Index Ventures Partner, Martin Mignot: Figma, Scale, Wiz: Inside Index’s Decacorn Factory

Summary

• Index’s 30-year ledger, disclosed on air: $11.5B invested, close to $30B returned, $20B+ still held — and “most of that is concentrated in eight, nine companies” out of 300-400 investments (Revolut, Figma, Wiz, Scale, Datadog, Roblox among them). Martin Mignot’s conclusion: AI-era value concentration is “not that different from before,” so being earliest, largest and most-referenced shareholder in category leaders “is the only thing that really matters.”
• “Beware of gross margin in the early days — that’s a mistake we’ve made a couple of times.” Revolut (a kind of negative gross margin, one of Index’s most controversial deals), Snowflake (he thinks), Deliveroo, and “all the LLM providers” fit the pattern; costs fall with scale, and Harry noted token prices were ~99% cheaper over 18 months. “If that’s the only thing that’s holding you up, in most cases I would totally ignore it.”
• Never turn down a deal on price at early stage — Index “never loses a deal or passes on a deal because of price,” because the industry has underestimated outcome sizes. The real danger is different: raising too much at too high a price pre-product-market-fit, then spending “a year or two reducing the team” — “that is dangerous.”
• The firm’s defining miss is Spotify — passed multiple times because the mediocre Last.fm investment showed “how the sausage gets made” with the labels, despite loving Daniel (likely Daniel Ek). The lesson Mignot now applies: exceptional founder plus real traction → “don’t overthink it.” He’s cycled back to “team, team, team and team” as factor number one.
• Against the barbell meme (megafund asset gatherers vs. boutiques), Index claims a “third way”: $300M seed, $800M venture, $1.5B growth. AUM gathering “can make a lot of sense financially for VCs who do it — I’m not sure it makes so much sense for the entrepreneurs.” Mega-AUM “distracts you and pulls you towards the later stage.”
• LLM exposure is likely Cohere plus a seed in Mistral. Dilution means “the pure venture multiple will likely be lower,” but outcome size and speed preserve absolute returns. Europe does need a sovereign LLM provider (governments may want, or even have to, use local providers), and social-network algorithms “should be public, should be able to be audited — they are utilities, critical infrastructure.”
• Liquidity is deliberately dumb: sell quarterly over three years post-IPO, preset. Index sold Robinhood “quite a bit lower” than today’s price, but schedule analysis showed the systematic approach beat alternatives — “you can never sell only at the top.” Otherwise it holds “pretty much everything until IPO.”
• Revolut, first seen at a Seedcamp demo day, is the masterclass: an FX “insertion point” instead of selling bank-switching, one Lithuanian license passported across the EU, and a founder who “wanted to be bigger than JPMorgan” from the first meeting. On the US, Martin says “I think they will”; Harry says “never bet against Nick.”

Deep dive

1. The ledger: concentrated in eight or nine names

• Mignot’s opening frame — venture is “playing the long game”: commit 10-15 years, “not a career… people shouldn’t join venture for the status.” The last cycle didn’t so much add tourist VCs as institutionalize the asset class, attracting people who choose it “as a career more than as a calling.”
• The numbers, volunteered on air: “We’ve been around for 30 years. We invested 11.5 billion. We’ve returned close to 30 and we still have 20-plus in holdings. Most of that is concentrated in eight, nine companies” — out of 300-400 investments. The roster: Revolut, Figma, Wiz, Scale, Datadog, Roblox — from seven different partners and five different locations, in a firm with “no CEO, no managing partner.”
• That concentration is why he shrugs at AI-era worries about value pooling in 5-10 giants: “I don’t think it’s that different from before, really” — Index has lived it internally for decades. The pattern also produced the classic miss-then-chase: Zendesk, passed at seed and A (“if you look at the memo and the valuation at the time, it’s quite funny in retrospect”), came in later via the growth fund.
• On persistence of returns: Ili at Quantum Light told him the best predictor of future returns is an early investor being on the Midas list. Mignot’s caveat: it’s “a really good list for who was a great investor 10 years ago.” Stebbings goes further — “none of the people on there did the deals that they said they did.”

