Robinhood CEO's Shocking Prediction On Stocks, The AI Bubble, & Gambling Controversy | Vlad Tenev

Summary

• Tenev says there isn’t one metric he obsesses over; the idea he wants Robinhood to own is broad ownership — now ~65% of US households, up from the low 50s pre-Robinhood, with a 95%+ target. His framing: “a future with relatively few owners is inherently fragile,” and the path runs through default ownership — 401(k)-style matches, Trump Accounts putting “$1,000 funded by Treasury” into every newborn’s account — plus private markets and tokenization to “force the rest of the world to catch up.”
• Robinhood grew to $125B in AUC, up 50% in 1 year, through record trading plus becoming a “financial home for life.” Q2 set record equities trading, and the latest quarter’s equities trading exceeded 2021’s GameStop peak “organically through compounding.” Robinhood crossed to #1 in retail options market share, while prediction markets — a rare first-mover launch, shipped within weeks of legalization — hit “hundreds of millions of annual revenue, our fastest growing business line of all time.” Deposit bonuses help, but Tenev says bonus-driven AATS activity is only a minority of the overall money moving in and out of the platform.
• Agentic trading has 100,000+ accounts and, per Tenev, no serious competition — “nobody non-trivial is working on agentic trading besides us.” The catch: agents sometimes refuse to trade because they are not optimized for trading, and Tenev says trading activity likely isn’t in the training data — “you don’t have agentic trading traces like you would programming traces.” He admits he doesn’t look closely at whether agent traders make money.
• His clearest bubble tell: every interview question suddenly became “when are you going to add Korean stocks,” and hedge funds are anecdotally buying AI chips just to resell them. “If a lot of the supply is being bought by speculators, that’s when you can kind of get into trouble” — even for assets with real fundamental growth like chips in an AI demand environment.
• Private markets are “the next frontier of our mission”: Robinhood Ventures Fund 1 includes pre-IPO OpenAI, Fund 2 does seed/Series A with YC at tens-of-millions valuations, and the end state is trading individual private names — requiring continued product innovation and accredited-investor reform. The premise: OpenAI and Anthropic are going into the trillions in valuation while “a small circle of wealthy insiders” captures the gains.
• Robinhood Chain is pitched as “the hottest chain in crypto right now” — top-five DEX volume, stock tokens in 120+ countries — but the US will probably adopt tokenization late. The tokens are not currently available in the US. Overseas users without functional banking leap straight to tokenized stocks; in the US, enabling 24/7 access would be “going from your very fast train to a high-speed train… you can already get from New York to DC in 2 hours. Shaving it to 1 hour maybe isn’t the biggest delta.”
• Financial advisers will likely survive AI, but their fees may compress: expect a middle ground between robo-advisory’s ~25bps and full-service’s 1%+. Responsibility is a key moat — “it’s hard to have the AI take responsibility” — even as Graham Stephan’s Claude portfolio review saved him $16k/year, which Tenev notes “doesn’t even need AI”; banks still hold trillions earning near zero because of inertia and relationship strength.
• A California billionaire tax could be an “own goal” on Tenev’s model of Hollywood’s policy-driven network breakdown — and, depending on implementation, could force people like him to sell company shares. Wealth taxes “start very very popular, but then eventually they cover everyone,” and the threat alone has already pushed significant taxpayers out. Parting hot take: more software engineers and more lawyers in 2035 than today.

Deep dive

1. Ownership—not a single metric—is the idea Tenev wants Robinhood to own: 65% of US households to 95%+

• Asked what he obsesses over instead of the stock price, Tenev rejects a single number — Robinhood now runs as a portfolio of businesses with de facto CEOs reporting to him — and lands on ownership: “a future with relatively few owners is inherently fragile.” US households owning equities sit around 65%, up from the low 50s before Robinhood, and he wants 95%+ via Trump Accounts, private markets, and tokenization that lets US assets be “distributed globally.”
• Graham’s pushback — is this an ownership problem, an education problem, or a most-people-don’t-have-enough-money problem? Tenev concedes the last and answers with default ownership: the employer 401(k) match “by itself moved ownership by tens of percentage points,” Trump Accounts mean “every single child born in this country gets $1,000 funded by Treasury,” and philanthropists like Michael Dell (targeting lower-income ZIP codes nationwide) and Ray Dalio are sponsoring their states.

