Jun 19, 2025 · 49m · y-combinator
Elon Musk: Digital Superintelligence, Multiplanetary Life, How to Be Useful · Y Combinator
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
At Y Combinator's AI Startup School, Elon Musk speaks with Garry Tan about the rapid arrival of digital superintelligence, sharing hard-won founder lessons from Zip2, SpaceX, and Tesla alongside his first-principles approach to AI scaling, robotics, and multiplanetary survival.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The partners hold 19.7% of the talking time here. How this is scored →
The partners as informed peer 2.8 Guest teaching 2.3 Guest disagreement 1.0 The partners pushing back 0.2
speaking balance: gold is the partners, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:15 Elon rejects the 'researcher' label in favor of 'engineer'
When Tan compliments the audience as eminent AI researchers, Elon immediately interrupts to reject the researcher framing, stating that unless there is a fundamental algorithmic breakthrough, it is purely engineering.
Hardest push from the partners ▶ 29:48 Tan argues hardware uniquely demands first principles compared to softwareTan pushes a distinction between hardware and software, arguing software allows fluff and extra CPUs whereas hardware strictly requires refusing 'no', prompting Elon to clarify that first principles apply equally to both.
Biggest teaching moment ▶ 27:00 First principles derivation of rocket material costsElon educates the audience and host on why conventional historical pricing is flawed, calculating that raw materials constitute only 1 to 2 percent of historical rocket costs and proving huge manufacturing inefficiency.
The partners hold their own ▶ 33:55 Tan contributes insight on reasoning training dataTan actively extends Elon's physics theme with an informed observation from AI researchers that hard science and physics textbooks train reasoning effectively while social sciences do not, earning Musk's direct concurrence.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The partners as informed peer | Guest teaching | Guest disagreement | The partners pushing back | Why |
|---|---|---|---|---|---|---|
| Building Something Useful vs Great and Engineering vs Research | 3 | 3 | 2 | 1 | Garry Tan sets up grandiose questions about building something great and addressing eminent researchers, which Elon immediately deflects down to building something useful and preferring the term engineer. Tan adapts comfortably, noting the impact of inflation on early tech valuations. | |
| AI Economic Impact and the Government Waste Side Quest | 2 | 2 | 1 | 0 | Elon reflects on his government efficiency side quest versus the impending tsunami of AI, while Tan warmly agrees and validates the frustration with politics using a San Francisco comparison. | |
| X.com, the PayPal Diaspora, and Founding SpaceX | 2 | 3 | 1 | 0 | Elon promptly corrects Tan on his Zip2 earnings, noting he walked away with 20 million rather than hundreds of millions. Elon then recounts his trip to Russia to purchase ICBMs, while Tan listens attentively. | |
| SpaceX's Launch Failures, Tesla's 2008 Crisis, and Survival | 2 | 1 | 0 | 0 | Tan prompts Elon to discuss the 2008 period when critics labeled him an unqualified software guy trying to do hardware. Elon details the near-bankruptcy of both Tesla and SpaceX in a collaborative storytelling tone. | |
| Founder Advice: Utility, True Work, Ego, and RL Loops | 3 | 2 | 1 | 0 | Tan asks how founders can evaluate their own utility and team candidates. Elon explains the physics definition of true work and how high ego-to-ability ratios break reality feedback loops. | |
| First Principles Thinking and Building xAI's Memphis Supercluster | 4 | 3 | 1 | 1 | Elon outlines first principles reasoning in rocketry and the rapid construction of xAI's Memphis supercluster. Tan suggests hardware uniquely demands this rigor unlike forgiving software, but Elon clarifies the principle applies universally. | |
| AI Scaling, Synthetic Data, Reasoning, and Humanoid Robotics | 3 | 2 | 1 | 0 | Tan brings up domain-specific reasoning data, noting researchers found physics textbooks useful and social sciences useless. Elon agrees and predicts humanoid robots will outnumber humans and other robots combined. | |
| Kardashev Scale, Mars Colonization, and AI Truth for Safety | 3 | 2 | 1 | 0 | Tan asks how multiplanetary ambitions fit into long-term survival and how to overcome the Fermi Paradox. Elon emphasizes that Mars must achieve self-sustainability within 30 years and insists AI safety requires uncompromising adherence to truth. | |
| Digital Superintelligence Timelines, Superabundance, and Neuralink | 3 | 3 | 1 | 0 | Tan inquires about superintelligence timelines and Neuralink's role in closing the human-AI bandwidth gap. Elon notes digital superintelligence will arrive imminently and details how low human daily output bandwidth currently is. |
Statements from this episode (24)
Musk: AI hype and early valuations are eye-watering, though some will pan out
“The hype level on AI is, is, is pretty intense as you've seen. You know, you see companies that are, I don't know, less than a year old getting sometimes billion dollar or multibillion dollar valuations. Which I guess could pan out and probably will pan out in…”
Musk: Giving legacy incumbents board control forces irrational decisions on new tech
“I guess for the first, my first startup, the big, the, really the mistake was having too much shareholder and board control from legacy media companies who then necessarily see things through the lens of legacy media and that they'll kind of make you do things…”
Musk: AI could expand the economy thousands or millions of times
“You know, the economy, assuming we don't, things don't go awry, and like AI doesn't kill us all and itself then you'll see ultimately an economy that is not, not 10 times more than the current economy. Ultimately, like if we become, say, or whatever our future…”
Musk: Government efficiency work is a side quest compared to AI
“I did sort of feel a bit like, you know, when I was in D.C. Taking a lot of flack for, like, getting rid of waste and fraud, which was an interesting side quest as side quests go. But Got to get back to the main quest. Yeah, I gotta get back to the main quest …”
Musk: Engineering requires maximum truth-seeking because math cannot be fooled
“If you're trying to build a rocket or cars, or you're trying to have software that compiles and runs reliably, then you have to be maximally truth-seeking or your software or your hardware won't work. Like, there's not, you can't fool math, like math and physi…”
Musk: Big defense contractors will not advance rocket technology
“If a startup doesn't do something to advance
rocket technology.
