The Squeeze on Series A: How European and UK VCs Secretly Recalibrated Valuation Multiples for GenAI

Read the headline numbers on European and UK Series A rounds for generative AI companies over the past 18 months and you could conclude that little has changed: £20m–£40m post-money valuations are still announced, and the press releases still say “oversubscribed”. Read the term sheets and a different picture emerges. The recalibration has happened, but most of it has moved out of the price and into the structure: how ARR is counted, which revenue earns a multiple, how large the preference stack is, and what happens if the next round comes in lower.

That is the sense in which the squeeze has been quiet: it sits in investment committee models, term sheet definitions and the gap between the headline post-money and what founders receive at a realistic exit. No specific fund, company or deal is described here; every figure that is not a public fact is modelled or a desk estimate.

Fund mechanics: why the recalibration happened

Series A pricing is not set by what a company is “worth”. It is back-solved from what a fund needs. Three pressures have pushed the required numbers in the same direction.

Vintage pressure. Funds raised in 2020–2022 deployed at high entry prices and are nearing their harvesting years with thin distributions. GPs raising successor funds are judged on those vintages, so they are reluctant to repeat the entry assumptions.

DPI over TVPI. Limited partners have shifted attention from paper marks (TVPI, total value to paid-in) to cash returned (DPI, distributions to paid-in). A mark-up from a priced Series B no longer carries much weight in an LP meeting. What matters is whether the position can plausibly be sold, and at what value, within the fund’s life. That pushes GPs to underwrite lower, more probable exit values rather than outlier outcomes.

Reserve ratios. Most Series A funds hold back a large share of the fund (often in the region of 40–60 per cent, desk estimate) for follow-on investment. GenAI companies tend to need more capital before profitability because of compute and inference costs, which means more rounds, more dilution and heavier calls on reserves. A GP who expects to be diluted more, or to spend more defending ownership, must buy more ownership at entry for the same cheque.

A GP sets a target multiple on the cheque, a realistic exit value and the share of the entry stake expected to survive dilution. The maximum post-money valuation follows.

The formula is: maximum post-money = cheque ÷ ((cheque × target multiple ÷ exit value) ÷ retention). Halving the underwritten exit value and adding a few points of expected dilution roughly halves the price a disciplined GP can pay. Reserves change the retention line: a fund that can afford its pro rata in every later round can model 75–85 per cent retention rather than 55–60 per cent, but only by committing more capital per company, which reduces the number of companies the fund can back.

When the model says £44m and the founder’s other offer says £80m, the gap is rarely closed by price. It is closed by terms.

How GenAI multiples were recalibrated

The second shift is in the denominator. A revenue multiple assumes revenue of roughly similar quality across the comparable set. For GenAI application companies that assumption fails in three ways, and diligence teams now adjust for each.

• Gross margin. Classic SaaS businesses commonly run gross margins in the region of 75–85 per cent. GenAI applications that pay for model inference on every user action frequently report materially lower margins, with 40–60 per cent a common range in diligence (desk estimate). Pricing on gross profit rather than revenue corrects for this automatically.
• Pilot and experimental ARR. Enterprise buyers have run many paid AI pilots from discretionary innovation budgets, and annualising a three-month pilot produces ARR that may not renew. Diligence teams now separate committed contracted revenue from pilots and usage without minimums, and discount the latter.
• Inference-cost sensitivity. A company whose cost of goods is dominated by a third-party model provider’s pricing carries a margin risk it does not control. Investors now stress-test margins against inference price changes in both directions, and against customer usage growing faster than revenue on flat-fee plans.

The same multiple is applied to both. The difference sits entirely in what counts as revenue and what it costs to deliver, which is why the recalibration rarely shows up as a public statement about multiples: the multiple did not move, the base did.

LP expectations and the UK capital push

UK pension reforms have, at a high level, been pushing more institutional capital towards private markets. In July 2023 the government announced the Mansion House Compact, a voluntary commitment by a group of defined contribution pension providers to aim to allocate at least 5 per cent of their default funds to unlisted equities by 2030. A further voluntary industry accord followed in 2025 with broader private-markets ambitions. These are commitments of intent, not mandated allocations, and the detail continues to change, so check the current government publications.

The British Business Bank, the government-owned development bank, is a significant LP in UK venture funds, notably through its commercial subsidiary British Patient Capital, and has been positioned as a channel for pension capital into growth-stage funds.

For Series A pricing, what matters is the type of LP this represents. Pension money is patient but return-disciplined, benchmarked against public equities and focused on realised cash. It reinforces the DPI emphasis rather than relaxing it: more capital does not mean looser entry prices if it arrives with tighter reporting.

Instruments: SAFE, ASA, CLN and BSA AIR

Before a priced Series A, or in place of one, many GenAI companies raise on convertible instruments. These postpone the valuation conversation without removing it: the cap and discount become the de facto price, and the unconverted stack becomes a negotiating issue in the next round.

The post-money SAFE, published by Y Combinator, fixes the investor’s ownership at conversion by reference to a post-money cap, so every additional SAFE dilutes the founders rather than earlier SAFE holders. Founders stacking SAFEs often underestimate the cumulative dilution.

