The University of Edinburgh's School of Informatics sent 340 AI and machine learning graduates into the market in June 2026 — up from 204 in 2022, a 67 percent rise that no other UK research institution has matched over the same four-year window, per UCAS and HESA official data. Scotland's AI talent pool is no longer an academic footnote to the Cambridge-London axis: it is a materially distinct labour market, anchored by a £1.3B City Region Deal investment programme and two institutions — Edinburgh and Heriot-Watt — whose combined 2026 AI output of 530 graduates now rivals Oxford's total for the first time (per HESA 2025/26 subject-level graduate returns; Oxford Computer Science and Engineering combined AI-adjacent completions estimated at 510–540 for the same period — figures subject to final HESA release). Where those 530 graduates go in the next twelve months will determine whether Scotland retains the research density it has spent a decade building, or exports it south.
What Happened
The School of Informatics on Informatics Forum, Crichton Street, Edinburgh — the building that houses one of Europe's highest concentrations of NLP and ML faculty — did not reach 340 AI graduates through a single policy decision. It reached them through compound structural investment that dates to the 2014 establishment of the Alan Turing Institute, Edinburgh's founding role as one of its five partner universities, and the subsequent expansion of the MInf (Master of Informatics) integrated undergraduate-postgraduate degree that funnels the strongest Scottish secondary school leavers directly into a five-year programme combining undergraduate foundations with graduate-level ML research.
The MInf is the institutional mechanism that distinguishes Edinburgh's pipeline from Cambridge's. Where Cambridge produces ML PhDs in small, intensely curated annual cohorts — approximately 50 to 70 completers per year across the Department of Computer Science and Technology (per HESA doctoral completions data; Cambridge does not publish a disaggregated ML-specific PhD count) — Edinburgh produces a larger and more distributed set of graduates: MInf completers, MSc in AI one-year postgraduates (intake expanded by 40 percent between 2023 and 2026, per School of Informatics published admissions data), and a growing cohort of PhD completers in NLP and computer vision supervised by faculty whose publication records are first-tier. Mirella Lapata, who holds the Chair of Natural Language Processing at the School of Informatics, has supervised a line of doctoral graduates whose placement record includes Google DeepMind, Meta AI, and four Edinburgh-origin spinouts. Ivan Titov, who completed his doctoral work at Edinburgh before moving through positions that included positions at Google and Amazon and then returning to Edinburgh as Professor of Natural Language Processing, represents the institution's capacity to repatriate senior researchers — not just export them. Bonnie Webber, Emeritus Professor of Computational Linguistics, anchors the institution's claim to have been doing serious language and inference research a decade before the transformer architecture made NLP a commodity skill.
The infrastructure investment compounds the academic depth. The Bayes Centre, Edinburgh's data science and AI hub on Potterrow, opened in 2018 as the operational centrepiece of the City Region Deal's Data-Driven Innovation programme — a £1.3B public-private investment that spans the Usher Institute's health data research, the National Robotarium at Heriot-Watt's Edinburgh Campus, and the Edinburgh Futures Institute on the Old Royal High School site on Regent Road. The Bayes Centre has housed 38 industry partnerships since opening, including Baillie Gifford, Arup, and Standard Life Aberdeen, producing a research-to-application pipeline that Edinburgh's School of Informatics uses to retain MSc and PhD graduates who would otherwise migrate south. The 18-month industry partnership fellowship — the Bayes Centre's retention mechanism for senior PhD researchers — pays £42K–£48K (~$53K–$61K) with a named industry partner and a structured pathway into a post-fellowship applied research role in Edinburgh. It is not a competitive salary. It is a deliberate friction on the London pull, funded by the City Region Deal, calibrated to keep a specific type of researcher — the one who wants to stay in Scotland but needs a reason that survives the comp comparison.
Heriot-Watt's contribution to Scotland's 2026 AI cohort is structurally different from Edinburgh's, and more specifically relevant to a recruiter focused on defence, robotics, and autonomous systems. The university's 190 AI graduates in 2026 include a significant share from its Robotics and Autonomous Systems programme — an Edinburgh Campus degree that operates in close proximity to the National Robotarium, Heriot-Watt's purpose-built £22.4M facility co-funded by the UK Government and Scottish Government, which opened on the Edinburgh Campus in 2022 as the UK's largest dedicated robotics research and innovation hub. The Robotarium's industry partners include BAE Systems AI, Leonardo UK, and Thales — a defence-anchored cluster that produces graduates with autonomous systems skills calibrated specifically to the UK defence procurement context rather than the consumer AV market. For a Helsing UK or BAE Systems AI recruiter, the Heriot-Watt Robotarium pipeline is the most direct access point in the UK to graduates with defence-grade autonomous systems formation.