2. Inside the $125B AUC year, up 50%: records everywhere, #1 in options, fastest business line ever

• AUC is the scoreboard for ownership, and growth splits between active traders and lifelong finance. Q2 set records, including equities trading; the latest quarter saw more equities trading on Robinhood than during 2021’s GameStop mania, and Robinhood has “crossed over into number one” in retail options market share even as incumbents keep merging with each other.
• Prediction markets are the anomaly: Robinhood is usually late to asset classes (crypto launched in 2018, years behind 2012-era natives like Bitstamp — which Tenev said he thought was the longest continuously running crypto exchange), but it shipped prediction markets within weeks of legalization for the presidential election. Result: “hundreds of millions of annual revenue, our fastest growing business line of all time.”
• The “financial home for life” leg runs birth to inheritance: Trump/custodial accounts, trust and joint accounts, and banking scaling to billions in assets with “really high direct deposit attach rate” — capture the paycheck, then route retirement, taxable, and strategies. Deposit bonuses help, but Tenev says they are a minority of the relevant activity: bonus-driven AATS is a good chunk of AATS activity, while AATS itself is a minority of overall money moving in and out of the platform.

3. Agentic trading: 100,000 accounts and agents that refuse to trade

• The MCP-server thesis: “anything that you can do on your phone with Robinhood, you should be able to have your AI agent do” via Claude Code, Codex, or others — launched with equities, expanded to options, crypto announced, tax-lot visibility added. Tenev’s own algorithmic-trading background frames it: that power used to require programming skill, deep legacy integrations, and $10-a-trade brokers; agents democratize it.
• Over 100,000 agentic accounts drive “decent volume,” pulling data sources from across the internet into strategies — though Tenev admits “I don’t really look too closely at their returns.”
• The episode’s best failure mode: agents sometimes decline to trade — not necessarily to protect the user; “it just doesn’t feel like it” — because they are not optimized for trading, and Tenev says, “I don’t think that trading activity is in the training data. You don’t have agentic trading traces like you would programming traces.” And the competitive claim: “nobody non-trivial is working on agentic trading besides us” — 100,000 users, not 10 million, so still early.

4. Financial advisers likely survive; their fees may compress

• Tenev’s call: humans likely persist because responsibility is a key part of the product — an adviser is “almost a member of the family,” and “it’s hard to have the AI take responsibility”; being a fiduciary “is a different bar.” Today people use AI to sanity-check advisers, not replace them. “The more likely scenario is you’ll see fee compression”: robo has consolidated around ~25bps, full-service sits at 1%+, and AI blurring the two could probably settle fees “somewhere in between.”
• Graham’s live example: he uploaded his portfolio to Claude, which flagged a fund with a higher expense ratio than a near-identical Vanguard fund plus a harvestable small loss — $16,000 a year saved after tax. Tenev’s deflation: much of this “doesn’t even need AI… it’s a very very simple algorithm.”
• The inertia point that makes it hard: despite Robinhood offering 3.5% APY, “banks still have trillions of assets that are earning almost zero because the inertia and the strength of that relationship is so high.”

5. AI disrupts trading more than passive investing — and Robinhood engineering is already agent-overseen

• His segmentation: buy-what-you-use investing (love the iPhone, buy Apple) is durable — “what are you going to ask it?” Active trading “is going to look significantly different” as AI augments technical analysis and data ingestion. Passive allocation? “I don’t think there’s juice there” — it’s already automated. Hence agentic trading first.
• Internally, Robinhood’s AI adoption is “close to 100%”: for software engineers, the workflow shifted from humans writing and reviewing code to “humans overseeing agents… seeing how many agents you can employ and can you keep them busy all the time, including when you’re sleeping.”
• Graham’s probe — could a few people run all of Robinhood from home? “It’s not our goal for sure”: someone still runs the agents, “it’s hard to imagine a world where a human is not responsible for decisions,” and the focus is leverage for existing teams, not headcount reduction.

6. No dwelling on the past — but the critics are right about coherence (and Korean stocks)

• Asked his biggest mistake of the year, Tenev resists dwelling on the past: “I don’t like thinking about the past very much… it’s a trap,” and he even “prefers to be behind and to be an underdog.” He does acknowledge postmortems and reviews. The real current problem: with so many products shipping, “how can we make sure the core experience is coherent” — every redesign “affects tens of millions of existing customers.”
• On the gambling criticism, it cuts both directions: some users want prediction markets removed (now a switch, once a support email), others complain they’re buried three taps deep. His defense: “if you look at what we actually incentivize in the product, it’s retirement” — a 3% match for Gold contributions — plus separate apps (banking, wallet) when experiences genuinely diverge.
• A concession worth keeping: “I think they’re right. I think we probably should have Korean stocks as well as stocks from every single market” — the catch-up bucket (fixed income, mutual funds, international) remains large alongside infrastructure polish and net-new innovation.

