Chasing Demand in Early-Stage Venture at Felicis – Aydin Senkut (EP.526)

Supabase

• Aydin Senkut · FelicisBullish

  Supabase's business has grown about 3,000x since Felicis's 2021 peak-valuation investment; valuation is catching up.

  Senkut says Felicis invested in Supabase at the top of the 2021 valuation peak. Since then the company's business has grown almost 3,000x, and its underlying growth has been steady and accelerating year after year. Market valuation took about three years to reflect this. He argues that looking only at valuation would make the company look stagnant, and that strong, accelerating business performance means valuation eventually catches up. He adds that most citizen-developer app companies, including Lovable and Replit, use Supabase as their underlying database, so Felicis benefits from their growth without having backed them.

  “The company's business performance has grown almost 3,000 X since we invested. It took three years for the market valuation to reflect what's been happening.”at 9:49

• Aydin Senkut · FelicisBullish

  Supabase's database demand is surging from AI apps and citizen developers, so it is a foundational winner.

  Felicis invested early when Supabase already had strong demand, much of it from crypto apps and then from AI and citizen-developer apps built on the platform. Aydin argues that database companies have historically become very valuable (he cites Oracle and Snowflake). Citizen developers are growing 10 to 100 times, and AI agents still need an underlying database even when they are not human. He sees this as a foundational layer with a large pool of potential acquirers, which supports the conviction behind the investment.

  “they had so much demand... all the crypto apps are being built on [Supabase], and all the AI and citizen developer apps are being built on [Supabase].”at 43:24

Skill AI

• Aydin Senkut · FelicisBullish

  Skill AI's founders and IP make it the robotics equivalent of Felicis's OpenAI and Anthropic bets.

  Senkut backs Skill AI, a robotics company founded by two Carnegie Mellon professors who were each former Facebook VPs. He cites their status as among the most-quoted people in robotics, their synthetic-data approach, their early use of reinforcement learning, and hires including people from Tesla. He argues that getting the best talent and having two ideas nobody else sees yet, applied well, is hard to do, and that this is why he compares the bet to Felicis's earlier OpenAI and Anthropic investments. The durable-IP point is central: he wants companies whose core IP is hard to copy.

  “This is the equivalent of our OpenAI or [Anthropic] bet in robotics.”at 14:42

Mercor

• Aydin Senkut · FelicisBullish

  Mercor, which supplies training data for AI models, is the fastest-growing company in Felicis's portfolio.

  Senkut says Felicis has done well in the AI wave by backing Mercor, which supplies the data used to improve AI models. He says its business has grown by leaps and bounds and is probably the fastest-growing company Felicis has ever had in its portfolio. He presents this as evidence that the data layer around AI models captures value, and he expects similar data companies in robotics.

  “We have done well in the AI wave, backing companies like Mercor that supply the data to improve these models. That business has grown incredibly by leaps and bounds, probably the fastest growing company ever in our portfolio.”at 15:15

Anthropic

Across the shows: 104 speakers in other episodes, 134 bullish and 86 bearish calls, past 30 days

• Aydin Senkut · FelicisBullish

  Anthropic's growth kept accelerating, defying skeptics who said such scale and acceleration were unlikely.

  Aydin uses Anthropic as an example of growth that consensus doubted. People argued that growth at that scale, let alone accelerating growth, was unlikely. It kept happening. He uses this to argue that for fast-growing AI companies the non-consensus call is whether growth continues and sharpens, not whether it exists.

  “Similar how [Anthropic] surprise people, people said at that scale, getting that growth, let alone accelerating is unlikely, lo and behold, that continued happening.”at 38:19

Venture capital & startups

• Aydin Senkut · Felicis

  Foundational infrastructure companies have produced Felicis's biggest winners, so Felicis prioritizes that layer.

  Senkut looks back at Felicis's major winners and says virtually all of them have been foundational-layer infrastructure companies. Examples include Shopify (commerce infrastructure), Adyen (payment infrastructure), Credit Karma (credit score infrastructure), Semgrep (code security infrastructure), and foundation model companies that became the platform for a part of AI. He concludes that each time Felicis chose a great infrastructure or foundational-layer company, those tended to have the biggest upside.

