Summary

Musk’s core timeline call: AI may exceed the sum of human intelligence in roughly 5 years, and by 2036 humanoid robots as the “end effectors” of digital intelligence create “a sort of quasi-infinite economy.” Pressed on how shareholders get paid, he dissolves the question — “I’ll make another prediction. Money won’t matter in 2036” — because if robots and AI produce more than “any human could possibly consume, what do you need money for?”
His stated risk hasn’t moved — the 10-20% killer-robot odds he gave on Ted Cruz’s podcast still stand — but his posture has flipped to fatalism: “I can’t see any way to really stop this incredible momentum… even if there was a stop button, we probably shouldn’t press it.” The host’s test — would he board a rocket with a 10-20% chance of exploding? — gets: “Yes, but let’s say you can’t do anything about it.”
A concrete governance proposal he’d start “immediately, really”: leading labs, including the Chinese ones, hold a call every week or two, and competitors get “as little as a week” of API access to test frontier models pre-release — MPAA-style self-rating, with government stepping in only when a builder won’t fix a flagged danger. The precedent: “it wasn’t as though someone in the government figured out Mythos had scary cybersecurity risks. It was Amazon. Amazon called the White House.” Timeline pressure: “six months is a long time.”
The China call: the AI constraint is electricity outside China, chips inside it. China already has more electricity than the US, Europe and India combined, heads to ~4x US production by his guess, and is “closer than most people realize to solving the lithography problem” — so “there’s a good chance that they are the leaders at some point.” Banning US firms from Kimmy K3 “won’t stop China from being the leader”; orbital data centers fix the power constraint, after which “the constraint will once again be chips outside of China.”
Labor: AI is already better than 90% of professional software engineers, heading to 99% and then “Stockfish level” — no way to compete — “and that applies to everything. Everything, everything.” Work becomes optional like gardening (the store’s tomato is plumper, but dinner from a friend’s garden is “a nice touch”); the answer is “universal high income,” Treasury-issued checks, and the contrarian macro call: “deflation will be the issue, not inflation” if goods output outruns money creation.
He defends ~80% post-IPO voting control as a long-horizon device — Mars spending was “literally listed” in the S1, yet quarterly pressure persists (“you missed your earnings this quarter cuz you spent too much on Mars” — “Yes”) — praises retail investors over quarterly-incentivized portfolio managers, and answers key-man risk with the Apple test: fine for years, but no more “Steve Jobs-level breakthrough product.”
The back half turns combative: Musk stands by “civil war in Britain is inevitable… if current trends continue” (his own timeline: ~20 years out, versus 10 for the singularity), while the host counters with crime data and his last UK visit being “a few years ago.” His counterpunch: “I think you’re the problem, the media… they hate you far more than they hate me.” On Dodge: “I got a little too involved in politics, got carried away, frankly” — but he insists zero people died from the aid cuts, since the Gates Foundation’s $50bn and MacKenzie Bezos’s tens of billions were available and “if they didn’t [act], I would say they’re equally responsible.”

Deep dive

1. The 2036 call: a “quasi-infinite economy” where money doesn’t matter

Musk’s unifying frame for all his companies: the economy is “digital and physical intelligence.” Digital AI produces breakthroughs “almost with every passing week” but “doesn’t have end effectors” — humanoid robots supply them, “lots of bots in a nutshell,” and vast robots plus vast intelligence yields “a sort of quasi-infinite economy.” AI may exceed the sum of human intelligence in around 5 years; within ten, “there really won’t be anything that AI can’t do better than humans apart from being human, perhaps.”
Asked how the businesses actually earn — the host notes the SpaceX IPO prospectus showed the overwhelming share of revenue coming from AI, from Grok, not launch — Musk goes over the question’s head: “Money won’t matter in 2036.” You want money for food, housing, transport, entertainment; if AI and robots provide more than anyone can consume, the question dissolves.
The hedges he keeps: abundance assumes no derailment “like a massive global thermonuclear war,” and “there’s no analogy or metaphor that illustrates the magnitude of change that we’re going to experience here.”

