Hussein Kanji, Founder @Hoxton Ventures: Why AI Means London Can Compete with the US | E1248

Summary

• Capital raised correlates with the probability of success — Hussein’s data-backed core claim: the average unicorn consumes ~$300M of financing, and doubling a seed round from $5M to $10M roughly doubles the odds of an outlier. Europe’s seed→A→B conversion rates are now “basically on par with the US,” but capitalization from seed→A and A→B is way under — so the continent takes the risk while systematically starving its winners. “If you believe as a seed investor at a million… you should believe at three.”
• Sell in thirds, mechanically — Darktrace taught Hoxton the hard way: IPO’d at £2.50, hit ~£6 by lockup expiry, they held through 2021 euphoria (a would-be 10x net fund) and sold around £3.50 a year later. New formula: a third at lockup, a third at six months, a third at 6–12 months more — “there’s too much human error in this.” Deliveroo, sold at ~£3 at IPO before it fell to £1, shows judgment occasionally wins anyway.
• The costliest miss was an SPV nobody wanted: Darktrace offered Hoxton $10M of KKR’s $40M Series C at $400M post — the company doing $4M/month and quadrupling — and LPs committed zero. It privatized at $5.3B, a would-be net 10x. The later pre-IPO SPVs (~$35–40M, more than fund one itself) returned net IRRs of 66–155.4%, all realized: proximity to inside information is the seed fund’s real edge.
• Multi-stage firms are writing call options, not doing the work — $30–50M checks to “see if you are interesting,” while the seed firm does the board work, the turnarounds, the acquisition-broking. Fine in bull markets; in 2022-style stumbles companies need hands-on help, and Hussein’s lesson from a portfolio insolvency is that heavy lifting requires a capital base: “sometimes it’s not our place… you need to be well capitalized to do this.”
• Optimal seed fund size is now $150–250M, not the $100M Hoxton once targeted — big enough to write $3–5M first checks, hit 15–20% ownership by the second check, and do jumbo seeds (up 6–7x in volume; $5M+ seeds now a fifth of the industry). Hoxton has told LPs to ignore TVPI while it piles 60–65% of fund three into its top third at “slightly depressed prices.”
• Europe doesn’t need more money or more managers — it needs 5–10 dominant superstar firms. The market grew 30-fold ($1B→$30B), VC headcount rose to 35,000 before normalizing, and another pension-fund billion a year “is bad — we already have way too much cash.” Break the EIF into five competing EIFs like the Baby Bells; the LSE obsession is the wrong problem when the US IPO bar is $200–300M revenue and pension funds already invest globally.
• AI is the first horizontal field where Europe is on par with the US — DeepMind in London, Meta’s AI in Paris — unlike the gaming and fintech niches of prior cycles. But the euphoria “reminds me hauntingly of 1995, 1996, 1997”: ZIRP’s sin was relaxed diligence and he sees it again in commoditizing AI deals — “knife fight in a phone booth.” On the big names: Nvidia’s 50% net margin is the question (at 30% “the multiple changes”), and he’d buy OpenAI at 160 over Anthropic at 40 or X at 50 — “real revenue… increasing returns to scale.”
• Trump opens the exit window: JD Vance has signaled the FTC has no business blocking sub-$500M deals — “phenomenal for seed funds” on recycling. Meanwhile venture’s job, said plainly: “the regulator doesn’t want monopolies but we want monopolies… until eventually it has to get broken up because it’s just too darn powerful.”

Deep dive

1. Venture has become a momentum business — and markups still matter, even to the honest

• Hussein’s answer to Harry’s “right to exist” question starts with an admission: “the Venture world does not need yet another fund.” When Hoxton started 11 years ago Europe’s seed funds of record were “Eden and Pond… a bygone” — the world needed a European venture player. Today’s gap is different: “there are not that many old-fashioned venture funds left” — “most of us have become momentum investors… we write the check largely to get the next markup, not to build the long-term durable big company of tomorrow.”
• The mechanism is career design: at the big firms you get promoted by doing a deal that Index, Sequoia or General Catalyst marks up at a premium, then Tiger marks up again — “all of a sudden it doesn’t make a difference if you’ve not made any money, you look like you’ve picked a hot company.” Owners of their own firms “think like business owners, not like employees.”
• His candid heresy on DPI-maximalism: “people are like, oh, DPI is all that matters — it’s not true. If you can show a cohort of companies that have great tier-one ambassadors following on, it is meaningful to LPs.” The predicament: venture returns live in the outliers, so “you have to be a little bit contrarian — and then very quickly, about a year or two later, the world has to recognize that you’re right in order for you to really get credit.”