2. The third way: rejecting the megafund-or-boutique barbell

• Doug Leone’s claim that venture went from high-margin boutique to low-margin commoditized industry gets a “not entirely.” Mignot rejects the two-camp meme: “there is a third way, and Index is in that third way” — enough scale to support founders inception-to-IPO, small enough to keep the trusted, personal relationship. Current funds: $300M seed, $800M venture, $1.5B growth — the right size.
• His judgment on asset gathering is one of the episode’s sharpest: it “can make a lot of sense financially for VCs who do it. I’m not sure it makes so much sense for the entrepreneurs themselves.” Mega-AUM “distracts you and pulls you towards the later stage.”
• Stebbings has changed his mind — with more trillion-dollar companies and few players able to write billion-dollar checks, he thinks the AUM gatherers investing in “your OpenAIs at 300 million and your Anthropics at 60 billion” could earn venture-like returns at scale. Mignot half-concedes: “I’m not sure you’re going to have venture-like returns at 300 billion… can you have amazing returns? Absolutely. The math makes sense.”
• Is seed just the AUM gatherers’ entry ticket to deploying $100-500M at C and D? “That is not our model. I don’t want to comment on their strategy” — Stebbings calls it “such a cop-out” and offers his clubbing analogy (seed is the door fee, the table is the C and D). Mignot’s counter: “every check is high conviction… we want to be as early as possible, become the largest shareholder and become the most valued and most referenced investor.”

3. What he buys: one simple, deeply original insight — plus execution

• The best founders share one trait: “they can come up with a very simple insight — something that sounds very simple but is actually incredibly deep and profound and defensible,” reached by genuine first-principles thinking. Nik at Revolut: FX in large corridors costs nothing, so give it away free as the hook. Will at Deliveroo: “the product is the delivery” — consistently under 20 minutes is the product; “everything else is a distraction.”
• Stebbings’ challenge via Fiverr’s Misha: with time-to-copy collapsing, does unique insight lose value? Concession with a hedge: “it goes down without great execution that comes with it… on its own it’s not enough.”
• Market-timing risk is real: a French delivery startup pre-smartphone had to phone drivers, took 50 minutes, couldn’t scale — “the concept was there but the technology just didn’t follow.” Cowboy and micromobility had the timing right but the model wrong: hardware without software attached means brutal supply chains, a volume game, capital-heavy with worse returns than pure software — and service players fight subsidized public transport. (Stebbings adds Lime’s confession: a 33% monthly break rate at one point.) He still thinks Lime and likely Dott will “be okay.”

4. Beginner’s mindset: the Spotify scar, and why he’s back to “team, team, team”

• Every investment biases you — including the wins. Being early in Revolut made every other fintech look worse (“Revolut can do this, does it better”), costing Index other neobanks — Qonto (likely) “would be a really good example.”
• The canonical case is Spotify, and the nuance matters: it wasn’t a bad prior investment but a mediocre one. Last.fm was “a totally reasonable outcome, but we saw how the sausage gets made in that industry and the power of the labels” — so despite loving Daniel (likely Daniel Ek, who worked at an Index company), a phenomenal product and early traction, “we passed multiple times, and when we wanted to come back it was too late.” Stebbings twists the knife: “Daniel wanted Danny every freaking round.” It’s the miss the whole firm still groans about — “only 148 billion.”
• The distilled lesson, repeated as the episode’s refrain: unique founder, unique insight, real signs of execution — “don’t overthink it.” Overthinking is “pretty widespread in the industry”; the desire to look smart and diligent kills returns. His quickfire echo: Hanno at Personio was “a yes immediately… if I had done that with every investment I would have done a lot better” — and Index ran the analysis: had it said yes to every company that presented at partnership, it would have outperformed itself.
• He’s traversed the famous career curve — team, then markets, then team again: “team, team, team and team,” now ranking team-traction-market. Traction hasn’t cheapened even with commoditized 0-to-$10M ramps, but in AI revenue quality is critical: project-based usage versus workflow-inserted products like Cursor, where “in all likelihood the stickiness is going to be a lot higher even if the numbers don’t show it yet.”

