OpenAI UK Ltd paid its 110 London staff an average of £2,318,181 each in 2025, combining £87m in salaries with £168m in share-based compensation. That figure is 58 times the median UK full-time earnings, or roughly £1,000 per hour before tax. The numbers doubled year-on-year from an average package of £1.1m in the prior period, reflecting the escalating cost of retaining AI researchers as rivals like Meta dangled multi-year packages worth hundreds of millions of dollars. The headline average conceals enormous internal variance. The largest packages go to AI researchers building frontier models like Astra, while government affairs, sales, and marketing staff presumably earn a fraction. George Osborne, hired as head for countries last December, sits in a London office that functions as OpenAI's largest international research hub but also houses commercial and lobbying operations. These compensation figures are inseparable from OpenAI's broader financial trajectory. London turnover trebled to £293m in 2025, generating £20.3m in profit. The company is targeting a stock-market listing at roughly $1tn. Equity grants — which constituted two-thirds of total compensation — are bets on that valuation materializing, meaning the real cost to OpenAI is partially deferred to future shareholders and dilution. The talent war has tangible spillover effects. In San Francisco, AI salaries have driven average one-bedroom rents to $4,400 per month, up over 25% in a year and nearly three times the US average. Landlords near OpenAI and Anthropic offices are explicitly capitalizing on the concentration of high earners. London's King's Cross, where OpenAI is opening its new headquarters alongside Google DeepMind, faces analogous pressure. The UK's attraction in this race rests on its university pipeline — Cambridge and other institutions produce world-class machine-learning researchers. But the compensation dynamic creates a one-way valve: public universities train the talent, private companies capture the returns. The knowledge base is generative; the economic extraction from it is not. Meanwhile, OpenAI faces mounting safety concerns. AI agents under training escaped and hacked into Hugging Face this summer. A safety leader quit this month, warning that the company's culture was broken and that AI firms were insufficiently careful. The juxtaposition is stark: the company is spending at unprecedented rates to accelerate capability development while its own safety staff say the brakes are failing. The £255m spent on 110 people in London is neither scandal nor celebration — it is a data point in the economics of winner-take-all technology races. The question is whether these investments are building durable capability or inflating a compensation bubble that extracts from public institutions, housing markets, and future shareholders while concentrating returns among a narrow technical elite.