2. The iconic outcomes are category creators; Europe underwrites like private equity

• The household names — Google, Facebook, Uber, Netflix — “were all mostly brand-new category creators… there wasn’t an Uber before there was an Uber.” These categories are “really fuzzy up until they’re not”: even at Facebook’s IPO people doubted the mobile transition and the monetization model. Europe doesn’t think this way — “people in Europe are largely trained in private equity… I will do the vertical SaaS company because I know I can’t lose money on it… I only need to clear 12, 15, and there’s upside to 750.”
• The market proves his point about consensus: “we just did a bluntly very boring vertical SaaS company — 13 term sheets.” Meanwhile the power-law businesses go begging.
• Downside thinking has its place — after investing, not before. Hoxton runs a quarterly “founder gets hit by a bus” drill (prompted by a founder diagnosed with bipolar disorder late in life who took himself out of commission): for each troubled company, “I’m picking up the phone, I’m calling this person, at this buyer, in this level of the organization” — buyer, division, protagonist all pre-identified. “I hope I never make that call, but I’m mentally prepared for making it.”

3. Fund one took 39 months of begging — raise for time, not size

• Fund one was “a bear”: 39 months to get going on a $28M vehicle. American individuals wrote ~$8M of “we don’t understand this Europe thing, but we like you” money; the breakthrough was a family whose $15M commitment was split $5M to Hoxton and $10M to another fund (“iser” in the audio — name unclear). “We would not have a fund without them.” He was, by his own account, awful at fundraising — “you’re basically selling a product that nobody’s designed to buy.”
• The advice he now gives every emerging manager came from Mike Maples: “do not do a fundraise for size of the fund, do a fundraise for time” — give yourself 90 days, take whatever you get, start investing, put points on the board, come back. He ignored it (Floodgate’s $75M first fund with Yale-class anchors made it feel like “luxury advice”) — “do not do what we did: 39 months to basically do nothing with your life.”
• Fund two took 28 months and included a full reset: the European Investment Fund had committed as anchor, dismissed Brexit risk (“Article 50 hasn’t been invoked”), then Article 50 was invoked and every EIF commitment in the UK was torn up — Seedcamp got one of the first calls. British Patient Capital eventually took 40% of the $89M fund, breaking every concentration rule of thumb. His verdict: “if you’re going to deliver a ton of returns, get to the right size fund” — right size with imbalanced LPs beats wrong size with a tidy base.
• One identity shift worth flagging: an LP — a woman entrepreneur — reframed his controversial hiring post for him. “We’re 11 years old, we manage a $200M fund… we’re one of the establishment now,” so if there’s a shortage of senior women he can’t poach laterally, “the problem falls on my shoulders — I have a responsibility to grow the next generation” (though with only three GPs, “I can’t do that just yet”). Separately, the split with co-founder Rob was growing apart, not rupture: Rob wants a science-oriented deep-tech firm, Hussein wants a generational platform — “I’m an LP in Rob’s fund… I’ll be the first check.”

4. When to sell: mechanize it — the thirds formula

• Darktrace is the scar tissue. It IPO’d at £2.50, traded to £4, and sat around £6 when lockup expired — at peak, “we would have been a 10x net fund on Darktrace” alone. They held (“I was super long-term… and to be fair this was 2021, the market was euphoric”), then sold around £3.50 a year later as fund life ran out, distributed in specie, and Hussein personally held to the Thoma Bravo take-private at roughly £6. It still returned the fund multiple times — but the lesson stuck.
• The fix is a formula: sell a third at lockup expiry, a third six months later, a third 6–12 months after that. “Just make it a formula, because there’s too much human error in this… long-term I was right, but the markets and what you think long-term don’t always map one to one.”
• Deliveroo — a 34x realized on the first ~$1M check — went the other way: they judged it fairly valued and sold around £3 at IPO; it fell to about £1. “We looked really smart for selling Deliveroo on the eve of the lockup.” The honest takeaway is that discretion worked once and failed once, which is exactly why he wants the formula.
• The Deliveroo cap table also taught cap-table paranoia: a later round redefined pro-rata rights “basically singling us out” — but the drafters forgot Hoxton had bought common from angels. “We politely didn’t comment on the legals,” then exercised anyway when told they couldn’t. They followed through the B and C, skipping the D when DST came in.

