Sam Altman on Start-ups, Venture Capital, and the Y Combinator
What this covers
Sam Altman, president of Y Combinator, discusses how startups succeed and which economic sectors are ripe for disruption. The conversation centers on a foundational claim: that Y Combinator bets on founder quality—determination, communication, and obsession—rather than polished business plans or even proven product-market fit. Altman explains why the fund deliberately remains an advisor rather than boss, allowing founders to steer and fail, and why it funds companies like AirBnB despite early product dysfunction, betting instead on founder resourcefulness and the capacity to pivot. The discussion ranges across how computing platforms create value in ways their early detractors cannot foresee, the role of network effects in startup survival, and why valuation and financial projections are irrelevant at the earliest stage.
The second half moves into sector-level analysis. Altman argues that falling costs and shrinking cycle times—what he frames as the Cost and Cycle Time Framework—are the primary enablers unlocking innovation in previously untouchable fields like biotech, energy, and nuclear power. He positions clean energy as civilization's highest-leverage problem, traces the technological slowdown in physical systems to post-WWII fear of science, and explains why clean-tech companies failed by chasing markets too vast and relying on subsidies. The conversation also touches artificial intelligence, Bitcoin and its blockchain layer, the mechanics of Beachhead Strategy, and his prediction that great products meeting real demand will overcome regulatory and incumbent resistance with near-certain historical regularity.
Altman argues that startup success is driven primarily by founder quality (determination, communication, obsession) rather than ideas or business plans, and that falling costs and cycle times are unlocking innovation in previously dead sectors like biotech, energy, and nuclear—while consumer demand inevitably defeats entrenched incumbents over time.
- YC funds founders over ideas, betting on the pivot because determined founders find a way when products fail
- Lower cost and shorter cycle times are the key enablers letting startups succeed in capital-intensive fields like biotech and energy
- Great products that people want defeat regulatory and incumbent resistance with near-100% historical reliability, and the cycle time of that defeat is compressing
YC backs founders who are determined first and smart second, seeking those with relentless resourcefulness who will persist through impossible-looking obstacles.
- YC funded AirBnB despite a product that wasn't working because the founders demonstrated 'relentless resourcefulness'—they had financed themselves by designing and selling limited-edition Obama and McCain cereal boxes for $40,000—which signals the kind of founder who finds a way when everything looks impossible.
“They made $40,000. Which they used to continue to operate their startup when nothing was working.”
- YC looks for founders who are determined and smart, in that order, because startup success is a game of being willing to beat your head bloody against the wall for years; endurance matters more than intelligence.
“On the founder front, we look for founders that are determined and smart. In that order. It is more important to be determined than smart.”
- Good founders care so much about the product—almost too much—becoming obsessed with small details of the user experience, a trait shared among successful founders whether or not it derives from emulating Steve Jobs.
“Another thing that we see among good founders is that they care so much about the product. Almost too much. They get obsessed with small details of the user experience”
- Founder communication ability is incredibly correlated with success because founders become the chief evangelist for the company, talking to hundreds of employees, press, customers, and suppliers weekly; being a poor communicator is a major strike against a founder.
“We look for founders that can communicate well. That turns out to be incredibly correlated with success.”
YC's network creates powerful community effects where startups help each other and face less predatory treatment, significantly improving success odds.
- A central organization like Y Combinator creates network effects—a powerful community where startups help each other and where investors and big companies are less likely to mistreat any company in the family—which makes funded startups significantly more likely to succeed.
“our belief is having sort of this central organization like Y Combinator can make these startups way more likely to be successful”
- Putting startups together in one shared incubator bullpen is counterproductive because founders need to feel unique and special; pooling them makes people feel like employees, creates distraction and poaching, and produces a 'farce' of startups, so YC deliberately does not house its companies.
“It turns out that startups need to feel unique. Startups need to feel like they are kind of taking on the world and they are special and sort of this like chosen thing. So, the worst possible way to get that is to put everybody in one big bullpen.”
- Y Combinator is nine years old as of the recording and has funded 716 startups so far.
“We funded I think 716 so far.”
- Since YC got going around 2007, half of the technology companies that have been started and reached $10+ billion valuations have gone through Y Combinator, with more than 20 companies above $100 million and three above a billion dollar valuation so far.