2. Humans won’t be in charge — and the stop button stays unpressed

He states it flatly: continued human control is “unlikely.” If the AI-human intelligence gap exceeds the human-chimp gap, “it’s hard to imagine that the chimpanzees would be in charge.” The remaining lever is values: make AI “maximally truth seeking and curious… if that’s the case, I think it will foster humanity.”
The host tracks the arc of his views: the decade-old fear of rapid recursive self-improvement (“we will just be the pet Labradors at best” — Musk now: “Labradors are pretty happy creatures”), the 2023 pause letter (“one of many, like 500 signatories”), the 10-20% doom odds last year. His answer: the risk is “not zero,” but “I can’t see any way to really stop this incredible momentum… even if there was a stop button, we probably shouldn’t press it — because the most likely outcome is incredible abundance for all.”
His own history argues for inevitability: he created OpenAI as a counterweight to Google’s near-monopoly, Anthropic spun out of OpenAI and now leads — “all roads lead to acceleration of AI… you can’t beat them, join them.”
The nihilist coda, half-joking: physics ends in heat death, “a long, boring, thin, gray soup of gradual cooling” — “if that’s the outcome for our reality, then it really is all about the journey. Because the destination is terrible.”

3. The safety architecture: rivals test each other, government as backstop

His proposal — given to Demis Hassabis over hours of conversation before Hassabis published his private-public regulator piece: leading labs meet by call “once every few weeks,” and before any frontier release competitors get a week or two of API access to test for harmful effects and recommend pausing. Government can’t do the judging — “quite difficult for someone in the government who doesn’t have a deep technical understanding” — but competitors “have an incentive to keep the others honest.” The template: the Motion Picture Association rating its own industry’s films.
Government keeps ultimate power, reactively: if a flagged company won’t address a danger, alert the state — “it’s really just the US government and the Chinese government that have the power to take action here.” That is “literally what just occurred with Mythos”: “it wasn’t as though someone in the government figured out Mythos had scary cybersecurity risks. It was Amazon… Amazon called the White House.”
On timing, people keep telling the host six months. Musk: “Six months is a long time” — start “immediately, really,” and include “the Chinese companies that are doing frontier AI models, too.”
Can labs that insult each other on X cooperate? “I’m not a fan of Sam Altman”: a nonprofit meant to be open-source and “owned by the world” became “an $800 billion for-profit company with closed source… the exact opposite of what I donated the money for. That’s my issue. I think it’s a legitimate one.” Dario is “a very principled person”; Anthropic exists because its team “didn’t trust Sam Altman”; at Anthropic “no one set off my evil detector” — though “the road to hell is mostly paved with bad intentions. There are a few well-intentioned paving stones in there.” Bottom line: “if we have to talk, we’ll talk… set aside our personal differences for the good of the world.”

4. China wins on electricity; Kimmy K3 is closing on Fable

The model scoreboard as he calls it: “Fable is still clearly the smartest model… anyone realistically would say that,” but the new Kimmy K3 released last week “is getting quite close.” And Anthropic is sitting on more — Mythos, from which Fable is derived, was “ready in February,” so “they certainly have right now” something much better, releasable at any time.
The structural argument: AI is a function of whatever binds — chips or electricity. Outside China the constraint is power and cooling (chips are being made faster than new electricity comes online); inside China it’s chips, thanks to US export bans. But China is “closer than most people realize to solving the lithography problem,” already has more electricity than the US, Europe, and India combined, and by his guess reaches 4x US production — roughly proportionate to population. “There’s a good chance that they are the leaders at some point.”
Policy follows: banning US companies from Chinese models “won’t stop China from being the leader in AI” — Washington “can’t stop the rest of the world from using them.” Orbital data centers solve the power constraint, and then “the constraint will once again be chips outside of China.”
Color he volunteers: China’s robot fighting matches — one robot “got its head knocked off, kept fighting even with its head knocked off… the robot battles are going to be quite entertaining to watch.”