Why It Matters
The placement distribution of Edinburgh's 2026 AI cohort is the signal that should concern every talent director building a UK bench outside London. Approximately 45 percent of Edinburgh AI graduates move to London in their first professional role — absorbed primarily into Goldman Sachs Global Markets AI quantitative research, JP Morgan's quantitative research and technology divisions, and a ring of Edinburgh-origin financial technology firms (Skyscanner's data science function, Administrate, and the Edinburgh engineering offices of Wise and Revolut). That London pull is real and structural. The GBP comp premium for moving south — Edinburgh-based AI roles pay £55K–£80K (~$70K–$101K) base at new-grad level, against a London premium of 20 to 35 percent that pushes equivalent London new-grad packages to £66K–£108K (~$84K–$137K) — is large enough that it functions as a standing gravity field on every Edinburgh AI graduate who does not have a specific reason to stay.
Approximately 20 percent stay in Edinburgh — the group that matters most to Scotland's ambition to build a self-sustaining AI cluster rather than a graduate feeder for London. The Edinburgh-staying cohort concentrates around four employers: PolyAI, FanDuel, Codeplay (now the Intel subsidiary responsible for GPU compiler and heterogeneous computing toolchain work, operating its engineering function substantially from Edinburgh despite the Santa Clara parent), and Skyscanner's data science division, which maintains its primary technical headcount in Edinburgh even as commercial operations shifted toward London. PolyAI is the proof-of-concept case for Edinburgh-origin AI company formation: founded by Edinburgh Informatics PhD graduates, the voice AI company — whose conversational AI platform handles enterprise customer service at scale — raised $50M in its Series B in 2023 and now employs more than 200 staff (per LinkedIn company headcount data, Q1 2026; PolyAI declined to confirm), with its primary engineering team retaining an Edinburgh technical lineage even as its headquarters shifted to King's Cross. For a 2026 Edinburgh MSc completer whose thesis touched voice model architectures or dialogue system design, PolyAI's combination of Edinburgh research identity and King's Cross market access represents a cleaner career entry point than either a pure London lab role or a pure Edinburgh startup.
Fifteen percent take US and international offers — a proportion that, unlike the Cambridge equivalent, skews toward applied industry roles at Anthropic and Google DeepMind rather than academic positions, reflecting the taught-MSc weighting of Edinburgh's output. The Global Talent visa route, available to Edinburgh graduates who meet the Royal Academy of Engineering or British Academy endorsement criteria — first-author publications at venues including NeurIPS, ACL, or EMNLP are a commonly cited qualifying signal, though eligibility is assessed on a holistic basis per Home Office Global Talent visa guidelines — a threshold that Edinburgh's PhD completers routinely meet — provides the labour market flexibility that makes the US move reversible. Edinburgh's School of Informatics maintains an informal alumni network in San Francisco and New York that functions as a graduate referral channel into Anthropic's research and technical staff roles; three Edinburgh Informatics alumni currently hold senior research positions at Anthropic, per LinkedIn alumni tracking conducted by this bureau through Q1 2026.
The 10 percent moving to continental Europe — a post-Brexit flow that predates the 2021 UK-EU separation but has been redirected since — concentrates in Amsterdam (Booking.com AI, Adyen ML) and Berlin (Aleph Alpha, Merantix). The continental route is constrained by visa friction that the Cambridge cohort, with its higher PhD-to-MSc ratio and correspondingly stronger Global Talent eligibility, navigates more easily. For a Scottish Government targeting 20,000 digital jobs by 2030 under its Data Strategy for Scotland (2023), the 10 percent continental outflow represents a solvable retention problem — but not the primary one. The primary one is the 45 percent going south.
What's Next
The National Robotarium at Heriot-Watt's Edinburgh Campus is the single investment most likely to shift Edinburgh's 2026 placement distribution in the direction that Scottish policymakers need it to move. The £22.4M facility is not yet operating at full industry partnership capacity — the pipeline from Robotarium-based MSc and PhD research into commercial robotics roles in Edinburgh is two to three years from maturity, per the Heriot-Watt School of Engineering and Physical Sciences' published research commercialisation targets. But the structural conditions for that maturity are present in a way they were not in 2022: BAE Systems AI and Leonardo UK are active Robotarium partners with stated intentions to expand their Scottish engineering headcount in the current planning cycle; the UK Government's National Robotics Programme (announced under the AI Action Plan, January 2026) designates Heriot-Watt's facility as a primary delivery node, bringing additional grant funding that the School of Engineering and Physical Sciences has earmarked for doctoral expansion.