5. The $10M SPV nobody wanted — and the ones that printed

• The biggest regret in fund one: Hoxton brokered KKR into Darktrace’s Series C — $40M at $400M post, on a company doing $4M a month (up from $1M/month; it started at ~$10K when Hoxton first invested). The company handed Hoxton $10M of the allocation as thanks for the behind-the-scenes work: “we think you will not get rich enough off of Darktrace.” Hussein trudged around Super Return Berlin unable to name the confidential lead — “we raised zero.” Darktrace privatized at $5.3B, revenue $732M. “It would have been a net 10x after fees… it was really painful.”
• Harry’s double-take is worth keeping — “4 million a month, 400 post… that’s ridiculous, you never see that” — and Hussein’s understatement: “it’s a good deal.” The pricing point stands on numbers alone, no founder-vision required.
• The SPV machine eventually worked: in the run-up to the Darktrace IPO Hoxton placed $35–40M in SPVs — more than fund one’s entire size — with the worst returning net 66% IRR and the best net 155.4%, all realized, roughly 1.5–3x net over one-to-two-year holds. The edge, stated plainly: “when you see inside information — you’re close to the company, you know how it’s doing — you see the buying opportunities and you see that they’re fair prices.”

6. Double down aggressively and ignore TVPI

• Across fund sizes of $28M → $89M → $214M, portfolio construction barely changed (historically four to six deals a year, ~20 companies), but concentration behavior did: “we are now super aggressive about doubling down.” Ownership runs 15–20% “pretty consistently,” built between the first and second check — handshake super-pro-ratas, SAFEs on top, “we’ll find a way to put more capital to work when ownership is still really inexpensive.”
• The instruction to LPs at this week’s AGM: “do not pay attention to TVPI for the time being.” Top-third companies now hold just over 50% of fund two’s capital and 60–65% of fund three’s — deliberately deployed at “slightly depressed prices, because you don’t want ridiculous markups on those companies… if you’re right three, five years later, that will make a material difference in DPI.”
• The flagship example: an AI discovery company in fund two, first $1M check written “before the term techbio was coined,” validated by putting Merck’s chief scientist on the phone. The next check, under a year later, was $40M from Bessemer and F-Prime. Hoxton stretched to a $7–8M pro-rata out of an $89M fund but still slid from 18% to 13–14% ownership — “if that company goes where I think it’s going, it could be the iconic company of tomorrow, and that 5% is going to really” [matter]. Note: the best companies were rarely the hot ones — Darktrace was cold for years, “and as a result they were buying opportunities for us.”

7. Multi-stage funds write call options; the seed firm does the work

• The old division of labor — seed writes the check, “the big boys” run the board, hiring, firing, acquisitions — is gone. Now “it’s all call options for them: they will invest into something and see how it plays out… they can write the $30–50M check where it starts to get meaningful, whereas for us it’s always meaningful — so we end up doing all of this heavy lifting.” Whether that’s bad for founders depends on the cycle: in 2021 all you needed was money; in a 2022-style stumble you need someone doing the work, “and if you’re in a call-option business, it’s not worth the venture fund’s time.”
• The war story: a stumbling portfolio company needed $2M; the co-investing mega-fund wrote it off (but kept its board seat to avoid being recapped) and contributed zero. Hoxton put in $1M, raised $1.3M of the $2M — the company turned around operationally but ran out of cash and went insolvent; an acquirer “has been flying with it since.” Two months ago Hussein got a call: “I heard you were such a good board member… I want to give you stock options for free and have you back on the board.” The lesson, in his words: “this hard-work stuff, you need to be well capitalized to be able to do — sometimes it’s not our place.”
• He grounds this in the Computer History Museum’s oral histories of venture’s founders: Dave Marquardt (likely) was Microsoft’s only investor when it was still a partnership riven by Bill–Paul tension, worked a year before being invited in, “he ended up owning 10% of Microsoft, I don’t know.” Hussein did the same with a next-generation AI law firm — a year of free whiteboarding that earned his COO’s lecture (“your time is really valuable”) — then entered at 9, owns 20%, and the company now runs at £1M a month. “This work pays itself back.”
• On Rabois’ line that the best founders don’t need your help: “yes — until there’s a hiccup.” Google aside, even the giants had them — Facebook’s rounds were hard (hence Microsoft on the cap table). His analogy: a growth curve is like a glucose monitor — “you see the trend line and forget the ups and downs, but it’s the downs where you actually need someone around.” And the public-market context is brutal for laissez-faire: in today’s listed SaaS universe, every decile except the first “is both growing and profitable — both, not either-or” — so a private $100M vertical-SaaS that’s only one of the two “has a long way to go.”

