State capital floods in, seed money drains

The UK government confirmed the Sovereign AI Unit will formally launch on 16 April 2026, backed by up to £500m in government funding and chaired by James Wise of Balderton Capital ¹. The unit will invest £1m to £20m+ per company in direct equity on commercial terms, not grants. Beyond capital, it will bundle GPU compute access (5,000 to 500,000 hours per company), exclusive government datasets, and advanced market commitments that guarantee procurement.

The structure is unusual for a UK government instrument. Six staff were hired by February 2026, with over 70 applications for three Managing Partner roles. Early hires include Josephine Kant (formerly Google and Y Combinator) and Konstantin Sietzy (formerly the AI Safety Institute). The unit will operate with what Wise calls "maximum autonomy" to move at startup speed.

The Balderton chair is structurally significant. It places a major London venture capital firm at the centre of a government programme that will co-invest alongside, or ahead of, private funds. The unit's initial portfolio, expected in May or June 2026, will reveal whether it targets genuinely early-stage AI companies or gravitates toward later-stage, safer bets where private capital already flows freely.

HM Treasury cut VCT income tax relief from 30% to 20% on 6 April 2026, the first reduction since Venture Capital Trusts were created in 1995 ¹. Simultaneously, the Enterprise Investment Scheme (EIS) lifetime company limit was doubled to £24m, and the annual company investment cap rose to £10m. The government projects the combined package unlocks approximately £100m of additional investment per year.

VCTs are the primary institutional vehicle for pooled early-stage investment. Angels use EIS (which retains 30% relief), but VCTs channel capital from retail investors into portfolios of small companies. The government argues investors will migrate from VCTs to EIS. That may hold for individual angel-backed rounds, but VCTs raised approximately £1bn in 2024-25. A 10 percentage point relief cut implies a contraction in new VCT fundraising that structurally smaller EIS angel rounds cannot fully offset.

The timing compounds the effect. Beauhurst data shows UK grant awards at their lowest count since 2016. VCT-funded early stages are the hardest to replace with alternative capital. For a pre-seed spinout needing £80,000 to prove a concept, the pipeline of institutional early-stage money just narrowed at the same moment the government is deploying billions at growth stage.

HM Treasury gave the British Business Bank (BBB) an additional £6.6bn mandate to deploy by 2030, with a new power to lead venture rounds and invest directly in startups at up to £60m per company from April 2026 ¹. A new £4bn Industrial Strategy Growth Capital Initiative targets eight priority sectors including advanced manufacturing, clean energy, defence, and life sciences. Total BBB financial capacity now stands at £25.6bn.

The BBB has historically operated as a fund-of-funds, distributing capital through intermediary managers rather than picking companies. Direct investment at Series B and C scale places government capital in direct competition and collaboration with VC funds. The £60m single-company cap is large enough for meaningful growth rounds; it signals an intent to anchor later-stage deals rather than seed the earliest companies.

The mandate also tasks the BBB with mobilising pension fund and insurance capital into UK growth companies, a structural gap that has left UK founders dependent on foreign investors for large rounds. The BBB's previous track record (£6.3bn in public funding generating an estimated £43bn in gross value added) will be the benchmark against which this expanded role is measured.

Granola, a London AI notetaking platform founded in 2023 by Chris Pedregal and Sam Stephenson, closed a $125m Series C led by Index Ventures and Kleiner Perkins on 28 March 2026 ¹. The round valued the company at $1.5bn, a sixfold increase from $250m in 2025. Granola's 2026 revenue already stands at 2.5 times its full-year 2025 total.

Three years from founding to unicorn. That pace matches Perplexity and Cursor in the United States, companies built on the same template: AI-native SaaS, rapid revenue growth, and product-market fit legible to US investors. Both lead investors in Granola's round are US-headquartered firms.

The pattern raises a question the government's industrial strategy has not answered. Ministers have steered public capital toward advanced manufacturing, quantum, life sciences, and defence. The market is validating AI applied to existing professional workflows, not deep science. Founders face a genuine fork: follow the policy subsidies into hard tech (longer runway, more patient capital now available) or follow the market signal into AI SaaS (faster validation, US investor appetite, and a proven unicorn template).

