Hussein Kanji on European Venture Capital and the AI Opportunity

Hussein Kanji, founder of Hoxton Ventures, joins Harry Stebbings to revisit European venture capital, fund construction and the opportunity he sees in AI. Their firms invest in startups, so both speak from the perspective of participants with commercial interests in the market.

Kanji says Hoxton started when Europe had relatively few active venture firms. Today it has many more, but he thinks fewer investors are willing to build a durable company over a long period rather than seek the next valuation increase.

He traces the shift toward momentum investing partly to cheap money and career incentives. A follow-on investment from a prominent firm can make an investor look successful even if no cash has been returned. Stebbings adds that such signals also help firm owners raise money from limited partners.

Kanji says follow-on investors can be useful signals, but venture returns are driven by outliers rather than averages. He thinks a venture investor needs to be contrarian initially, while finding enough subsequent support for the market to recognize and finance the opportunity.

He points to companies such as Google, Facebook, Uber and Netflix as examples of businesses that created or transformed categories. Their models were not always obvious early: investors questioned how Facebook would make money on mobile even around its IPO.

In Kanji's view, too many European investors use a private-equity mindset, focusing on protecting downside rather than pursuing the power-law outcomes that define venture capital. This is his characterization of the market, not a description of every European firm.

Hoxton nonetheless prepares for downside after it has invested. Each quarter the team identifies potential buyers, divisions and individual contacts who might be relevant if a company runs into trouble. Kanji hopes not to use the list, but wants an acquisition path ready.

The first Hoxton fund took 39 months to raise. Early American supporters backed the founders personally despite skepticism about European opportunities, providing roughly $8 million; Kanji says it took much longer to secure the remaining commitments needed to make the fund viable.

A family and another investor ultimately supplied major commitments. Looking back, Kanji wishes he had followed Mike Maples's advice: set a fixed fundraising period, such as 90 days, then invest what is raised and build a record. He acknowledges that the advice can be easier to follow with strong existing backers.

He says the original pitch was hard because many prospective investors did not believe Europe could produce major outcomes and there was little historical evidence to show them. Hoxton remained relatively methodical after closing, typically making four to six investments a year, though the market accelerated in 2021.

Kanji says Deliveroo was the first fund's best realized investment on its initial check, returning about 34 times that check. The fund invested around $1 million in the first round. He initially tried to talk founder Will Shu out of the idea and doubted that restaurant delivery would become mainstream.

After several months Kanji saw Shu's attention to operations, technology for tracking drivers and strong weekly growth. Index entered a larger financing than Hoxton could lead alone, and Shu helped Hoxton participate alongside it. Kanji credits the founder's commitment to solving a problem he personally felt.

Preserving ownership was difficult for a $28 million fund. Kanji recounts a dispute over pro rata rights and says Hoxton participated in the subsequent B and C rounds but not the D. Rights and ownership changed over time rather than remaining at the original level.

Hoxton's funds grew from $28 million to $89 million and then $214 million. Kanji says the firm now puts more capital into its strongest companies, sometimes building toward 15–20% ownership across its first and subsequent checks. It seeks additional allocation without undermining later investors.

He says Hoxton generally values preferred-stock protections and does not use uncapped notes. Financing terms, ownership and fund size all matter when the firm decides how much to reserve for follow-on investments.

Darktrace taught him a difficult lesson about selling. Kanji says the fund did not sell when its IPO lockup expired, despite a higher share price, and sold about a year later at a lower price as the fund approached the end of its life. A portion was distributed in kind, which he personally held until the acquisition.

At the peak, he estimates the Darktrace position could have produced a 10-times net fund outcome; that is a hypothetical peak, not the fund's realized return. He says Hoxton still made substantial money, but now favors a formula: sell one third after lockup, one third six months later and the rest six to twelve months after that.

Deliveroo was different: Hoxton sold after lockup because it judged the valuation fair, and the stock later fell. Kanji says the first fund has distributed multiples of its capital, but notes that even multiples of a $28 million fund are not necessarily large sums for its limited partners.

With returns established, Hoxton has been concentrating capital in its strongest companies rather than prioritizing immediate markups. Kanji says a little over half of the second fund's capital was in its top third of companies, and the third fund was approaching 60–65% in its best holdings. These are portfolio allocations, not realized returns.

The best investments are not always the hottest early on. Kanji says Darktrace faced skepticism and presented opportunities to buy more. He recalls introducing KKR to a $40 million round at a $400 million post-money valuation, when he says the company was generating about $4 million of monthly revenue.

The company offered Hoxton an opportunity to raise a $10 million special-purpose vehicle for part of that round, but it failed to raise the money. Kanji contrasts the $400 million post-money valuation with the later roughly $5.3 billion privatization value and describes the missed allocation as painful; the resulting return he discusses is hypothetical.