  “Virtually all of our major fund maker companies have been foundational layer infra companies... Every time we have chosen a great infrastructure or foundational layer company, those tend to be the ones that had the biggest upside.”at 18:10

• Aydin Senkut · Felicis

  Abundant venture capital is inflating valuations, shrinking ownership and concentrating deals in a few companies.

  Senkut says the abundance of capital has become almost a tax. When that much capital chases fewer companies, ownership must decline and valuations rise. Round-to-round valuation jumps have moved from 2x to 10x for great companies. He sees a few companies producing business performance he hasn't seen even in early Google or Shopify days, which he thinks is justified for those few. But capital is concentrating in the same few companies while whole sectors and geographies get no investor interest. This makes finding the right companies early more important, and it is why Felicis holds some less obvious bets alongside the hot ones.

  “The abundance of capital has become a tax almost. When you have that much capital chasing fewer companies, it's inevitable that ownership is going to decline, valuation is going to increase.”at 20:50

• Aydin Senkut · Felicis

  Mega venture funds that grow past about $1B are essentially fighting the public index for basis points.

  Aydin argues that capital is concentrated in fewer companies and that capital-intensive companies require minimum viable scale of about $500 million to $1 billion for a venture fund. Beyond that, he says the fund stops multiplying capital and its returns look more like index-style basis points. He says that at roughly $1 billion, his firm's strategy can still return three, five or seven times capital because companies are reaching scale much faster than before.

  “Once the numbers get to a certain level, you're essentially fighting the public index. You're fighting for some BPS rather than multiples of capital.”at 52:42

AI infrastructure & data centers

• Aydin Senkut · Felicis

  AI infrastructure and robotics are trillion-dollar markets, with early companies landing billion-dollar contracts.

  Senkut describes a new world of 'trillion dollar markets with billion dollar contracts.' He says the demand for AI infrastructure and robotics is well documented at around a trillion dollars, and a few companies are already winning contracts, letters of intent or interest worth $100 million to a billion dollars. Such companies reach billion-dollar revenue faster but with more concentration. They need years of capital investment before the platform is ready, and they are complex hardware rather than software that can be built in a few months. He says this makes them capital intensive, that the scale and speed of scaling can be an order of magnitude bigger, and that Felicis may need to reserve for more than one round.

  “We are in a new world. I like to call it trillion dollar markets with billion dollar contracts. You have these areas, whether it's robotics or AI infrastructure, the demand is well documented to be around trillion dollars.”at 19:12

• Aydin Senkut · Felicis

  Inference and neocloud firms funding their infrastructure with debt pay much higher interest, which hurts economics.

  Aydin says many inference and neocloud companies are building their infrastructure on debt, and that this means paying about 10% higher interest rates. He argues that financing costs matter a lot when a company's margins are otherwise the key differentiator, so debt-funded AI infrastructure operators face a structural disadvantage against better-capitalised or higher-margin rivals.

  “A lot of these inference and neocloud companies, they're building their infrastructure on debt. You're paying 10% higher interest rate for these things. Finances do matter when you have great margins that can make a big difference.”at 41:16

Defense

• Aydin Senkut · Felicis

  Defense demand is structural because conflicts are here to stay and the sector needs better solutions.

  Senkut notes that three years ago few people were talking about defense, global resilience or energy. He argues that defense is not going to go away and that conflicts are here to stay, so the industry needs to do a better job. He treats defense as one of the areas that has become important in the last few years, alongside scalable energy solutions. The argument is brief and gives no company-level detail.

  “Three years ago, not a lot of people were talking about defense... Defense is not going to go away, the conflicts are here to stay, we need to do a better job at that.”at 8:25

Software & SaaS

• Aydin Senkut · Felicis

  AI lets anyone build custom software, which is driving big changes in the SaaS world.

