Where AI-named companies are forming in the UK
Independent dataset of every active UK company on the Companies House register whose name signals AI. Refreshed monthly from Companies House bulk exports plus the live API. Free, public, methodology open.
Monthly UK AI-named company formations, all time
From the pre-ChatGPT baseline to the post-ChatGPT acceleration. April 2026 set a new monthly record. (Older months undercount companies since dissolved, so the early baseline looks flatter than it really was; the post-2022 rise is the robust signal.) Toggle the "AI agent" search-interest overlay to see how closely public attention tracks the formation curve.
The same story by year, with 2026 on pace
Annual AI-named formations. 2026 is only part-filed so far (solid); at the current run-rate the full year projects to ~4,800 (faded). Read the steepness with care: older years undercount companies since dissolved (we hold the live register), so the raw climb looks steeper than the true births. The rise holds honestly as a share of all formation (about 6x) and in the near-complete recent months.
All UK company formation by primary sector, and where software sits
Every new UK company per month, stacked by its primary SIC code (the principal activity given at incorporation, so each company counts once, no double-counting). Software/IT (blue) rises as a share of all formation, from about 5.7% before ChatGPT to roughly 8% in 2025, a real but modest signal. The bigger driver of the overall surge is retail/wholesale (much of it e-commerce, online-seller and bulk registrations), not AI. Toggle Count vs Share; dashed lines mark major model releases. UK only: the one register where we hold the full company population with SIC codes.
The sectors against each other, and what the models did to them
The same primary-SIC sectors as separate monthly lines (not stacked), so you can compare their trajectories and toggle any sector on or off. Software/IT and retail/wholesale pull away from the pack after ChatGPT and again through 2025. The data runs to the last complete month (May 2026); the dashed markers extend to Mythos (disclosed April 2026) and Fable 5 (released 9 June 2026), the build-up the curve has been leading toward.
The attention wave: what people search, and what the press writes
Two different attention metrics on one chart, both worldwide. Lines are Google Trends search interest (left axis, relative 0-100): "Artificial intelligence" (grey) rose after ChatGPT, but "AI agent" (blue) and "Agentic AI" (green) were near-zero until late 2024 and then surged. The thick amber line is global news-article volume for "AI agent" (GDELT, right axis, raw article counts). Search interest and press coverage move together, and both step up in the same window the monthly company formations do. Read the search lines as timing not magnitude (relative index); the press line is absolute counts. Suggestive context for the formation spikes, not proof of cause.
And it shows up in the company names: the "agent" wave
UK AI-named companies whose name contains the word "agent" or "agentic", by incorporation year. "Agent" has a long baseline, software agents and multi-agent systems are decades-old AI terms, but it jumps to 155 in 2025 (from ~60-90 a year before). "Agentic" is a brand-new coinage: essentially zero before 2024, then arriving in incorporation paperwork exactly as the "AI agent" search and news waves peak. The naming wave tracks the attention wave. (2026 is a partial year.)
Britain is renaming itself into AI
Companies that already existed and changed their name to add "AI", by year the rename was filed. 300 UK companies rebranded into AI in 2025 alone, up sharply from the 2018 baseline. These are existing businesses repositioning, not fresh incorporations. (2026 is a partial year, shown faded.)
Where the rebranders come from
The primary sector of companies that renamed into AI. The majority were already pure software or IT, repackaging under the AI label, but roughly one in three came from outside tech (retail, finance, education, healthcare, legal): the genuine cross-vertical pivots. Tellingly, rebranders are less software-heavy than fresh AI startups (about 50% vs 60% software), so existing-business pivots are more cross-sector than companies founded as AI from day one. "Other" is mostly generic catch-all SIC codes the companies filed rather than a real sector, "other business support" (82990), "other professional/scientific/technical" (74909), "other service activities" (96090) and dormant-company codes, plus a scatter of real estate, R&D, telecoms and construction.
AI-named companies are building for the jobs AI most automates
Each AI-named company mapped to the occupation its product targets (from its website), then matched to that occupation's AI-exposure score in Anthropic's March 2026 labour-market study. Every vertical we can map cleanly lands on a high-exposure white-collar occupation, marketing, programming, financial analysis, medical records, security, averaging ~65% exposure. The supply side (who's building AI-named companies) mirrors the demand side (whose work AI automates). Caveat: a vertical isn't an occupation, so this is a directional bridge; 56% of companies don't map cleanly to one occupation and are excluded from this chart.
Growth versus solo-founders, by year
As UK AI-named company formation multiplied several-fold (about 6x as a share of all new companies) from 2018, the share that are solo-founded rose from ~68% to over 80%. The growth is being driven almost entirely by individuals. Note: the raw bar counts undercount older years, which are depleted by company dissolutions in the live register, so the rising share is the honest measure of growth.
