The first half of 2026 is over. The number that defines it: 340,000 net-new AI roles posted across the top 50 global AI employers between January 1 and June 28, a +42% increase against the equivalent H1 2025 period, per ENTRA job board signal aggregation across LinkedIn, Indeed, Greenhouse, and Lever. That is not the number of AI jobs that exist. It is the number of new ones created — in six months. For context, H1 2025 produced roughly 240,000 net-new roles on the same methodology. The acceleration, not just the growth, is the story.
Three structural forces drove the print. First, the frontier lab spending race entered an unconstrained phase: OpenAI committed to doubling its workforce to 8,000 by December 2026, per Financial Times reporting in March, while Anthropic filed for a near-$965 billion IPO that transformed its equity into the most aggressively competitive recruiting instrument in the private market. Second, sovereign AI programs in the Gulf moved from planning to execution: Abu Dhabi's Stargate UAE campus entered its first 200-megawatt operational phase, and G42 converted sovereign capital allocation into active hiring across its portfolio companies. Third, the EU AI Act's August 2, 2026 high-risk enforcement deadline created a Europe-specific demand spike for AI governance and compliance engineering that had no equivalent in H1 2025, generating role categories — model-audit specialist, GPAI compliance architect, AI risk systems engineer — that did not meaningfully exist at this time last year.
Compensation moved in parallel. Senior Research Scientists at Anthropic, OpenAI, and Google DeepMind cleared a median of $680,000 to $1.4 million in total comp in H1 2026, per Levels.fyi and 6figr data through June. That band is 40 percent above the equivalent H1 2025 figure. The equity instrument reset — OpenAI's conversion from PPUs to RSUs, Anthropic's 15.7x valuation step-up, Google DeepMind's one-time retention payments of $400,000 to $900,000 against 24-month cliffs — means the headline comp number understates the actual value on offer. The talent supercycle that began in 2024 did not moderate in the first half of 2026. It accelerated. These are the bureau dispatches from the final week of the half.
United States: OpenAI Crosses 3,800, SF Comp Resets Again
OpenAI ended H1 2026 with more than 3,800 employees — its fastest-ever six-month expansion, adding roughly 1,300 net staff against its January base, per ENTRA US Bureau tracking and headcount data reviewed from multiple aggregators including Unify and Revelio Labs. The company's stated target of 8,000 by December means H2 must produce a comparable rate of growth while simultaneously absorbing the integration cost of a workforce that has scaled faster than any frontier lab in the industry's history.
The comp number that closed the half: senior researchers at San Francisco frontier labs are averaging $680,000 in total compensation, per Levels.fyi data current to June 2026. That figure reflects base, equity, and bonus — and it spans the OpenAI, Anthropic, and Google DeepMind senior-IC bands as a blended average. The high end of the disclosed range, at the staff and principal research levels, clears $1.4 million at OpenAI and $1.05 million at Anthropic. The practical floor for anyone classified as "senior" at a frontier lab in San Francisco is now above $500,000. Hiring managers below that threshold are recruiting from a pool their best candidates have already left.
OpenAI's 1-million-square-foot San Francisco footprint — anchored by the former Dropbox headquarters sublease at 1800 Owens Street, reported by The Real Deal in March — is the physical infrastructure that will absorb H2 headcount additions. The research-to-applied-engineering ratio of those additions will determine whether OpenAI's per-employee productivity converges with or diverges from Anthropic's. At Anthropic's $47 billion annualized revenue run-rate disclosed in its June 2026 IPO filing, on roughly 3,000 to 5,000 employees, the per-head figure sits between $9 million and $16 million — still the highest in the industry by a substantial margin.
Middle East: G42 at 12,000+ as Abu Dhabi Locks In Its Lead
G42 closed H1 2026 with more than 12,000 AI-adjacent employees across its portfolio — a 28% increase from its H1 2025 baseline, per ENTRA Middle East Bureau tracking of public filings, LinkedIn headcount data, and portfolio company disclosures. The portfolio includes Core42, Inception, Presight, and the company's healthcare and energy AI subsidiaries. The Abu Dhabi AI corridor — anchored by G42, MBZUAI, ADNOC Digital, and the incoming Stargate UAE campus — is now the fastest-growing sovereign AI talent hub globally by headcount velocity.
The structural enabler is the UAE's January 2026 Golden Visa revision. The "Scientists and Specialists" track now ties a 10-year residency to a verifiable basic salary floor, decoupled from employer sponsorship. Researchers who leave G42 for a spinout, or move between sovereign entities, carry their residency. That single policy change removed the primary structural disincentive — visa dependency on a single employer — that previously made Gulf AI roles a lesser option for internationally mobile talent.
Compensation at the senior lateral level now reaches $340,000 in purchasing-power-adjusted terms for PhD-level researchers at G42's Inception unit, combining a base in AED with housing supplements and zero income tax. For MBZUAI's Class of 2026 — 140 graduates from 23 countries, the institution's largest cohort — 80% stayed in the UAE within twelve months, per ENTRA tracking. The Abu Dhabi corridor is not just growing. It is retaining what it builds.
Europe: Paris Posts 1,200 Senior Research Hires, Closes the Comp Gap
Paris's five primary AI labs — Mistral, Hugging Face, Kyutai, and the Parisian research teams of Google DeepMind and Meta FAIR — collectively added more than 1,200 senior researchers in H1 2026, per ENTRA Europe Bureau analysis of hiring data from Welcome to the Jungle, LinkedIn, and institutional disclosures. That makes Paris the highest-velocity AI research hiring market in Europe for the half, outpacing London, Berlin, and Stockholm by net senior-IC additions.