7. The CEO job at scale: the highest-signal five minutes, and talking to interns

• The evolution he describes: small-company CEO means “how quickly can I ship code”; mid-scale means learning domains outside your competency (“a lot of engineers don’t really have an interest in learning about marketing or legal or ops, and I think it’s limiting”); at scale it’s systems — “if I wanted to spend five minutes on something, how can I get the best and highest signal five minutes possible?”
• His anti-pattern for operators: relying solely on management’s assessments. Get “both ends of the barbell” — the senior-executive view and ground truth: read the code, read the shipping copy, “spend a lot of time talking to interns,” because a broken link in the chain feeds you bad information. Graham’s parallel: their friend Ben Mallah has his son live inside the apartment complexes he buys.
• Also find the quiet customers who silently stop using the product without posting to Reddit or filing tickets — and obsess: “every little problem is just going to be on my mind until I figure out a way to solve it… not ideal for staying calm and having great mental health, but actually really really good for the business.”

8. Policy wishlist: PDT repeal done, accredited-investor reform next — Ventures Funds 1 and 2

• The PDT repeal (day-trading limits under $25,000) was “very very good”: it penalized small accounts, flagged risk-management exits as day trades, and was per-broker — restricted traders just moved accounts and started from scratch elsewhere.
• Private ownership is “the next frontier of our mission”: top AI labs like OpenAI and Anthropic are “going into the trillions of dollars of valuation and there’s a small circle of wealthy insiders that’s benefiting from all the growth.” Robinhood Ventures Fund 1 is a late-stage pre-IPO basket (OpenAI is in it); Fund 2 flips to seed/Series A — typical valuations in the tens of millions — partnered with YC.
• The differentiator: every Ventures deal is direct “with the company’s blessing,” unlike some competitors doing second-layer SPV/LP interests that companies later say they know nothing about. Graham’s needle — doesn’t broad access just push the wealthy in even earlier? “There’s always a cat and mouse game… our job is never done.” End state: trading individual private names, which needs continued product innovation and accredited-investor reform.

9. Robinhood Chain is “the hottest chain in crypto” — and the US will likely adopt tokenization late

• Chain launched weeks ago: top-five in DEX volume, fast-growing TVL, stock tokens (Tesla, Nvidia) in 120+ countries outside the US — sendable like Bitcoin, swappable in active Uniswap pools, usable for lending and collateral. USDG (with Paxos) anchors it, but developers built things Robinhood “probably wouldn’t have thought to make ourselves,” like protocols airdropping stock tokens to memecoin holders. It was built for AI agents too.
• Why the US may lag: overseas, places without functional banking leap straight to tokenized stocks — “a huge leap” — while in the US, enabling 24/7 tokenization would be “like going from your very fast train to a high-speed train… you can already get from New York to DC in 2 hours. Shaving it to 1 hour maybe isn’t the biggest delta.” The tokens are not currently available in the US, and Tenev says the US is probably going to be a late adopter.
• Robinhood Social’s edge is verified trades and P&L widgets — elsewhere “you don’t really know whether they’re BSing or faking the screenshots” — and screenshots are now getting cross-posted to X. Graham’s live bug report: no profile verification yet, so similar usernames can make it unclear whether an account is really his; Tenev: “we’ll have to figure this out.”

10. Alarm bells: Korean stocks, and hedge funds buying chips to flip

• On incumbent brokerages: they “suffer from just being large.” His consultant playbook is a talent beachhead — “you can’t eat the elephant in one bite” — inject elite engineers into one circumscribed area, run it standalone, and use it as the roadmap to rebuild the whole company. Robinhood, meanwhile, captures a younger base “growing wealthier over the next couple of decades as opposed to aging out” — beneficiary of the great wealth transfer.
• The bubble tell he describes is feature-request clustering: international stocks were never top-ten, then suddenly “every question I would get in interview was, when are you going to add Korean stocks… that does raise the alarm bells a little bit.” He still plans to add them.
• On chips and energy: the fundamentals are obvious if AI demand keeps growing, but there have been anecdotes of hedge funds buying chips because they know, even if they don’t need them, they can resell them at a higher price. “If a lot of the supply is being bought by speculators, that’s when you can kind of get into trouble” — even for something with real fundamental growth.
• Is a high market bad for the young? In the 2020 COVID crash Robinhood customers bought while other discount brokers’ customers sold — younger customers “take corrections as opportunities” — though he doesn’t want one: “everyone generally should just want things to go up smoothly up and to the right, but it’s not the reality.”