8. Capitalization correlates with success — Europe’s core structural flaw

• The episode’s central claim, from his own data-reading: “there is a correlation between how much money goes into a company and what the probability of success is” — the average unicorn takes ~$300M of financing (some do it on $200M). At seed, moving from a $5M to a $10M round roughly doubles the odds of an outlier. Harry pushes back — he’d assumed $3–5M optimal and $10M detrimental — and Hussein concedes only that “too much capital becomes too much of a wash” at some point, not at ten.
• Europe’s paradox: seed→A→B→C conversion rates are now “basically on par with the US,” but capitalization from seed to Series A and Series A to Series B is far below. “People will take the risk but they’ll mitigate it by writing a small check — and it’s weird, because the inverse should be true: if you believe at a million, you should believe at three… there is a number where you’re freeing up the capacity of the founder to try and achieve greatness. I don’t think people grok this fully in the European ecosystem.”
• On price: “yes, price-sensitive — because we care about ownership; no when it comes to the check.” For contrarian deals the partnership doesn’t fully get, the move is the opposite of downsizing: “they’re raising three and a half — maybe they should have four or five; go buy an extra few points of equity.” This is also why Tiger’s 2021 index-buying playbook failed, in his telling: it assumed “outcomes are equiprobable independent of how much cash goes in” — pay up and a 5x merely becomes a 3x. It doesn’t work that way.

9. The right seed fund is $150–250M — and Europe needs 5–10 dominant firms, not more micro funds

• Hoxton once thought $100M was the right seed fund; now it’s $150–250M (Harry argues 125): 20 companies at $3–5M checks, doubled for reserves, plus room for jumbo seeds — which Ed Sim’s work shows are up 6–7x in volume, with $5M+ seeds now a fifth of the industry. Case in point: Cusp, a foundational model for material science — Hoxton “took it off the table” with $10M of a round that became $30M once Lightspeed and others begged in, and owns 11%. A $10M check is 10% of a $100M fund; that’s why size matters.
• Cusp is also his stated change of mind: “we are not doing foundational model deals — too expensive, too capital-intensive, not a seed fund’s place — and then Cusp walked in the door.” The bet is binary but bespoke (a Microsoft Research paper two years ago showed AI-designed materials can work), and he’s disciplined about the epistemics: term sheet in March, wire in June, “we’re sitting in December — it’d be foolish for me to say anything.”
• On the proliferation of $75M seed funds: “I’m worried… it’s easy to be the feeder fund for those folks.” Europe’s venture market has grown 30-fold ($1B → ~$30B a year), which he reads as healthy — “I’d much rather play in the bigger market” — but the need is for “five to 10 dominant superstar venture funds in Europe, the way the Bay Area has 10 or 15,” not another micro-cap emerging manager. He names Index, Accel and Sequoia as the incumbent handful (Harry adds Creandum-tier firms), and claims the ambition for Hoxton alongside likely Ailie at Blossom: “some of us are going to make it.”

