only way to accelerate is to solve the underlying problem faster.
The Bet That Sets the Tempo
PsiQuantum's operating philosophy crystallizes around a single, uncompromising proposition: photonic qubits manufactured in standard semiconductor fabs can reach the target machine faster than any competing approach. That conviction, articulated repeatedly by the founding team, shapes every priority and every resource allocation.
"We just completely rejected that," co-founder and Chief Scientific Officer Pete Shadbolt said of the industry's prevailing strategy — building small, noisy intermediate-scale quantum (NISQ) processors and offering them via cloud access. "So we just went fully 100% committed to realizing a very large machine. And so every investor that we've spoken to, every government that we've spoken to, the deal has basically been that if you give us money, we will spend that money exclusively on the technologies that are needed to get to a very very large system that is actually commercially useful." The refusal to chase near-term revenue or publicity cycles is not rhetorical. It has not. It has not pursued the cloud-access model that rivals such as IBM, Rigetti, and IonQ use to generate early engagement. The founders decided, as Shadbolt put it, "not to do that."
The decision traces back to a physics-first assessment the team made after more than a decade of academic research at Bristol and Imperial College. The consensus: commercially impactful quantum computing requires roughly a million qubits. Anything short of that scale cannot run fault-tolerant algorithms for chemistry, materials, or optimization — the problems that justify the enterprise. "Ten years ago there was sort of defensible optimism that small systems might be useful and we took a risk in not pursuing like 100 qubits or 500 qubits or something," Shadbolt said. "Five, ten years later that feels like a really good bet."
That bet only looked plausible because the photonic approach fits into existing semiconductor infrastructure. "The only reason that we could afford to make that really extreme bet that we're going to go straight to a million qubits... is that our technology fits into the fabs and the OSATs and the contract manufacturers who are building millions of devices for the semiconductor industry every year." The company designs its Omega chipset and manufactures it at GlobalFoundries' Fab 8, a tier-1 foundry running standard CMOS processes. Photons don't feel heat or electromagnetic interference the way superconducting qubits do, so control electronics can sit inside the cryogenic cabinet. Networking uses standard telecom fiber, avoiding transduction. The architecture is modular by design: manufacture wafers, couple chips with fiber, stack cabinets, scale.
The roadmap mirrors how leadership-class supercomputers are built. "We have intermediate milestones uh scattered along the the road map to that big system. And to me it looks a lot more uh like how you build a conventional like leadership class supercomputer... Eventually they put one rack in a building somewhere, bring it up, debug it. Then they go and put 10 racks together... And that's as much as possible. That's really how we're approaching uh the the road map and the staging. And we're clear that none of that is useful until you get to that final very large scale machine." Internal milestones exist (first rack, then 10, then hundreds), but they are treated as engineering checkpoints, not product releases.
Government partnership is not a sideline; it is a structural pillar. The Australian Commonwealth and Queensland governments committed roughly A$1 billion for the Brisbane utility-scale system at Moreton Bay Central, where construction began June 2026. The U.S. side anchors a half-billion-dollar quantum campus in Chicago (Illinois Quantum and Microelectronics Park). "Large-scale quantum computers will prove enormously consequential for governments worldwide," the company states, and it has organized its delivery model around sovereign-site deployments rather than cloud tenancy.
Leadership reflects the founder-led, physics-first orientation. The four academic co-founders retain the technical authority. Victor Peng, appointed interim CEO in February 2026 and confirmed in July, brings semiconductor-operations experience from AMD and Xilinx. Fariba Danesh (COO), Susan Kim (CFO), Dani Kleinman (Chief People Officer), Rob Soderbery (EVP), and Sriram Sitaraman (CIO) round out an executive layer built to translate research-grade physics into industrial-scale delivery.
The cultural consequence is a workplace organized around a single, decade-plus horizon. Resources flow to the million-qubit target; distractions are cut. Employees describe an environment where the technical bar is the primary filter, where progress is measured in wafer yield, coupling loss, and cryogenic integration — not in press releases or cloud-user counts. The trade-off is explicit: no incremental product cycles to celebrate, no near-term customer feedback loops, and a pace that assumes the timeline is fixed by physics and fab schedules, not by quarterly planning.
What the Interview Loop Filters For
PsiQuantum's interview loop does not hide what it values. The topic weights published by InterviewQuery as of April 2025 tell the story: Data Structures & Algorithms leads at 60, followed by Machine Learning at 35, with SQL, Statistics, and Analytics trailing. These are not arbitrary buckets — they map directly to the physics-first stack the company builds. A photonic quantum computer needs control systems that move petabytes of measurement data in real time, error-decoding pipelines that run at clock speed, and simulation infrastructure that lets engineers iterate on chip designs before tape-out. The bar selects for engineers who can write the software that sits two layers below the quantum layer.
Glassdoor's aggregate (two dozen reviews, nearly two-thirds positive, difficulty rated just under three out of five) suggests the process is rigorous but not performative. Candidates face multi-stage loops that blend coding assessments with deep-dive technical discussions. The relatively modest difficulty score compared to FAANG norms reflects a different filter: PsiQuantum cares less about puzzle-solving speed and more about whether you can reason through a noisy, under-constrained physics problem.
The roles themselves confirm the profile. Zero G Talent's board data shows recent postings clustered at the principal and staff level:
| Role | Salary Band |
|---|---|
| Principal Power Systems Architect | $239,500–$281,400 |
| Principal Characterization-Failure Analysis Engineer | $208,300–$244,700 |
| Staff Hardware Design Engineer | $201,400–$236,600 |
| Staff Digital Design Engineer | $201,400–$236,600 |
| Manager Optical Process Engineering | $201,400–$236,600 |
These are not entry-level slots. The company hires people who have already shipped silicon, characterized photonic devices, or built cryogenic control electronics at scale. The 2026 appointments of Rob Soderbery as Executive Vice President and Sriram Sitaraman as Chief Information Officer (both veterans of large-scale hardware and infrastructure organizations) signal that the hiring bar now includes the ability to operate inside a manufacturing-grade discipline, not just a research lab.
