Two markets, one job title

The split, in one paragraph

If the market feels confusing right now, it is because you are looking at two markets wearing one job title.

One is frozen. Across UK tech the median engineering salary moved 0.24 per cent in the six months to April, which is statistically nothing. The other is running hot: AI role pay moved roughly twelve times faster over the same window, US labs are advertising London engineering roles with £225,000 salary floors, and three quarters of all UK venture money in the first half of 2026 went to AI companies.

You can see the split in whichever dataset you pick up. Payroll data says the median mid-level engineer in London earns about £72,000, up barely 2 per cent on the year. Advertised salaries for London software engineers actually fell 9 per cent this year, after rising 13 per cent the year before. Meanwhile AI and ML hiring is up 88 per cent year on year, and those roles carry a 12 per cent like-for-like premium on verified payroll.

Both stories are true at once. Which is why one engineer is fielding three recruiter messages a day while a friend with a near-identical CV is sending applications into a void alongside eight hundred other people. Neither of them is imagining it.

King's Cross became a gravity well

The single biggest change this year is not a number, it is a postcode. In April, Anthropic signed for a new King's Cross office with room for 800 people. OpenAI answered with an 88,500 square foot lease nearby, its largest research hub outside the US. Around 900 net-new engineering and research seats between them, on the same few blocks as DeepMind, Meta and Wayve.

The prices are not local prices. Every engineering role Anthropic has listed in London carries a base of at least £225,000, and the top of its ranges runs past £600,000. The door swings one way, too: lab retention is unusually high, running around 80 per cent at two years.

Observed clearing prices

These are base salaries at which offers were actually accepted or lost. London, seed to Series B, mid-2025 through mid-2026. All figures plus equity.

One thing that has not changed: advertised bands still understate the top. Across our client base, the gap between the job description range and the number quoted on real calls runs £20,000 to £30,000. Everyone's band is still fiction at the top end.

For the product engineer level specifically, including where deals die and what the offer mechanics look like, see what a product engineer earns in London.

Teams got smaller, more senior, and better paid

The other quiet shift is that startups simply employ fewer engineers than they used to, and pay each one more. Seed-stage startups now average 5.3 people, down from 6.9 two years ago. AI-native startups run about 25 per cent leaner at the same valuation, and within that smaller team they carry 13 per cent more engineers, fewer managers, and noticeably fewer juniors.

At 5.3 people, your next engineer is a fifth of the company. Leaner teams price every hire like it matters, because it does.

The junior market

The junior numbers are brutal. In 2023, tech companies hired roughly three in ten new graduates. Last year it was one in ten. UK graduate vacancies fell by more than a third in twelve months.

If you are mid-level or senior, this is the scarcity underneath your leverage: the industry stopped minting people like you, then converged on hiring exactly you. It also explains a strange feature of 2026. Engineering attrition is at historic lows, around 9 per cent, so fewer people are testing the market at precisely the moment the market pays most for the ones who do.

Equity: the rules moved in your favour in April

The most under-reported comp news of the year is a tax change. From April 2026 the UK dramatically expanded EMI, the tax-advantaged option scheme most startup engineers are granted under.

In plain terms, scaleups that had outgrown EMI, roughly Series B and beyond, can grant it again. If a company told you in 2025 that it was too big for EMI so your options would be unapproved, that is mostly no longer true, and it is worth re-asking.

The tax logic underneath got stronger, not weaker. Between £100,000 and £125,000 the effective marginal rate is still 62 per cent, the thresholds are frozen until at least 2031, and over two million people will cross the £100,000 line this tax year. £10,000 of base in that band is £3,800 in your account. The same value moved into options, pension or an extended exercise window is worth materially more.

What to do about it

If you are an engineer

1. Work out which market you are in before you negotiate. Payroll medians describe people who already have jobs. Advertised bands describe what companies wish they could pay. The clearing prices above describe what profiles like yours actually closed at.
2. The AI premium follows proof, not titles. Hiring managers report a wave of CVs rebadged "Senior AI Engineer" and filter them out in one interview. What moves you across the line is production LLM work: shipped, measured, in front of users.
3. Above £100,000, negotiate structure rather than the number. The 62 per cent band means base is the least efficient thing you can win.
4. Assume the counter, then discount it. Counters still arrive within the week, and the data still says roughly half the people who accept one are back on the market within months. I used to think that stat was recruiter folklore too. It keeps being true.
5. Low attrition cuts both ways. Almost nobody is moving, which means when you do test the market you are competing with fewer people than the vibes suggest.

If you are hiring

• Benchmark against who you are actually bidding against: funded, AI-adjacent startups, with the labs' shadow over your senior band.
• The EMI expansion is the cheapest lever you were handed this year. If you are near Series B and wrote off EMI on headcount or assets, re-check your eligibility before your next offer goes out.
• Price every hire like it matters. At 5.3 people, it does.

Common questions