AI is concentrating in pure software, not spreading into industries
Each line is the share of that year's new AI-named companies tagged with a given sector (self-reported SIC codes). The takeaway: pure software climbs from 56% in 2018 to 79% in 2026, AI-named companies are getting more concentrated in software over time, not less. Marketing and advertising is the one non-software field gaining ground (1% to 6-8%), while specialist verticals like healthcare and finance stay tiny and flat despite the technology getting cheaper. So far, agentic tools are not visibly pushing AI into specialist industries. Note: companies can hold several SIC codes, so these are independent shares and do not sum to 100%.
What they file versus what they actually build
The SIC-code view above understates how vertical AI really is, because the code is chosen once at incorporation and rarely revised. Among AI-named companies we can classify from their website, 70.3% file a generic software/IT SIC code, but only about a quarter are genuinely horizontal plays (developer tools, general-purpose infrastructure). The rest are building AI for a specific industry, finance, retail, healthcare, marketing, education, and so on. The vertical AI economy is real; it is just invisible to the official industrial classification. (Website-classified companies only; "unclassified" excluded.)
Under-30 owner share, AI versus all UK
Owners here are people with significant control (PSC), the people who hold the company, not necessarily those who run it day to day. Before ChatGPT, AI owners were OLDER than the average UK owner. After ChatGPT, AI owners are younger, with the crossover around 2023. The 2026 point is drawn faded/provisional: a partial year with incomplete date-of-birth coverage (the dip appears in both series, a coverage artifact), so we do not extrapolate it upward. This is a share, not a count.
Under-30 director (founder) share, AI versus all UK
Directors are the people who run the company, a different population from the owners above. Now measured from the complete officer register (all ~5.7M UK companies, not a sample): founders are getting younger across the whole register, and AI founders are younger still. The under-30 share of new directors is shown for all UK companies, the AI keyword cohort, and the stricter website-confirmed AI subset. The confirmed cohort skews even younger, so the youth signal is not an artifact of name-tag noise. Age thresholds (not birth-year generations) are used here because age is comparable across years. The 2026 tail is partial/provisional.
Top UK postcode areas by AI density
AI-named companies per 1,000 active UK companies. Virtual-office formation agents filtered out. Cambridge beats Oxford. Bloomsbury beats the City.
Where the companies register is mostly an artifact. Where the people are is not.
A registered-office address is a weak signal of where AI is actually built: 76.8% of AI-named company directors list the exact same postcode as their company, and a large share of those are formation-agent virtual offices. So we tested whether the people are more spread out than the office map suggests. They are not. Directors whose address differs from the registered office (the only independent location signal we have) distribute almost identically to those who match it, London ~43% either way. The clustering at specific postcodes is inflated by virtual offices, but the regional concentration in London, the South East, Cambridge and Manchester is real. Caveat: Companies House withholds directors' residential addresses, so these are service addresses, and the regions cover England only.
The UK AI map
Where AI-named company directors are, one circle per postcode area, sized by the number of directors. London (the EC cluster) dominates, but real secondary clusters show up across the UK: Manchester, Birmingham, Bristol, the Oxford-Cambridge-Reading arc, plus Glasgow, Edinburgh, Cardiff and Belfast. Positioned at each area's centroid from director service-address geocodes (same caveat as above: service, not residential, addresses).
Do they survive? About as often as any UK company, just slightly more likely to sit dormant
Outcome of each incorporation cohort, from live Companies House status: still trading, dormant (filed dormant accounts, i.e. registered but not operating), or closed (liquidation, dissolved, struck off). Benchmarked against the whole UK register, two honest findings: closures match the all-UK rate at every vintage, so there is no AI-specific failure wave; and dormancy runs only modestly above the all-UK norm, a few points per cohort (see below), not the multiple a smaller sample first suggested. Read it by cohort, not all-time: recent (2025-2026) companies are too young to show failure or dormancy, and dormancy is undercounted until a company has filed its first accounts.
Dormancy by cohort: AI versus the whole UK register
UK AI is increasingly founded by foreign nationals
Among AI-named companies founded in 2023 or later, 35.7% of owners hold a foreign nationality (64.3% UK). And the foreign share has climbed from about 31% pre-2022 to the high-30s since. Note: this is the nationality of UK-company owners, most are UK-resident, so it reflects who is building here, not migration flows.
Where foreign AI founders come from, and how young they are
Top foreign nationalities among owners of post-2022 UK AI-named companies, with each group's median birth year. India, China and Pakistan lead, and the international founders skew notably younger than British ones (median births around 1991-93, versus 1982 for British founders). The youth story and the international story overlap.
The .ai domain is now a default for AI founders
Domain ending of UK AI-named companies that have a live website. The global .com still leads (46%), but the AI-native .ai (a tiny Caribbean ccTLD repurposed by the industry) is already second at 19%, ahead of the national .co.uk (15%). These UK companies brand for a global, AI-native audience rather than a domestic one, and .dev is almost unused.
What founders are calling their AI-named companies
Beyond the literal "AI", UK founders reach for mythology, sci-fi, and birds of prey. Count of active AI-ecosystem companies whose name contains each word. Athena (wisdom), Nexus, Nova and Phoenix lead, an iconography of intelligence, rebirth, and watching from above.