9fin, an AI-powered debt intelligence platform for financial analysis, closed a $170m Series C led by HarbourVest at the end of March 2026 ¹. Canada Pension Plan Investment Board, Redalpine, Highland Europe, and Spark Capital also participated. Founded in 2016 and operating from London, New York, Hong Kong, and Belfast, 9fin has now raised over $250m total.

Two UK unicorns in one week, and neither is infrastructure, defence, quantum, or life sciences. Both Granola and 9fin are B2B AI workflow tools targeting knowledge workers in professional services. 9fin's route was slower (founded 2016, unicorn 2026) but its investor base is different in character: pension funds and institutional allocators, not pure venture capital. That investor profile suggests a company generating stable, recurring revenue rather than chasing hypergrowth.

The divergence between policy and market is sharpening. 9fin operates in a niche (leveraged finance analytics) that no government programme would think to subsidise. Its success is a reminder that the most durable companies often build in verticals that policymakers overlook, serving buyers who pay because the product saves them time, not because it advances a national strategy.

Innovate UK published a new prospectus in April 2026 replacing open-competition grant programmes with a portfolio management model ¹. The Velocity programme offers ongoing relationship-based account management instead of fixed-term grants. Growth Sector Teams will actively scout and back high-potential deep-tech businesses across six priority areas: advanced manufacturing, clean energy, creative industries, defence, life sciences, and digital technology.

A new High Potential Business Framework evaluates companies on four criteria: team capability, technical breakthrough, talent pipeline, and market readiness. The model is explicitly DARPA-like: proactive identification and sustained backing rather than reactive competitions.

The budget does not match the ambition. DARPA's 2025 budget was approximately $3.7bn (roughly £2.9bn). Innovate UK's total annual budget is approximately £1bn, of which innovation grant programmes are a fraction. DARPA's power comes from absolute scale: it can back 50 speculative projects knowing most will fail. Innovate UK's increased selectivity without a proportional budget rise means the same money reaches fewer companies. Active competitions in April 2026 include Battery Innovation (£25m), Robotics and Autonomous Systems (£38m, DSIT-backed), and Frontier AI Discovery (£2.5m). Companies excluded from the Velocity portfolio have fewer fallback options than under the old regime.

The Ministry of Defence launched a £20m fund in January 2026 targeting startups with no prior MoD experience, an unusual eligibility criterion in a sector that typically requires existing security clearances and supply chain relationships ¹. The fund offers accelerated contracts (revenue-generating, not grants) to fully or majority UK-owned businesses working in AI/ML, data science, robotics, autonomy, and precision weapons.

A new Office of Small Business Growth aims to reduce MoD procurement complexity. The fund sits alongside a £400m ringfence for novel technology and a broader £7.5bn SME spending target. A Dragon's Den-style pitch event connects funded startups with the Defence Investors' Advisory Group for private follow-on capital.

£20m is seed money for a procurement pipeline, not a venture fund in its own right. The signal matters more than the sum. The MoD is building a pathway for dual-use civilian tech companies to enter the defence supply chain, starting small and scaling through contracts rather than grants. For founders in AI and robotics, this is a new revenue channel with a buyer who has a £2.5bn additional SME budget to deploy by May 2028. The contract model (revenue from day one) is more attractive to early-stage companies than traditional grants, which require match funding and impose spending restrictions.

London startups raised £2.14bn across 45 deals in March 2026, up 21% year-on-year ¹. Six late-stage deals captured £1.88bn, or 87.7% of the month's total. Nscale alone accounted for roughly £1.5bn, approximately 70% of everything raised. Excluding Nscale, the remaining 44 deals raised £644m.

AI companies took 88% of capital across 22 of 45 deals, but two-thirds of that AI allocation was a single cheque. At the early stage, 28 deals raised £106m total, with a median deal size of £1.9m. That is healthy deal volume but modest scale.

The distribution tells the story more clearly than the headline. UK AI funding is bifurcated: a handful of mega-rounds at the top, a large field of small rounds at the bottom, and very little in between. A seed-stage founder raising £2m is operating in a different capital market from a growth-stage company raising £200m. The headline number of £2.14bn creates an impression of abundance that the median founder does not experience.