5. Gross margin is a fallacy at seed

• Revolut was “one of the most controversial deals,” and may end up the most successful. Why controversial: a very European product (“FX frankly is not a big topic” in the US), a kind of negative-gross-margin business giving away FX with only thin interchange, burning faster the faster it grew — and Mignot thought Nick at the time wasn’t a natural storyteller and fundraiser.
• The generalized lesson: “Beware of gross margin in the early days. That’s a mistake we’ve made a couple of times” — Snowflake was, he thinks, a similar story, Deliveroo another, and “all the LLM providers were very clear examples.” His rule: “if that’s the only thing that’s holding you up, in most cases I would totally ignore it.”
• The mechanism: early-stage companies should optimize growth and product, not margin, because in pure software the cost curve does the work — Stebbings noted that price per token was ~99% cheaper over 18 months, and infrastructure optimizes with volume, exactly as at Revolut. “If you made your judgment based on gross margin at the price of tokens at the time, you would have missed great companies.” Stebbings: that was precisely his take on Lovable’s gross margin.

6. Never lose a deal on price — but overcapitalization pre-PMF kills

• On Peter Fenton’s “price is a mental trap”: “he’s absolutely right… we never lose a deal or pass on a deal because of price in the early stage.” He rejects the are-we-paid-for-risk frame: “you don’t really have a good sense for the size of the reward” — and the industry has underestimated outcomes, so in hindsight high entry prices might have been a fair reward.
• The hedge: it must stop somewhere. At $200B, “is one trillion still a mental trap? There must be some moment where you get into a slightly different realm” — closer to IPO the outcome distribution narrows and risk-reward becomes computable.
• Pressed, he admits deals priced too high have damaged companies: the dangerous pattern is tons of capital at high valuation before real product-market fit, growth subsidized by the raise, then “a year or two reducing the team, going back to basics… trying to find product-market fit after you’ve grown so much is really, really tricky. That is dangerous” — especially in Europe. Yet Index has done this on some occasions when it found extraordinary founders with revenue or open-source traction.
• Ownership has repriced: 20% was the bar 15 years ago; now the goal is “double-digit ownership at exit — most of our returns have been generated by companies where we own close to or more than double digits at exit.” Seed is deliberately collaborative (“we’re not going to have sharp elbows”); A and B is where minimum ownership is fought for. And on Stebbings’ claim that Series A is the worst place to invest (3-5x price inflection against sub-2x progression): labels are noise — the real stages are pre-PMF, post-PMF, and scale, and “great companies are created at any time in the cycle.”

7. The machine: votes one to ten, and no fence-sitting allowed

• Quorum scales with check size, every office is always represented, and partners vote 1-10 — “you can’t vote five and six.” Average above six and the deal is approved: forced conviction, qualified majority. High-conviction latitude exists — a $5M check “not exactly on the spot, but you can definitely make the deal happen” — grounded in “collective trust in the partner’s judgment” closest to the team.
• Stebbings’ culture-chasm point lands: American founders market brilliantly, French founders say “we’re doing 50 million in ARR… it’s okay.” Mignot: “the answer is yes, by and large” — which is exactly why Index leaves latitude to the local partner when there’s no data to speak for itself. The split stays disciplined: roughly half Europe, half US, one global bar.
• The frothy-market failure mode, told with unusual candor: you run your own analysis, someone offers 2x the price with 2x the money and “incredibly high conviction and speed, and you’re like — do they know something I don’t?” Proximity cuts both ways — seeing how the sausage gets made can make you too negative — and sometimes doing the pro-rata on that external validation “was the wrong call.”
• No elaborate outcome-scenario modeling: “we don’t waste cycles” — sensitivity analysis on the few levers that matter, “much more on the founder than on number crunching.”