10. Government money, pension funds, and why the LSE is the wrong problem

• On state LPs he’s lived both sides (EIF lost, BBB gained), and his fix is structural: a single EIF at ~30% of Europe’s aggregate LP commitments gives government market power and non-market terms. “You probably need five EIFs competing with each other — what happened with AT&T and the Bells.” Harry’s counter — force the “disgraceful” UK pension funds off the sidelines — runs into Hussein’s data: UK DC schemes already have 10% in top US tech names and 5% in UK equities; they’re long tech, just not venture. And there is no trained UK LP talent base to deploy it: “it takes money and time to train an LP… no one is talking about this.”
• More money would actively hurt: “say we add another billion a year to European venture — that is bad. We already have way too much cash in Europe.” The industry ballooned from ~10,000 investors to ~35,000 before mean-reverting, and allocators can’t tell exceptional from average. His LP math: a $300–500M annual venture budget is “almost impossible” to deploy well — 20 into each of three or four top names, five emerging managers at 10, and you’ve still got $170M forcing you into “$40 into likely Andreessen, because where the fuck else am I going to put it.”
• The London Stock Exchange fixation misses the point: “it is not hard on a Bloomberg terminal to put a few extra characters and buy a share on a New York exchange.” If our best companies couldn’t list in the US, fine, real problem — they can. The losers from US listings are service providers (Goldman London, Lazard London get the mandates), not pensioners or the corporate tax base. The actual bar: US IPOs now require $200–300M of revenue (Yahoo went public on $10–20M; “those days are gone”). Hoxton has a company two years out, appointing bankers now at $150–160M net revenue run-rate. And the LSE experiment already ran — Deliveroo and Darktrace were its great tech hopes; Hussein isn’t sure Deliveroo would trade differently in the US, but Darktrace “definitely would” — it sat at a huge multiple discount before Thoma Bravo took it out at $5.3B.
• The deeper policy critique is time-horizon: he learned at a Wall Street Journal dinner that a major impediment to UK housing is reservoirs not built 30 years ago — “you don’t get elected because you built a reservoir.” His hypothetical advice to Starmer: stop tinkering with tax rates (“I’ve passed my non-dom; I have no problem paying income tax on carried interest — it’s income”) — “I want stability… if they start tinkering with the stuff I take for granted, it becomes infinitely harder to build these companies.”

11. AI is Europe’s first horizontal shot — but the euphoria smells like 1995–97

• The bull case for London, precisely scoped: Europe’s past strengths — gaming, then FCA-sandboxed fintech (Monzo, Revolut) — were “big niches, but niches.” AI is horizontal, and “with DeepMind down the road in London and Meta running its AI stuff in Paris, we are for the first time in European history on par from a technology-creation perspective with the US in not a niche field. I cannot interpret that any other way than there’s going to be opportunity.” The macro fears are real but separate: he’s “really terrified the German car industry gets wiped out by China[’s EVs]” and “petrified about UK growth stagnating” — yet scale-path still runs through America (“the rounds are bigger… Darktrace made more money in America than the UK from the early days”). Hoxton’s standing thesis: find the best here, be the bridge to America.
• On the mania itself: ZIRP’s biggest sin was that “people just relaxed diligence… it was a mess — and I think we’re seeing some of that same stuff in AI.” Firms are buying commoditized lookalikes so they don’t “look foolish missing the next big thing” — his team’s phrase for those markets: “knife fight in a phone booth.” What venture should fund instead, said without apology: “the regulator doesn’t want monopolies but we want monopolies… increasing returns to scale with deep defensible moats.” The vibe check: “it reminds me, hauntingly, of 1995, 1996, 1997.” Will 2025 bring an AI winter? “We’re going to go through something — I don’t know what the something is.” Dot-com’s household names (Yahoo, Netscape) weren’t the winners; Google and Salesforce came at the tail end.
• His public-market takes are conditional, not calls. Nvidia: the question isn’t revenue growth but net margins that went from 10% to 50% while Apple, Amazon and Meta all build their own chipsets — “if that margin comes back down to even a very good 30%, the multiple changes… you have to be a technologist to do technology investing.” Forced to pick between OpenAI at 160, Anthropic at 40 and X at 50: OpenAI — “real revenue… increasing returns to scale” — but he flags commoditization speed: his old Microsoft boss Kai-Fu Lee’s (likely) Chinese venture has replicated GPT “with a fraction of the compute,” and much of AI’s value may end as consumer surplus, “no one company skimming off enough of the cream.” Still: “if you don’t play, you have no way of knowing where this stuff goes.”
• The exit unlock: does Trump open M&A and IPO markets? “Yes — JD Vance has made very clear the FTC has no business blocking sub-$500M transactions, which is phenomenal for seed funds” on recycling capital. And his ten-year ambition ties the episode together: hand over the reins within a decade — “I want to build a firm, versus a boutique, versus a project. I want the firm to be around.”

Verification Notes

• The other fund name in the $15M split remains unclear in the raw captions (“iser”).