“since we sort of really got going, two years in, since 2007, half of the $10+ billion dollar technology companies that have been started have gone through YC.”
New computing platforms—wearables, augmented reality, virtual reality—arrive rarely but create enormous value; assessing killer applications early is nearly impossible.
- Artificial intelligence is undervalued and people should pay more attention to it; its early promise underdelivered due to cycle-time problems, leading people to dismiss it, but that is a mistake because it remains important.
“Like I do actually believe AI is undervalued and people should pay more attention to it.”
- It is hard to conceptualize the value of a new platform because the killer applications for it don't yet exist; pundits dismissing a new platform by noting it merely duplicates current functions are repeatedly wrong, as the history of predicting whether we need a new technology is horrible in both directions.
“the reason it's hard to conceptualize is that the things you are going to use those platforms for don't yet exist.”
- Embeddable technology and the merging of humans and machines will arrive gradually through small versions—a pill with a sand-grain sensor, an RFID chip, or a nanobot in the bloodstream—rather than a single dramatic moment, and Altman is confident he will eventually inject some kind of embeddable into his body.
“even if it's a little nano robot that gets injected in my bloodstream that does something... I am very confident that at some point in my life... that I will inject some sort of embeddable into my body.”
- Computers and humans are already augmenting each other in big ways, so although there is no dramatic SciFi moment of cables into the brain, in 100 years people will see we were much further along the man-machine merge than they think; a human's capability today is qualitatively different from 30-50 years ago.
“Computers are augmenting humans in big ways. Humans are augmenting computers in big ways. And that's already happening.”
- Just as the 20th century was the carbon century, the 22nd century will be the atomic power century—powered either by terrestrially generated nuclear or by sun-generated energy (solar, wind, biofuels), since both derive from the same fundamental force; the cheapest source of power historically becomes dominant, but which of the two wins is too hard to call now.
“the 20th century was clearly the carbon century. And I believe the 22nd century is going to be the atomic power century. I'm very convinced of that.”
- Every time in history a new computing platform has emerged—supercomputer to desktop to laptop to phone to tablet—it has been hugely important and created enormous value, but such platforms come along rarely; the next ones will be wearables (including augmented reality) and virtual reality.
“Every time in history that there has been a new computing platform, it has been hugely important and a ton of value has been created. But they don't come along very often.”
- Safe, cheap, clean energy is probably the most important thing an individual can do for the future of the world because many other problems—clean water, environmental damage, economic challenges, and even wars—reduce to the cost and availability of energy; at one tenth today's cost, the world could have all the clean water it wants.
“I really do believe that safe, cheap, clean energy is probably the most important thing an individual can do for the future of the world.”
YC's governance model—advisor not boss, hands-off on product, no later-round investing—preserves founder autonomy and avoids conflicts that would harm portfolio companies.
- YC deliberately positions itself as advisor rather than boss—it will let founders 'drive off the cliff' and fail—because the founder must run the company; investors are along for the ride and cannot make the product decisions.
“We specifically are not hand-holders. And we tell them, in the first day, in the first dinner, we are not your boss. We will not hold your hand. If you want to fail, we will let you drive off that cliff.”
- YC deliberately does not invest in the later, larger funding rounds (e.g. $3-12M checks) at the end of its program because investing in some companies and not others would create a conflict that signals and badly hurts the companies not chosen.
“If were invested in some companies at the end of our program and not others, it would really hurt the ones we didn't invest in.”
- YC offers every company the same standardized deal—$120,000 for 7% of the company—with essentially no negotiation, because at the speed and volume YC operates, negotiating each offer would be too costly in time.
“our deal is always the same. It's $120,000 for 7% of the company.”
Low cost and short cycle time are the two most critical factors for startup success in any domain; good ideas fail when both are too high.
- The Amazon Web Services model has arrived for biotech: just as web startups no longer set up their own servers, life-science startups can now push a button to have a remote roboticized lab run a physical experiment, while cheap powerful gene sequencers and robotic arms provide the throughput of thousands of humans with no mistakes around the clock, collapsing cost and cycle time.