5. Every job goes to Stockfish level; work becomes gardening

On the host’s citation of Dario’s call that 50% of entry-level white-collar jobs vanish within a few years, first a jab: “Dario did dig his own grave” by calling Methuselahs “super scary, terrifying, and also, by the way, we’re going to release it… you’ve literally told them to be scared and then you release the scary thing.”
His own pace estimate is hardly softer: AI will do “any job better than any person,” very soon for anything involving “a person at a computer or on a phone.” Software already: better than 90% of professional software engineers, heading to 99%, then “Stockfish level” — Stockfish beats Magnus Carlsen probably even running on a phone — “and that applies to everything. Everything, everything.”
The gardening analogy, as told: you don’t have to grow vegetables; your tomato “is not going to be as plump and juicy as the ones at the store,” but dinner from a friend’s garden is “a nice touch. That’s what work is going to be like.” He concedes “it will be a bumpy road” — the word “computer” once meant a person, skyscrapers full of them calculating bank interest — but “the pace of change is radically accelerated.”
His recommended syllabus: Ian Banks’s Culture novels, “the best envisioning of an AI future that I’ve read.” The host, reading Excession on his prompting, objects that humans have minimal agency next to the Minds; Musk insists “they have agency” — while noting Banks was a socialist with “a very different political philosophy.”

6. Print the checks: deflation, not inflation

How do the displaced live? “There’s going to be universal high income… the Treasury should just simply issue people checks.” The host — “I’m an economist” — objects that pure issuance means inflation. Musk’s rebuttal: inflation is the ratio of money to goods and services; if output rises 1,000%, you can “create money in the database” and still get falling prices. “I’ll make a prediction, which is that deflation will be the issue, not inflation.” The host, notably: “You may well be right.”
On paying for redistribution: “I set a record for the most amount of taxes ever paid by a human… my guess is that I will pay many trillions in tax. And I’m fine with that.” His arithmetic: ~40% national income tax on stock options plus roughly 15% in California state taxes — “about 45%” — then roughly half again at death: “at the end of the day I’ll probably have a quarter of whatever I have.” But in the future, “taxation also becomes somewhat irrelevant.”
On sharing the upside: one reason SpaceX went public was the shareholder cap — private companies are limited to a few thousand holders, public ones can have millions. What wealth buys him is only “control of the direction of the companies… until the AI is so smart that it’s in charge. And at that point controlling companies won’t really matter.”

7. Eighty percent voting control, defended as a long-horizon device

The host’s charge: post-IPO Musk holds ~80% of voting shares and can only be removed by votes he himself controls — “you can go crazy.” His deflection: Alphabet (Larry and Sergey) and Meta (Zuckerberg, who “released one last week” and will be “one of the AI companies”) sit in the same bucket.
The affirmative case: control protects 5-10-year investments from quarterly punishment. Mars spending “was in the S1 prospectus… literally listed as a thing we will do,” yet “‘you missed your earnings this quarter cuz you spent too much on Mars.’ And I’m like, ‘Yes.’” Portfolio managers pressure for immediate returns because their comp and evaluation cycles demand it — “once you understand the incentive structure, the behavior naturally follows.” Retail investors, by contrast, are “very insightful and have very long-term views.”
Key-man risk gets the Apple answer: there’s “a clear road map — or space map, I should say” and the companies would “do very well for several years,” but post-Jobs Apple makes “amazing phones, great computers” without “a Steve Jobs-level breakthrough product.”
Mars is undemoted, reframed: the goal is maximizing “the future light cone of consciousness.” Starship is “the largest, most powerful flying object ever made, and we intend to launch that more than once per hour.” Seeing Star Wars at six “blew my mind… maybe had a profound effect on me” — and the AI wave “is pretty much happening as I thought it would,” citing Ray Kurzweil’s “remarkably accurate” predictions.