8. Revolut from the inside: the trigger, the passport, and never betting against Nick

• Origin: a Seedcamp demo-day pitch — but the real signal was multiple touchpoints in a short window (the pitch, a friend’s mention, using the app, a partner’s intro): “for me that’s a big signal… they’ve really hit a nerve.” He came with a prepared mind, having studied Simple (the first real neobank, sold to BBVA) and met Monzo, hunting for one thing: the trigger that makes people switch bank accounts. Revolut never sold switching — it sold “you’re traveling to Portugal for a stag weekend, you’re going to get fleeced by your bank, why don’t you get a Revolut card?”
• The contrarian masterstroke was going global from day one. Conventional wisdom said banking is local, go deep in one market; Nik’s first-principles view: banking is a digital service — “a single piece of code that works across the globe.” The decisive mechanic: one Lithuanian license passported across the entire EU. Mignot’s policy conclusion: “when you give European founders one unified market to compete on, they can be as big if not bigger than anyone in the world” — hence Index’s involvement in EU Inc. Notably, Mignot says Nick would probably get a banking license earlier if rerunning it — it’s far easier pre-scale.
• Will Revolut (now ~$75B in Stebbings’ framing) crack the US? “I think they will. Never bet against Nick.” What makes him special: he never accepts “that’s how it’s done,” breaks problems down and answers them himself (Dalio (likely) and Bridgewater as the visible inspiration), sustains intensity “over a very long period in very difficult situations,” and scaled ambition from the start — “when we first met he wanted to be bigger than JPMorgan… there’s no law of physics that says it can’t be as big.”
• The same no-limits theme anchors his book pick — Lionel Barber’s Gambling Man on likely Masayoshi Son, who went out to raise $10B and decided mid-flight to raise $100B, “lost it all multiple times but never stopped”: “a lot of the limits are in your own head.”

9. LLM exposure, European sovereignty, and algorithms as utilities

• Index holds likely Cohere and a seed position in Mistral. On whether LLMs are even a good venture product given the dilution: “the pure venture multiple will likely be lower than some other categories — that is clear. The difference is the size of the outcome and, very importantly, the speed at which it gets reached” — so absolute returns stay high, especially deploying a lot of capital. He openly worries about value accruing to OpenAI and Anthropic.
• Does Europe need an LLM provider? Yes: tech sovereignty is “a real notion” — governments and quasi-government entities “may want, or even have to, use local providers,” plus a genuine enterprise market in localization — with the hedge “assuming they are close to the frontier.” And no illusion: “is it going to be smaller than OpenAI? Yes, for sure.” Government’s role: “they should be customers… rather than as investors.”
• On TikTok: allowed, but the bigger conversation is algorithms across TikTok, X and Facebook — “should be public, should be able to be audited by anyone, including independent auditors… they are utilities, they are critical infrastructure for the economy and for political systems.”
• On the talent war against Meta-and-OpenAI pay packages: worry, yes — but ESOP is the startup’s weapon: “if you can tell a good enough story about this future value creation, there is no amount of package that can compete with that.” And the seven-day-a-week discourse? Nothing new — Revolut’s and Deliveroo’s early teams worked nights and weekends; the change is founders now say it openly, which he calls positive: “there’s no mismatch of expectation.”

10. Sell mechanically, hold to IPO — and accept you can’t pick your winners

• Index times neither entry nor exit: post-IPO it sells every quarter over three years, preset, adjusted only at the margin by a four-person exit committee that pairs the deal partner with one who isn’t close. The philosophy: “don’t try to be too smart.”
• The cost is visible — Robinhood, where Index was a very large investor, was sold “quite a bit at a lower price than where it is today.” But the firm’s schedule analysis showed: “if we had taken different schedules, we would have been worse off” across the portfolio — “you can never sell only at the top.” To Stebbings’ Shopify counter (mechanical selling would have surrendered 98%), he concedes only “there are always contra-examples.” Secondaries aren’t taboo — Revolut is 10 years in, near end of fund cycle — but “by and large we hold pretty much everything until IPO”: concentrated winners deserve riding, and “the best price discovery is on the public market.”
• Stebbings’ structural point survives the exchange: winners aren’t instantly obvious, so reserve allocation is often wrong. Mignot, flatly: “Yes — it’s inevitable.” Figma is his own proof — Dylan (likely Dylan Field) came to the CEO retreat year after year with nothing launched (“why are we still inviting him?”), while Danny (likely Danny Rimer)’s conviction was “unparalleled… I don’t think there are many examples in the business of that level of conviction for so long.” Index invested in every round (Greylock led the A — “you don’t invest in every single round of every one of your companies”), and once the product shipped, “the traction was undeniable.”
• The close is the worldview under it all: a self-described techno-humanist — technology is how humans escaped being “eaten alive,” and “every technology will come with its downsides, but then you have more technology to solve the downsides and keep that wheel going. It’s an incredible human adventure.”