  Senkut argues that AI has greatly lowered the bar to developing custom software, so anyone can build and customize a piece of software. This makes him ask where to invest when creating software is no longer that difficult. He says some fundamental software still matters, but he sees big changes underway in SaaS as a result. This is why Felicis has chosen to pass on many software deals and shift toward more capital-intensive areas.

  “We are now in a new world where because of AI, anybody can build any piece of software and customize it. The bar to developing custom software has gone down so much.”at 10:50

Robotics & autonomous vehicles

• Aydin Senkut · Felicis

  Robotics and physical AI are a huge opportunity; humanoid robots could replace low-wage labor at scale.

  Senkut is bullish on robotics and physical AI, arguing from market sizing. US, Western Europe and Japan have about 190 million low-skilled jobs paid roughly $20-25k a year, and robots that can work 24/7 could take them. He notes that a robotics company Felicis backed already has projects with Fortune 100 and S&P 500 companies, and that capturing 0.2% of this market would equal $10 billion in revenue. He says robotics is capital intensive but demand is vast, and he expects the wave to be as big as ChatGPT or bigger, with the market possibly expanding beyond current labor use. He also notes that many past robotics companies failed, which he sees as a reason to pick carefully. Felicis is investing across the robotics stack, including synthetic-data, data and parts makers.

  “One of the areas we're bullish on is robotics and physical AI... If all of a sudden we had robots that could do jobs 20 fourseven reliably, it might open up whole new areas”at 11:54

AI: overall outlook

• Aydin Senkut · Felicis

  Countries and continents will want their own AI and energy champions, not only US or European ones.

  Senkut argues the world has changed. Five to ten years ago sovereignty didn't matter much, and key areas like energy, AI and maybe robotics were dominated by global companies. Now he expects each continent to want its own champion for national-security reasons. He says this is not mutually exclusive with global leaders, and it opens up multiple ways to play the same wave, which is why Felicis spreads bets across data, manufacturing and sovereign elements.

  “five to ten years ago, the sovereignty didn't matter... maybe each continent will want to have their own champion. They're not gonna be satisfied with a champion from The US or Europe for different national security reasons.”at 16:55

Enterprise AI adoption & AI software

• Aydin Senkut · Felicis

  Search demand is shifting from consumers to AI models and companies, which could accelerate growth for search infrastructure.

  Aydin describes a search infrastructure company Felicis backs. He says many AI models and AI companies now use it, and that search use has changed because it is no longer mainly consumers going to Google. Felicis thought this company could grow five to 10 times faster than its already strong traction implied, which is the non-consensus part of the bet.

  “A lot of the AI models and AI companies are using it. The way that we use search has changed, it's no longer consumers going to Google.”at 37:45

AI pricing & unit economics

• Aydin Senkut · Felicis

  Inference vendors chase the same few big customers, whose buying power squeezes their margins.

  Many companies offering AI inference are all chasing the same five or six large customers. Those customers know their market power. Winning them brings scale, but it also badly hurts margins. Felicis's preferred inference-adjacent company avoided this by serving aggregated long-tail demand with an unconventional customer mix, which gave it margins an order of magnitude better than peers and insulated it from any single customer's capex or revenue swings.

  “One of the hot things right now is inference. There are a lot of companies helping with inference. Most of them are chasing the same five or six names. Those five or six names know the market power they have... it also absolutely kills your margins.”at 41:06

IPOs, M&A & deal activity

• Aydin Senkut · Felicis

  IPOs are rare, so startups need $1B to $10B revenue scale or they will exit through acquisition or PE-style sales.

  Aydin says IPOs are few and far between, and companies without revenue scale of roughly $1 to $10 billion may not get one at all. So Felicis looks for companies that can reach that scale very fast, or plans niche companies for PE-style or acquisition exits. He adds a rule of thumb that acquisition prices are almost always around 1% to 2% of the acquirer's market cap, so Felicis maps likely acquirers in advance to engineer exits.

  “The IPOs are fewer and far between, if you don't have a revenue scale of 1 to 10,000,000,000, it doesn't even matter.”at 48:30