He describes an AI drug-discovery company in the second fund that initially attracted little interest. Hoxton invested $1 million, then a $40 million round followed. Because maintaining its original 18% stake would have required a large check from the $89 million fund, its ownership fell to about 13–14%; he believes the lost stake could matter if the company succeeds.

Hoxton later raised special-purpose vehicles for Darktrace. Kanji estimates roughly $35–40 million was invested through them, with realized investor returns generally around 1.5 to 3 times net over one to two years. He cites net IRRs of approximately 66% to 154% for individual vehicles, reflecting the short holding periods rather than a uniform portfolio return.

To reduce attachment to founders when considering follow-ons, Hoxton asks a growth investor who is not involved in the company to examine the numbers. Kanji says independent analysis can help separate the appeal of a relationship from the underlying business performance.

Hands-on support does not always save an investment. Kanji recalls a company that needed about $2 million to turn around: Hoxton put in $1 million assuming others would supply the rest, but the total raised fell short and the company entered insolvency. Another owner later acquired it, and he says he was invited back; the transaction was still being negotiated.

His lesson is that even when an investor can help operationally, rescue work requires enough capital to finish the job. He thinks larger multistage funds may treat smaller checks as options, while seed investors often do more of the difficult board and operating work when a company struggles.

Kanji invokes the early venture industry's hands-on history, including an account of investor Dave Marquardt helping Microsoft before investing. He contrasts that approach with a modern industry where larger funds can find it hard to justify extensive work on a relatively small position.

He offers an example from Hoxton's portfolio: an AI-enabled law firm in which he says the fund owns 20%. He spent substantial time helping the founder work through business problems and says the company later reached a £1 million monthly run rate. His favorable assessment comes from an investor in the business.

Stebbings asks whether a large Series A check accompanied by little support is good for founders. Kanji says it can be sufficient in a rising market, but when conditions deteriorate companies may need an experienced sounding board and operating help. He argues that even eventual successes can experience serious setbacks.

They discuss whether investors should explain why they decline a deal. Kanji is skeptical that detailed feedback to founders outside the portfolio is always productive, although he thinks frank debate with founders in whom the firm has invested is important.

Hoxton's second fund took 28 months to raise. Kanji says a European Investment Fund commitment fell away after the Brexit process advanced, followed by a prospective co-anchor, forcing the firm to restart. British Patient Capital ultimately supplied a large commitment that he estimates at about 40% of the fund.

Concentration in one limited partner is usually discouraged, but Kanji argues that getting a viable fund raised can matter more. He says public institutions have had an important role in European venture, while worrying that too much capital controlled by one government-backed investor can give it excessive market power.

His proposed remedy is more competition among funding institutions. He also questions whether UK pension funds entering venture have enough experienced people to select managers. He warns that simply adding more capital to an ecosystem already flush with money could lead to poor allocation.

Stebbings describes the difficulty large endowments face deploying hundreds of millions of dollars a year into a limited number of strong venture funds. Kanji sees the appeal of an index-like strategy that absorbs large commitments, while arguing that paying too much can undermine the resulting returns.

Kanji says his analysis found a correlation between capital raised and the probability of an exceptional outcome. He estimates that reaching unicorn status takes around $300 million on average and argues that European companies often raise smaller rounds than US peers. These are his estimates and correlations, not evidence that raising more money alone causes success.

He says his data suggests better odds for a $10 million seed round than a $5 million one, while acknowledging that too much capital can also be wasteful. His argument is that a startup needs enough resources to pursue its opportunity, and investors should size checks accordingly.

Hoxton is sensitive to price because price affects ownership, but Kanji says it sometimes considers writing a larger check even when the opportunity is contrarian within the partnership. Rather than automatically making a small bet, the firm asks whether more capital would give a founder a better chance.

Partners can challenge a deal with concrete evidence, such as deteriorating customer cohorts. Kanji values that error correction, but says subjective doubts that cannot be tested should not necessarily block a partner from pursuing a conviction. A partnership, in his view, should permit both scrutiny and room for disagreement.

For a seed fund making around 20 investments, Kanji suggests $150–250 million may be a useful size, depending on check sizes and reserves. He says larger seed rounds and the desire to double down can make an even larger fund appropriate, while emphasizing that portfolio construction differs by firm.

He describes an AI materials-science model investment: Hoxton committed $10 million and a larger round followed with other funds. He regards the outcome as uncertain and says it would be premature to claim a return. The firm generally limits total exposure to a company to around 10% of a fund.

Kanji estimates a 3% probability that a seed investment becomes a unicorn and says a concentrated portfolio therefore requires strong selection. He worries that many small seed funds depend on later investors to provide capital and support, especially when markets are less buoyant.

He says venture investment in Europe grew from roughly $1 billion to $30 billion over the period since Hoxton began. Rather than treating a larger, more competitive market as a threat, he sees room for more durable European firms and wants Hoxton to be one of them. These are his broad market estimates.