The UK government committed £2bn to Quantum computing via the ProQure programme in March 2026, claiming to be the first country to target large-scale quantum deployment by the early 2030s ¹. Of the total, £1bn is earmarked for procuring large-scale quantum machines, £500m for computing applications, and over £400m for sensing and navigation.

The allocation reveals the government's thesis. Sensing and navigation (GPS-independent positioning, underground mapping, submarine detection) are closer to commercial deployment than general-purpose Quantum computing. The £400m sensing budget is more than three times the £125m for quantum networking and dwarfs the £90m for infrastructure. Partners include Infleqtion, IonQ, Vescent, HSBC, and BT Group, alongside universities Cambridge, UCL, and Glasgow.

For UK quantum startups, ProQure creates a domestic buyer. The programme's projected £200bn economic contribution by 2045 and 100,000 jobs are government estimates, not independent forecasts. The nearer-term reality is a procurement pipeline that gives UK quantum companies a revenue path they would otherwise have to find in the US or Asia, where competing national programmes are already operational.

PISCES (the Private Intermittent Securities and Capital Exchange System) completed its first live trading events in March 2026 ¹. JP Jenkins ran a trading window from 18 to 24 March; the London Stock Exchange's Private Securities Market also completed a first event. The Financial Conduct Authority sandbox that governs PISCES runs until 2030.

PISCES allows intermittent trading of shares in private companies, creating a secondary market for pre-IPO startup equity without requiring a public listing. UK startup employees holding options have historically waited seven to ten years for an exit event. Early-stage investors in UK companies face the same illiquidity. Both groups have had strong incentives to push for US listings, where secondary markets and pre-IPO liquidity mechanisms are established.

The sandbox is early. Two operators are approved; trading windows are intermittent, not continuous. Volume and price discovery will determine whether PISCES becomes a functioning market or remains a regulatory experiment. If it works, it could reduce the pressure for UK high-growth companies to list in New York simply to provide their investors and employees with liquidity.

Beauhurst's 2026 spinout report shows UK grant awards fell to their lowest count since 2016, even as the average grant size rose 10.96% to £423,000 ¹. Fewer companies are receiving larger awards. The total grant pool is not collapsing, but its distribution is shifting toward mid-stage companies that can absorb bigger tickets.

For a university team seeking a small grant to test an early hypothesis, the narrowing is material. Proof-of-concept funding sits below the threshold of most state programmes. The new Innovate UK Velocity model favours companies with demonstrated team capability and technical breakthrough; by definition, the earliest-stage ventures cannot yet demonstrate either. The expanded £40m proof-of-concept fund (up from £20m over three years) partially addresses the gap but does not restore the volume of small grants available a decade ago.

All of this is happening while the government deploys growth-stage capital at record scale through the British Business Bank, the Sovereign AI Unit, and sector-specific funds. Almost all of it targets companies that have already survived the earliest stages. The pipeline that feeds those programmes is quietly thinning at its base.

Tropic Biosciences, a Norwich-based gene-editing company focused on tropical crops, closed an oversubscribed $105m Series C co-led by Forbion and Corteva Catalyst ¹. Temasek, Just Climate, and IQ Capital also participated. The round is one of the few large UK deep-tech raises located outside the London, Oxford, and Cambridge corridor.

Tropic works on gene-editing staple crops (bananas, coffee, rice) for disease resistance and yield improvement. The investor mix reflects the company's position at the intersection of food security and biotechnology: Corteva is a global agricultural sciences company, Temasek is Singapore's sovereign wealth fund with a strong food security mandate, and Just Climate focuses on climate-positive investment.

Norwich's presence in the round is notable. The city hosts the John Innes Centre and the Earlham Institute, two of the UK's strongest plant and genomic science institutions, but the region has not historically attracted venture capital at this scale. Tropic's success suggests that deep-tech companies rooted in genuine research clusters can raise internationally even without a London address, provided the science is strong and the commercial application (food supply resilience) has a clear global buyer.