Mistral's contribution is the largest and the most structurally significant. The company ended H1 at approximately 280 employees, up 70% year-on-year, with a senior research engineer band that reset 38% above its mid-2025 level to €280,000 base plus €240,000 equity over four years — approximately €520,000 total, or $568,000 at current rates. The equity instrument is BSPCE, the French employee warrant scheme, against a company valued at €11.7 billion after its September 2025 Series C. Arthur Mensch told Le Monde plainly: "L'écart avec les Américains est réel. Nous ne le nions pas." The gap with US frontier labs in nominal terms is real. In purchasing-power-adjusted, tax-adjusted terms — California's 13.3% top marginal rate against France's housing and transit cost structure — it is closing faster than the headline differential suggests.
France Travail, the country's main public employment agency, documented a 67% jump in AI-related job postings between 2024 and 2026. The mix of that growth matters: the largest share is in senior research roles, not junior or applied positions, which indicates that the Paris market is competing for the same narrow PhD-level cohort that Anthropic and OpenAI recruit from. For the first time, that cohort is considering the comparison.
United Kingdom: DeepMind at 3,200, Cambridge-London Corridor Adds 8,400
Google DeepMind closed H1 2026 with more than 3,200 UK-based employees, cementing its position as the largest single AI research employer in Britain. The Cambridge-London AI corridor — defined by ENTRA as the axis connecting King's Cross in London to the Cambridge Science Park cluster — added 8,400 net-new AI roles in H1, more than the rest of the UK combined, per ENTRA UK Bureau analysis drawing on ONS data and job board aggregation.
The King's Cross cluster has a density that no other European city can match at the senior IC level. Google DeepMind at Pancras Square, alongside ElevenLabs, Wayve, Synthesia, and Isomorphic Labs within a five-minute walk, constitutes a recruiting funnel of more than 10,000 AI-trained engineers within London commuting distance. Demis Hassabis said on Stratechery in February: "London remains our research engine. Mountain View is our scale engine. The split is intentional." That split produced 3,200 UK employees and a concentration of active-research output — AlphaFold 3 derivatives, Gemini multimodal work — that remains comparable to the Bay Area's best.
The compensation "London discount" has narrowed from 35 to 40% in the prior decade to approximately 18% on a purchasing-power-adjusted basis, per ENTRA's Q1 2026 recruiter survey. At the pre-IPO equity level — ElevenLabs, Wayve, and Isomorphic Labs all carry pre-liquidity instruments — the discount falls below 10%. Candidates from US frontier labs who are relocating to London are no longer taking a material financial step down. That structural change, more than any single hiring announcement, explains why the Cambridge-London corridor's net-add figure of 8,400 is the number UK hiring managers will cite when H2 budget conversations begin next week.
What to Watch in H2 2026
Three forces will define the second half.
Return-to-office mandates arriving in Q3. OpenAI's three-days-in-office baseline — Monday through Wednesday, per current policy — is established but not uniformly enforced. Anthropic's softer "approximately once a month in SF" expectation for remote employees has been operating case-by-case, per Blind discussion and recruiter-sourced accounts. Q3 is when both companies face the first real test of whether their stated policies hold against a workforce that grew 30 to 70% in six months and includes large numbers of remote hires who joined under pandemic-era flexibility norms. RTO enforcement risk is not evenly distributed: the applied engineering and go-to-market functions added in H1 are more geographically dispersed than the research teams, and they are the ones most likely to friction against a tightening office-attendance expectation.
EU AI Act high-risk enforcement, effective August 2, 2026. The August deadline activates mandatory obligations for covered high-risk AI systems: risk management documentation, bias testing, human oversight requirements, and continuous post-market monitoring. The Digital Omnibus proposal to defer the deadline to December 2027 remains under negotiation and has not been adopted. For labs and enterprises operating in the EU, the compliance build-out — staffed by the AI governance engineers, model-audit specialists, and GPAI compliance architects that did not exist at scale eighteen months ago — accelerates through July. Every enterprise tech company with EU contracts is a new buyer entering the European AI talent market simultaneously. That is a demand spike on top of the research hiring spike Paris and London already documented in H1.
Whether the Abu Dhabi corridor can hold 28% growth through summer. G42's 28% H1 headcount expansion was enabled by a capital allocation environment — Stargate UAE, sovereign fund commitment, Golden Visa policy normalization — that remains supportive. The friction point for H2 is talent supply, not capital. Abu Dhabi is competing for a pool of PhD-level AI researchers that is growing at roughly 15 to 20% annually globally. A 28% growth rate at the corridor level requires either pulling talent from markets that were not previously competing with Abu Dhabi, or retaining a higher share of the MBZUAI pipeline before US and European labs recruit from it. Both are possible. Neither is guaranteed.
H1 2026 produced 340,000 net-new AI roles, a +42% year-over-year print, and a comp reset that moved the senior frontier floor to $680,000 in the US. The question for H2 is not whether the market is tight. It is whether the supply of people who can fill these roles grows faster than the rate at which those roles are being created.
Global Career PlatformFind AI talent. Find your next role.
Booking is hotels. · Airbnb is apartments. · ENTRA is global careers.