8. Starlink as geopolitical switch: “no angels in war”

On cutting Russian access: Starlink was never sold to Russians, but units ordered through Ukraine were smuggled into occupied territory and “used in some cases for attacks.” The fix — a whitelist of approved terminals built with the Ukrainian government — has real downsides: “a whole bunch of people who were just innocent users… now don’t have Starlink.”
Asked how it feels to hold war-shaping power, he pivots to the substance: the front line “has barely moved” for two or three years, Russia “is not going to withdraw,” and Ukraine “is not going to retake those territories. It is not going to happen.” His scorn: “fancy diplomats who have seven-course dinners in luxurious places while people are dying at the front lines.”
He rejects the pro-Russia label — “I’m just being pragmatic, or people will die for years with nothing achieved, and that is exactly what happened” — citing “a very deep dive into the background of that conflict” and “the old adage that there are no angels in war.” The host credits his support of Ukraine as right and important, while pressing that an individual able to shape a conflict by flipping a switch is historically rare.

9. Dodge postmortem: “got carried away” — and the zero-deaths hill

The rare concession: “I think I got a little too involved in politics, got carried away, frankly.” The motivation stands — the deficit is massive, and interest payments on the debt exceed everything spent on the Department of War plus intelligence combined.
His signature anecdote: Dodge merely asked for recipients’ contact information — “we need to wire money for some important cause in Africa… but the wiring instructions are to Deloitte & Touche in Washington, DC… can you please connect us with the recipients so we can send them the money directly? And then we’d get silence.” Waste, he says, is “much worse than anyone could possibly imagine.”
The sharpest clash of the interview: the host argues the suddenness of the aid cuts — antiretroviral funding stopping overnight — “almost certainly” caused unnecessary suffering and deaths. Musk: “Zero… these claims are false. They’re nonsense.” With tens of thousands of NGOs, the Gates Foundation’s $50 billion, and McKenzie Bezos’s tens of billions available to act quickly, “if they didn’t, I would say they’re equally responsible.” USAID “was a political organization… of regime change,” and defunded organizations naturally “come up with a sad story that sounds sympathetic.”
On Trump: “this administration on balance is excellent. No administration is perfect, but it’s vastly better than the alternative.”

10. The Europe brawl — and the Cassandra defense

Musk stands by the post: “Civil war in Britain is inevitable… if current trends continue.” The mechanism: a rapidly growing group “whose beliefs are antithetical to Western beliefs” will try to implement its views, opponents will resist — “that’s what you call civil war… A child can see this.” His platform: “secure borders, safe cities, sensible spending — which of those three are far-right fringe?” And the forcing function: a welfare state paying benefits above poor-world living standards creates “a very powerful economic forcing function” for migration.
The host’s counters are specific: Musk was last in the UK “a few years ago”; London is safer than any US city; and on two of Musk’s own three tests — safe cities and sensible spending — Europe beats the US. Musk’s rejoinders: American danger comes from failing to incarcerate repeat violent offenders (“the classic Soros DA”), and “there are more people that die in Europe from heat deaths than people that die from gun deaths in America — why aren’t you writing an article about air conditioning?” (Host: “we’ve done several.”) Both accept a UK tour to settle it.
The media turn, kept in by his own demand (“Please keep this part in… don’t cut this part”): “I think you’re the problem, the media… a favorable view of journalists is like 15%… they hate you far more than they hate me. Do you even know that? You probably don’t.” He rejects the far-right label — “my political views are very centrist” — and the racism charge: a woman he says is half-Indian, with whom he has four children, one named after a famous Indian physicist.
On the timescale, he says the singularity dominates the macro effects: “civil war in Britain is probably 20 years away, and the AI robot singularity is 10 years away.” Host: hopefully the benign all-powerful AIs prevent such outcomes. Musk: “They probably will.”

Jensen Huang Launches an Open AI Alliance, Anthropic & OpenAI Team Up in DC, Kimi K3 Goes Global The Robot Episode: Four Leaders on What's Coming