11. The billionaire tax as California’s potential own goal

• The real risk isn’t personal, it’s the ecosystem: Tenev says Hollywood has “just gotten decimated in the past 10 years,” with a lot of the damage tied to policy own-goals. He warned that if filming and studios keep moving elsewhere, the network could eventually break enough that talent and award shows leave LA. Tech looked “fairly bleak” in San Francisco during COVID until “AI kind of resuscitated things” — “what happened in LA shows that we shouldn’t take this for granted.”
• The mechanics: these taxes “start very very popular, but then eventually they cover everyone”; the lost revenue from consistent moderate taxpayers who leave can exceed the one-time hit, and “a lot of people have already left the state” on the threat alone. Personally, since the bulk of his wealth is Robinhood shares, “on a cursory reading” he thinks people like him could have to sell shares in their companies. There’s even a Robinhood prediction market on it — he could technically hedge, and it currently shows minority support.

12. How Vlad trades, why he’s long Bitcoin, and defense in depth

• His own shares aren’t held at Robinhood — not egg-basket caution, but because his and employees’ shares are administered under ESPP and 10b5-1 requirements; he notes companies provide this kind of administration as a B2B service. Personal trading: “I have to be really careful trading individual stocks” because of the restrictions around material nonpublic information; ETFs are clean, crypto is “a fun trading product,” and he holds “a pretty diversified portfolio” (ex-Redditor, sadly no longer posting).
• On Bitcoin: “It was the first asset. It was the original memecoin… nobody can ever take the position of being the first and most trusted.” Long-term bullish on that singularity — “but I can’t tell you if it’s going to hit a million or anything like that.”
• On custody risk: “defense in depth” — the lion’s share in cold storage so a hot-wallet breach touches only a small minority of assets — plus a new reality where AI tools make vulnerabilities cheap to find: “you can’t rely on it being just hard to find or it being too expensive.” Answer: penetration testing, internal red teaming, and turning AI attackers on yourself, while banks holding trillions still have a “relatively poor cybersecurity posture.”

13. Manufacture hunger, journal the week, and bet on math generalizing

• On raising kids: his parents told him directly “we don’t have the money to send you to college,” and visa status meant failure could send the family back to Bulgaria — whereas for his own children, “it can be hard to manufacture hunger if it’s not really there.”
• His journaling loop — what went well, what could have gone better, a ranked priority list carried week to week — exists to escape being “completely input driven”: “you wake up in the morning and you’re like, let me look at what other people want from me… the risk is you just don’t get the important things done.” Even crises, he argues, are influenced by mindset: stressful at the time, “almost fun in retrospect.”
• Against the well-traveled-networker parenting thesis (Chris Camillo’s, relayed by the hosts), he cites Peter Thiel — “I’m much more worried about the math people than the word people,” with Thiel’s explanation tying math’s recent egalitarian advantage to progressive trends, possibly reversing — but his own view is that “math generalizes”: abstract math is pure problem-solving that trained him for a business career he never studied for. It’s why he started another company a couple of years ago building “mathematical superintelligence”: a system truly great at math “can basically be really good at any economic activity, starting with coding… but probably it can write you a better history essay eventually as well.”

14. Reverse interview — and two hot takes for 2035

• Consulting the hosts on their own show: “inspiring guests” is a north star you’d hear from many podcasts; the gap is differentiation. His example is Acquired — “I just don’t think anyone else does exactly what they do… I know exactly what it is,” a four-hour deep dive on one company — and he suggests sharpening their edge: finance questions put to successful people you wouldn’t associate with finance.
• On Robinhood as a media company: after acquiring MarketSnacks, which became Robinhood Snacks, and building Sherwood, he now favors platform over first-party — “it’s probably unrealistic for the best content to be created first party” — so be a neutral, meritocratic conduit for creators with “no hidden agenda or hidden favoritism that can tip the scales… and get you demonetized without you understanding why.” Both network products (Chain, Social) are “month one.”
• Rapid fire: the biggest investing myth is waiting until you have savings and knowledge — compound interest means “the earlier you start, the better off you’ll be… investing at age zero.” Best product builder outside Robinhood: Spotify. What surprises by 2035: how far private markets get democratized — plus the closer, “I think there will be more software engineers and more lawyers in 2035 than today,” which Graham jokes is a dangling carrot.