Kanji says his firm backs first-time founders and looks for people building what feels like their life's work. In a discussion of founder age, he suggests experience and the ability to focus may make the late twenties through thirties attractive, but presents this as a judgment rather than a rule for individual founders.

He believes some European startups benefit from more hands-on help from investors than their US counterparts may expect. Founders may not always want that guidance when things are going well, he says, but it can become valuable when a problem surfaces.

Kanji distinguishes his concerns about European economic stagnation from his outlook for startup creation. He sees AI as a horizontal technology in which Europe has meaningful research and company-building capacity, citing DeepMind in London and Meta's AI work in Paris. He believes that creates opportunity, not a guarantee Europe will win.

For scaling and financing, however, he argues the US still offers larger funding rounds, a larger market and deeper experience. Hoxton's own thesis is to find strong European companies and help connect them to America; he notes that Darktrace built a substantial US business early.

Kanji worries about European growth, German carmakers' exposure to Chinese electric vehicles and the consequences of energy prices. These are his concerns and predictions. He questions whether policymakers devote too much attention to the London Stock Exchange rather than ensuring promising companies can obtain capital.

Stebbings argues that a strong local liquidity market matters. Kanji replies that investors can buy US-listed securities and the strongest companies can list there. He acknowledges that local listings support professional services jobs, but questions whether location of the exchange is the most important constraint on building companies.

Kanji estimates that US IPO candidates now typically need more than $200–300 million in revenue, much higher than in earlier eras. He notes that private equity and private markets can also provide liquidity. His suggested threshold is a market observation, not an exchange listing requirement.

He says an unnamed portfolio company has a $150–160 million net revenue run rate and could reach $250 million before a potential IPO in about two years. Both the future revenue and listing timeline are projections. He also says he would have preferred Darktrace to list in the US, while being less certain whether Deliveroo would have traded differently there.

Kanji supports removing friction from European incorporation and would welcome an improved London exchange, but says capitalization of ambitious companies should be a higher policy priority. He also uses reservoir capacity and housing as an example of infrastructure planning whose benefits can arrive long after an election cycle.

On tax, he says he is willing to pay income tax on carried interest, which he regards as income for work. His broader request to government is stability: frequent changes make it harder for investors and founders to plan. This is a policy preference from someone directly affected by the rules.

Asked about companies staying private longer, Kanji says he does not know whether the change is ultimately good or bad. He expects the larger private market to persist and says an early investor should still consider selling some shares when offered an exceptional return, whether that opportunity comes early or at an IPO.

Asked about criticism of his comments on hiring women, Kanji says an LP made him rethink Hoxton's responsibility. He describes the difficulty of recruiting experienced women laterally, and says an established firm should invest in developing the next generation. He adds that his small partnership was not yet in a position to do so.

He says he accepts criticism and recognizes that part of it was fair, though he was unhappy that a critic did not meet him after he sought a conversation. His account of the disagreement is his own perspective.

Kanji says he and former partner Rob grew apart over the kind of firm they wanted to build: Hoxton aims to become a large early-stage platform, while Rob prefers a science-oriented deep-tech firm. Kanji says he invested as an LP in Rob's fund and supported the separation.

On AI, Kanji says he changed his mind about whether a seed fund could invest in a foundation model, after backing materials-science company Cusp. He says Hoxton committed a larger initial investment and held about 11%; its eventual outcome remains unknown.

His investing advice is to play the long game and be contrarian, while recognizing that companies still need the market to support them. He believes the zero-interest-rate era weakened diligence and sees similar euphoria in some AI investing.

Kanji wants to back companies with defensible advantages and increasing returns to scale. He worries that many AI products are quickly commoditized and that investors chasing the same opportunity may end up in intense competition. Despite the risk, he thinks investors must participate to learn.

He cannot predict whether an AI winter will arrive or what the market will look like in five years. Looking back at the internet boom, he notes that today's major companies were not necessarily the most celebrated names at the time.

Asked about Nvidia, Kanji focuses on its high net margins and the possibility that customers' own chips or future technology could challenge them. He explicitly says he is not a semiconductor specialist and does not know whether those alternatives will succeed; the margin figures he cites are his characterization.

Given a choice among OpenAI, Anthropic and xAI at the valuations proposed by Stebbings, he picks OpenAI, while questioning how much value model providers can retain if the technology commoditizes rapidly. He says AI could produce substantial consumer benefits without any single provider capturing the majority of the value.

Kanji expects a Trump administration to be more open to smaller acquisitions, a prediction he sees as potentially helpful to seed funds. He closes with an ambition to build Hoxton into a durable firm that can eventually be led by a new generation rather than remain dependent on him.

Stebbings thanks Kanji for returning to the show after nine years. Kanji hopes they can revisit the conversation again in another decade.