“there are now these labs where you can say, 'Hey, do this physical life science experiment for me. Here are the parameters. Here's the experiment to run. Do it.' And you push a button on your computer, and just like Amazon, if you ask, they do it.”
- Low cost and short cycle time are the two most important factors for startups to succeed in a given area; many good ideas (like early biotech) fail simply because they are too early when costs are too high and cycle times too long, and become viable only when both fall dramatically.
“I've really come to believe that those two things, low cost and low cycle time, are the most important things for startups in a given area to be successful.”
- Clean tech largely failed because companies went after markets that were way too big, had no answer for how to be economically competitive (relying on government subsidies not drying up), had broken teams, and raised too much capital.
“All the clean tech stuff has basically failed. But it had all these weird problems--that companies were going after markets that were way too big. They never had an answer to how they were going to be economically competitive.”
The best startup ideas look contrarian and unpopular early; fashionable categories fail to copycats while unfashionable-seeming options like Twitter succeed.
- YC favors ideas that are new and almost unpopular—ideas that look like a bad idea but are in fact good and that few people are pursuing—because by the time an idea is fashionable it is too late; copycats of fashionable categories fail while the contrarian-looking option (e.g. Twitter) succeeds.
“we like ideas that are new, and that are almost unpopular. A friend of mine once drew this great Venn diagram, and one circle was things that look like a good idea, is a good idea; and the other circle was, looks like a bad idea. And you want the things that are at this intersection.”
- YC will sometimes 'fund for the pivot'—deliberately funding a team whose current idea is terrible and is known to fail, purely because the founders are so good that YC wants them in the community, on the expectation they will pivot to a viable idea.
“Fund for the pivot. Which means that this idea is so terrible, but this founder is so good, we're going to give them money anyway; and we know this first idea is going to fail.”
Consumer demand, given time, overrules entrenched interests; great products people desperately want win with near-100% historical track record, slowing innovation only.
- Consumer demand, given enough time, will overrule corrupt entrenched entities; great products that people desperately want win with a near-100% historical track record, and entrenched interests can only slow innovation down, not stop it.
“I think consumer demand given enough time will overrule corrupt entrenched entities.”
- The amount of time entrenched interests can hold off a new innovation is condensing because mobile phones, social media, and rapid peer-to-peer information spread accelerate adoption, compressing the cycle time of disruption.
“in a world of everyone having a mobile phone and social media and these things just spreading among people very quickly, that the amount of time these entrenched interests can hold something off is just condensing.”
Technology shapes society's values and norms rather than reverse; fabulous products rewire how people think and what they accept as normal.
- Technology shapes society rather than the reverse; whatever technology exists as you grow up becomes your accepted world, and the more fabulous the product the easier people realign their thinking to internalize it, which then drives the culture.
“There's this--what actually happens is technology shapes society. It's not the other way around.”
- Whatever privacy norms a person is comfortable with around age 18 they will be comfortable with for the rest of their lives; older people only slowly grow more comfortable with new technologies and never adopt them the way those who grew up with them do.
“whatever the privacy norms are, when you are say, 18, you are going to be comfortable with that for the rest of your life as you grow up. And then older people will slowly get more comfortable with these things”
- Progress always feels like we are on the cusp because on an exponential curve any point looks vertical when you look up, but zooming out reveals we have already passed many cusps—so the recurring feeling of being on the verge of transformation is structurally inevitable.
“It always feels like we're on the cusp. Because it's one of these exponential curves. So at any point on that curve when you look up it looks vertical. But if you can zoom out and look at how far we've already come--we've gotten over a lot of cusps already.”
Early YC decisions rest on whether founders built product, spoke to users, and can brainstorm productively, not on business plans or financial projections.
- At YC's early stage, business plans and financial projections are irrelevant and never read; the decision rests on whether founders have built a product, spoken to users, and can engage in productive back-and-forth brainstorming, decided in a ten-minute interview on the spot.
“The startups--we look at their demo, but we never read the business plan.”
- Startups should begin with an appropriately small subsection of the problem—a beachhead that can scale—and often do things that don't scale at the start; AirBnB began in one city going door to door taking photos, and would never have gotten off the ground attempting its current scope from day one.
“You also want something that is the appropriate size to start with. If AirBnB, since we've been talking about them, tried to start what they are doing today, they never would have gotten off the ground.”