What gets filtered out? Engineers who need a spec sheet before they start coding. Physicists who cannot translate a Hamiltonian into a Verilog module. Managers who optimize for quarterly milestones instead of yield curves. The hiring process selects for a specific kind of technical depth: the ability to hold a full-stack mental model from photon generation to error-corrected logical qubit, and the patience to debug it when the yield drops at 3 a.m. in the GlobalFoundries fab. That profile aligns with the founder-led, physics-first mindset — but it also means the median tenure will skew toward people who treat the work as a vocation, not a career step.
The Numbers and the People Who Stay
Glassdoor's aggregate data offers the clearest public window into employee sentiment at PsiQuantum, though it comes with the usual caveats: reviews are self-selected, anonymous, and undated in the summary metrics. As of the latest available snapshot, nearly two-thirds of reviewers said they would recommend the company to a friend. The overall rating breaks down to nearly four out of five for work-life balance, and just over three out of five for both culture and values and career opportunities.
Those numbers sit in a revealing tension. A near-4.0 work-life balance score suggests the day-to-day grind may be more manageable than the "high-intensity environment" described in founder interviews would imply. Yet the lower culture and career-opportunity scores hint at friction around how decisions get made and whether individual contributors see a path forward. In a founder-led organization where long-term quantum advantage trumps short-term milestones, career ladders can look different than they do at growth-stage SaaS companies: promotion may track technical breakthroughs rather than headcount ownership, and that mismatch frustrates engineers accustomed to conventional progression.
The recommend rate of 64 percent lands squarely in the mixed band for hard-tech ventures, comparable to what you see at other deep-science shops where the mission attracts true believers but the pace and ambiguity wear on people who joined for stability. Glassdoor does not surface the tenure or role distribution behind these aggregates in its public summary, so we cannot tell whether the dissatisfaction clusters in hardware bring-up, software tooling, or the corporate functions that support them.
No public source in this research set provides named, dated employee accounts with specific praise or criticism. The company's own newsroom highlights philanthropic investments in South Chicago STEM programs and executive appointments like those leaders in July 2026, but those are outward-facing signals, not internal sentiment. Without attributed quotes from current or former staff (whether on Glassdoor, Blind, LinkedIn, or in reported journalism), any finer-grained portrait would be speculation.
The numbers confirm a workforce that broadly tolerates the conditions but registers meaningful reservations about culture and career trajectory. That aligns with a shop betting everything on a single, decade-scale technical bet. People who thrive on that bet stay; people who need quarterly validation leave. The more than a third who would not recommend the company are likely the latter.
The company has grown to more than 500 people across Palo Alto, Milpitas, Daresbury, Malta, Chicago, and Queensland. That geographic spread, combined with a mission to deliver the target machine, creates a specific profile of who lasts.
People who thrive share three traits. First, they treat ambiguity as a design parameter, not a bug. PsiQuantum's roadmap stretches toward utility-scale systems in Australia and Chicago — projects backed by nearly A$1 billion in Australian government funding, a $125 million DARPA QBI contract, and the CHIPS Act letter. Milestones shift as physics dictates. Engineers who need fixed sprints and clear quarterly OKRs tend to leave. Those who stay build their own structure inside the chaos.
Second, they possess depth in a hard discipline (photonics, cryogenics, semiconductor process integration, quantum error correction) and can translate across boundaries. The Omega chipset integrates photonic components manufactured at GlobalFoundries' Fab 8, then tested at cryogenic temperatures in Daresbury before final assembly in Milpitas or Brisbane. Senior hardware engineers debug yield loss that can trace back to design choices made years earlier; characterization engineers work from wafer maps through cryostat validation.
Third, they accept centralized, founder-driven prioritization. The four co-founders set the technical north star. Victor Peng's recent appointment as CEO, alongside the two executives, signals an operational scaling phase — but the physics agenda remains founder-owned. Decisions on architecture, manufacturing flow, and government deliverable sequencing flow top-down. Contributors who need consensus-driven roadmaps or visible career ladders (career opportunity rated that figure) will frustrate quickly.
Who burns out? Candidates hired for incremental execution include program managers expecting waterfall Gantt charts, software engineers wanting clean API boundaries, and technicians seeking stable shift schedules. The reorgs mentioned in reviews aren't cosmetic; they reflect a company rewiring its org chart every time a new physics result changes the integration plan. The 2025–2026 sprint (Series E close, Construct platform launch, groundbreaking in Illinois and Moreton Bay, DARPA phase advancement) compressed a decade of infrastructure build into eighteen months. People who measure progress in shipped features quit. People who measure progress in reduced loss budgets stay.
The hiring bar selects for the latter. Offers land in the $200k–$280k band for senior individual contributors. That compensation buys intensity. The trade-off is explicit: you work on the hardest problem in computing, with government-scale resources, alongside peers who speak your native technical language. You don't get predictability. You don't get a mapped promotion path. You get a shot at the first useful quantum computer — and the grind that comes with it.
As previously described, the machine will be housed in a cryogenic facility resembling that cryogenic facility, with 100 stainless-steel cabinets cooled by liquid helium to near absolute zero. When it powers on, the first customer-facing utility will arrive — a sovereign machine running algorithms that would take that long on a conventional GPU. The grind continues until it does.
Working in frontier tech? Zero G Talent tracks the openings: see every open PsiQuantum role, browse frontier tech jobs, the companies hiring, and the people building the field.