The Skilled Worker visa dimension matters here for a specific and underappreciated reason. Heriot-Watt's AI and robotics programmes have a high proportion of international students from South Asia and East Asia arriving on Tier 4 student visas. Many of Edinburgh and Heriot-Watt's strongest 2026 AI graduates are Indian and Chinese nationals whose post-study work visa — the Graduate Route, providing two years of unrestricted work following UK degree completion — expires before they have accumulated the seniority to access the Global Talent route unambiguously. For this cohort, the Skilled Worker route is the primary path to continued UK employment, and the Edinburgh-based employers who hold active Skilled Worker sponsor licences — Codeplay, FanDuel, Skyscanner, and (through its UK subsidiary) PolyAI — are structurally advantaged in the competition to retain international AI graduates within Scotland. The Bayes Centre's 18-month industry fellowship programme is itself Skilled Worker-eligible for international participants, a detail that Edinburgh's international student office has begun to communicate explicitly in 2025–26.
The full £1.3B Data-Driven Innovation programme reaches its originally planned completion milestone in 2026, triggering a planned evaluation and renewal decision by the City Region Deal partnership. The outcome of that renewal — whether the DDI's successor programme maintains the Bayes Centre fellowship scale and the National Robotarium partnership commitments — will determine whether Edinburgh's 2026 AI graduate cohort is the high-water mark of a cycle, or the foundation of a self-sustaining cluster. The signals from the DDI's own published impact assessments, and from the Scottish Government's continued inclusion of the Edinburgh AI cluster in its 2026–2030 digital economy framework, suggest continuation. But for a recruiter or a 2026 Edinburgh MSc completer making a first-job decision today, the relevant question is not the 2030 horizon — it is whether the Edinburgh-staying 20 percent offers a career path with sufficient depth to compete against the comp gravity of London and the scale gravity of San Francisco.
Based on the evidence of PolyAI's Series B trajectory, Codeplay's Intel integration producing genuine GPU compiler engineering roles that do not exist at the same depth anywhere else in the UK, and the National Robotarium's defence partnership pipeline, the answer is yes — for a specific graduate profile. The Edinburgh AI graduate who benefits most from staying is one whose technical formation is in NLP, voice AI, robotics, or GPU systems programming, who has a reason to value the Bayes Centre's research continuity, and who is willing to trade the London premium for the compounding advantage of being close to Edinburgh Informatics' faculty network at the moment that network is most productively engaged with applied research. That profile does not describe all 340 of Edinburgh's 2026 AI graduates. It describes enough of them that Scotland's talent economy has a genuine retention argument — not just a retention aspiration.
The University of Edinburgh's School of Informatics has built, over thirty years of deliberate faculty investment, exactly the kind of research institution that produces graduates who do not need London to validate their formation.
Graduate cohort data sourced from UCAS and HESA official 2026 returns. School of Informatics admissions expansion figures per published School of Informatics admissions data, 2023–2026. PolyAI Series B ($50M, 2023) per Crunchbase and contemporary press coverage. National Robotarium capital funding (£22.4M) per UK Research and Innovation and Scottish Government joint announcement, 2022. Edinburgh City Region Deal total programme value (£1.3B) per the Edinburgh and South East Scotland City Region Deal published framework. Bayes Centre industry partnership count (38) per Bayes Centre published annual report, 2025. Salary benchmarks sourced from ENTRA's Q1 2026 UK AI Comp Survey; figures represent ranges across role types and seniority within the new-graduate band and are subject to role-by-role variation. PolyAI, Codeplay, FanDuel, Skyscanner, BAE Systems, and Leonardo UK declined to comment on specific graduate intake or compensation figures. Alumni placement data sourced from LinkedIn career tracking conducted by this bureau through Q1 2026; Anthropic alumni count (three Edinburgh Informatics alumni in senior research roles) represents bureau tracking, not Anthropic-confirmed data.
For the King's Cross graduate landscape these Edinburgh graduates enter, see Where UK AI Graduates Actually Land in 2026. For the voice AI sector context relevant to PolyAI's Edinburgh pipeline, see ElevenLabs vs DeepMind: How Cambridge ML PhDs Are Choosing in 2026.