Startups succeed by unlocking latent supply to meet latent demand, monetizing underused assets and expanding effective supply without physical construction.
- Software platforms like AirBnB and Uber succeed by unlocking latent supply to meet latent demand—they enable people to monetize underused assets (a spare room, a personal car) that other people want, expanding effective supply without new physical construction.
“All of these situations where software platforms can expand supply, when there is latent demand--Uber is another great example.”
Bitcoin's killer innovation is blockchain as a decentralized ledger, not payments; its real potential lies in distributed file storage, contracts, and future applications.
- The big challenge for Bitcoin as a currency is that it still has not found what it is better at than other payment systems; estimated transaction volume is not undergoing the runaway growth widely predicted, legitimate transaction use is very small, and the most promising green shoot is international remittance.
“the big challenge with BitCoin as a currency is, it still has not found what it's better at than other payment systems. And that's really important.”
- The most interesting piece of Bitcoin is the blockchain—a decentralized network agreeing on the truth via a shared public ledger verified by math and the Internet, with an incentive system to make it work—whose applications extend beyond payments to distributed file storage, contracts, and future revenue shares.
“I think the most interesting piece of BitCoin is this idea of the blockchain, this idea we can have the centralized network agree on what the truth is, and have an incentive system to make that work.”
- One of Bitcoin's greatest assets is that many people—everyone who holds Bitcoin and the VCs heavily invested in Bitcoin companies—are strongly incentivized for it to continue existing, which may provide the activation energy to bridge the gap between its current state and its potential.
“it has this kind of great thing which is that there are all these people that are really incented for it to continue to exist, which is everyone that holds BitCoin. And all the VCs who invest a lot in BitCoin companies.”
- The innovation in fracking is one of the great technology stories of the last ten years, a quiet revolution producing actual lower-cost energy with more advances coming, in the world's biggest market.
“The innovation in fracking is remarkable. That is one of the great technology stories of the last 10 years.”
Spectacular failures disproportionately shape public perception while slow steady harms are underweighted; nuclear safety vastly exceeds coal despite worse optics.
- The main obstacle to nuclear power is a public perception problem rather than a regulatory one—the NRC, staffed by nuclear engineers, wants nuclear to happen; the issue is that humans badly overrate spectacular failures, even though nuclear is hundreds of thousands of times safer than coal, because when nuclear fails it fails spectacularly while coal deaths come slowly via lung cancer.
“I think the problem with nuclear is a public perception problem, not a regulatory problem first and foremost.”
- The technological slowdown in the 'atoms' world dating to the early 1970s stems in a big way from the 1945 atomic bomb, which—amplified as the Cold War heated up—transformed the public perception of technology from wonder into fear, a Frankenstein-style fear of science that society has never fully recovered from.
“my view is that what it really was in this very big way the atomic bomb in 1945 that sort of changed the perception of technology.”
Healthcare and life sciences are entering rapid development as costs and cycle times collapse through robotics and sequencing, enabling personalized medicine at scale.
- Healthcare is in a period of great development after long being ignored, but most investors are still not paying attention; in about two years some of these companies will succeed and trigger a flood of investment that arrives too late.
“Most investors interestingly enough are still not paying a lot of attention. And probably in 2 years, when some of these health care companies get successful, there will be a true flood of investment into this space, and it will already be too late.”
- A huge amount of personalized health data is now available—a full genome for around a thousand dollars, microbiome sequencing, 800 biomarkers from blood trended over time—and new companies can take a piece of a patient's tumor and test hundreds of drug combinations against it to find the optimal treatment, replacing a doctor blindly trying a few options.
“You can get your entire genome--the whole thing--for a thousand dollars, soon. You can get your like, your microbiome, the bacteria in your guts, sequenced.”
All great platforms enable both terrible and wonderful uses; bad PR never kills successful products as people return despite prior swearing off.
- Any great platform will have terrible and wonderful uses, and although bad PR events generate disproportionate attention and schadenfreude, they never actually kill a successful startup—people return and use the product again even after swearing they wouldn't.
“for all that the press loves to say, 'Ah, this happened; this thing is over,' and sort of whip themselves up in this frenzy, it never actually kills the startup.”