The Advance Subscription Agreement is the UK’s closest equivalent, and is used largely because it can be compatible with SEIS and EIS. An ASA carries no interest, cannot be repaid and must convert into shares, with a longstop date by which conversion happens even without a funding round. For SEIS/EIS purposes HMRC has generally expected the longstop to fall within six months of the agreement; at the time of writing, check the current HMRC guidance and advance assurance practice before relying on that.

Convertible loan notes are debt: they accrue interest and can fall due for repayment, which gives noteholders leverage if the company misses its next round. Because they are debt, they do not qualify for SEIS or EIS relief.

France’s BSA AIR (bon de souscription d’actions, accord d’investissement rapide) is a warrant whose subscription price equals the investment amount, converting into shares at the next round on cap and discount terms. It plays the SAFE’s role in a civil-law framework. Tax treatment for French investors depends on the structure, so check with French counsel.

Bridge rounds and extensions

When a GenAI company’s metrics do not support the next priced round at the price its insiders want, the default response is a bridge or an “extension” of the previous round. The common structures are:

• Discounts and caps. Convertible bridges typically carry a discount to the next round price (15–25 per cent is a common range, desk estimate) and often a cap set at or below the last post-money, which is a down round in all but name.
• Warrant coverage. Bridge investors may receive warrants for additional shares, worth a percentage of the bridge amount, as compensation for risk.
• Pay-to-play. Existing investors who do not take up their share of the bridge may have their preferred shares converted to ordinary shares or lose anti-dilution rights. This concentrates protection in those who keep funding.
• Extensions at the old price. Selling more Series A shares at the original price preserves the headline but often adds new preferences or warrants to make the old price acceptable.

The signalling risk is real: a Series B lead will ask why insiders bridged rather than led. Bridges work best when tied to a measurable milestone (for example, converting a defined set of pilots into annual contracts) with enough runway to reach it.

Down-round protections in London and Paris

Anti-dilution. In UK venture documents, including the BVCA model documents, broad-based weighted average anti-dilution is the market starting point. It adjusts the investor’s conversion price by reference to how much new stock is issued at the lower price, across the whole fully diluted share capital. The standard formula is new price = old price × (A + B) ÷ (A + C), where A is shares in issue before the round, B is shares the new money would buy at the old price, and C is shares actually issued.

A worked example (modelled): 10m shares in issue, old price £2.00, a down round raising £6m at £1.50. C is 4m and B is 3m, so the adjusted price is £2.00 × 13 ÷ 14 = £1.857. A Series A investor who put in £12m (6m shares) would convert into about 6.46m shares. Under a full ratchet, the price simply resets to £1.50 and the same investor receives 8m shares. Full ratchets are uncommon and are a warning sign when they appear. In France, anti-dilution is often delivered through “BSA ratchet” warrants issued to investors and exercisable at nominal value if a down round occurs; the economics can be weighted average or full ratchet depending on the drafting.

Liquidation preferences. A 1x non-participating preference (the investor takes either its money back or its pro rata share, whichever is greater) remains the market norm in London and Paris. Participating preferences, where the investor takes its money back and then also shares in the remainder, and multiples above 1x are where the recalibration has shown up.

This is the “structure instead of price” trade-off. A company can keep a headline Series B valuation by accepting a 1.5x or 2x preference, or participation, in return. The table below models what that does to the waterfall.

The pattern is the core of the thesis. At a modest £80m exit, a 2x preference takes £28.7m from founders and team compared with a clean 1x round, while the priced down round costs them nothing, because the Series B investor takes its preference in both cases. At a £300m exit the position reverses: the 2x preference becomes irrelevant because the investor converts, while the down round costs founders £22m. A 1.5x participating preference hurts at both exit values.

Founders who accept enhanced preferences to protect a headline are betting on a large exit and selling protection on the modest one. Most venture-backed exits are modest, which is why investors offer the structure.

What founders should negotiate

• Agree the ARR definition in writing before price is discussed: what counts as committed, how pilots and usage without minimums are treated, and which period the metric uses.
• Present gross margin with an inference-cost sensitivity of your own, including your plan for model routing, caching or provider switching. If you do not model it, the investor will, conservatively.
• Prefer a lower clean price to a higher structured one unless you have strong reasons to expect a large exit. Run the waterfall at three exit values, including one below the post-money.
• Hold the line on 1x non-participating and broad-based weighted average anti-dilution, the market norms in BVCA-style documents. Treat full ratchets, participation and multiples above 1x as price concessions and value them as such.
• Count the convertible stack. Model every SAFE, ASA, BSA AIR and loan note converting at its cap before signing the next one, and check SEIS/EIS longstop compliance for UK instruments.
• Tie bridges to milestones and cap their size, discount and warrant coverage. Avoid a second bridge unless it is the route to a defined outcome.
• Ask what the fund’s model assumes about exit value and retention. A GP willing to share its back-solve gives you something concrete to negotiate against.

The recalibration has been less a collapse in multiples than a migration of risk from headline price into revenue definitions, margin adjustments and preference terms. Founders who negotiate only the press-release number win the visible part of the deal and lose the part that decides what they receive.

This article is general information about market practice and tax schemes, not legal, tax or investment advice. Take professional advice on any specific transaction.