It's definitely not coming from the big defense contractors because they just impedance match to the government and the government just wants to do very conventional things.”
Musk: A high ego-to-ability ratio breaks feedback loops to reality
“Like a major failure mode is when ego to ability ratio is double greater than sign one. You know like if you, if your ego to ability ratio is, it gets too high, then you're gonna basically break the feedback loop to reality. And in, in AI terms, your arrow, yo…”
Musk prefers the term engineering over research and calls xAI a company, not a lab
“That's kind of like why I actually, I prefer the term like engineering as opposed to research. I prefer the term and I don't, I actually don't want it to call XAI a lab. I just want to be a company.”
Elon Musk: Raw materials are only 1% to 2% of rocket costs
“Actually a rocket, the raw materials of a rocket are only maybe one or two percent of the historical cost of a rocket. So The manufacturing must necessarily be very inefficient.”
Musk: xAI rented a quarter of all US mobile cooling capacity
“We rented about a quarter of the mobile cooling capacity of the U.S. And put the chillers on the other side of the building.”
Musk: xAI deploying 140K GB200s and 150K H200s in Memphis
“So now we've got 150,000 H 150 K H 200 and 30 K GB 200 in the Memphis training center. And we're about to bring a 110,000 GB 200 online at a second data center also in the Memphis area.”
Musk: AI developers have run out of high-quality human pre-training data
“We, we've kind of run out of pre-training data or human generated pre like human generated data. You run out of tokens pretty fast certainly of high quality tokens.”
Musk: xAI is currently training Grok 3.5 with reasoning focus
“And, like, right now we're training Grok 3.5, which is a heavy focus on reasoning.”
Musk: Humanoid robots will outnumber all other robots combined tenfold
“There's gonna be so many humanoid robots, and robots of all sizes and shapes, but my prediction is that there will be more humanoid robots by far than all other robots combined, by maybe an order of magnitude, like a big difference.”
Musk: Humanoid robots will outnumber humans five to tenfold
“I mean, I'd predict that there's gonna be,
At least five times as many humanoid robots as there are humans, maybe 10 times.”
Musk predicts Mars will be self-sustaining within roughly 30 years
“I think we'll have enough mass transferred to Mars within like roughly 30 years to make Mars self-sustaining, such that Mars can continue to grow and prosper, even if the resupply shifts from Earth stop coming.”
Musk: Forcing AI to believe falsehoods makes it dangerous
“Something that I think is extremely important in building AI is, is a very rigorous adherence to truth, even if that truth is politically incorrect. I, my intuition for what could make AI very dangerous is if you force AI to believe things that are not true.”
Musk: There will be 5 to 10 frontier AI models globally
“I do think there will be several deep intelligences, maybe at least five maybe as much as 10. I'm not sure that there's going to be hundreds, but it's probably close, like maybe there'll be like 10 or something like that of which maybe four will be in the US.”
Musk: Digital superintelligence will arrive by 2026 for sure
“Like, I think we're quite close to digital superintelligence. It may happen this year, and if it doesn't happen this year, next year for sure. Digital superintelligence is defined as smarter than any human at anything.”
Musk: There is a 10% to 20% chance AI causes annihilation
“I think it most likely will be a good outcome. I guess I'd sort of agree with Jeff Hinton that maybe it's a 10 to 20% chance of annihilation. But look on the bright side, that's 80 to 90% probability of a great outcome.”
Musk: Digital superintelligence will happen before Neuralink scales
“The Neuralink is not necessary to solve digital superintelligence. That'll happen before Neuralink is at scale”
Musk: Five humans have received Neuralink implants
“We have now five humans who have received the kind of the read where it's reading signals, and you've got people with ALS who really have no, they're tetraplegics, but they can now communicate at, with Similar bandwidth to a human with a fully functioning body…”
Musk: Neuralink will do first human vision implants in 6-12 months
“And then I think in the next six to 12 months, we'll be doing our first implants for vision, where even if somebody's completely blind we can write directly to the visual cortex.”
Musk: Augmented Human Intelligence Will Probably Be One Billionth of Digital Intelligence
“Human intelligence, even assuming a significant increase in human population and intelligence augmentation, like massive intelligence augmentation, where like everyone has an IQ of a thousand type of thing. Even in that circumstance collective human intelligen…”