Chamath Says Your Portfolio Is Worth 75% Less Than You Think. Karpathy’s Data Suggests He’s Right.

---

THE NUMBER: 60-80% — the share of a typical equity valuation derived from terminal value. That’s the portion of every stock price that assumes competitive advantages persist for a decade or more. Chamath Palihapitiya just argued that AI makes that assumption unpriceable. If he’s even half right, the math doesn’t bend. It breaks.

---

Chamath Palihapitiya posted a note this weekend titled “The Collapse of Terminal Value” that should be required reading for anyone who allocates capital — including the capital of their own career. His thesis: AI accelerates disruption so fast that no company can credibly project cash flows beyond five years. Terminal value — the 60-80% of every DCF model that assumes moats endure — collapses. The S&P 500, currently trading at roughly 22x earnings with a $58 trillion market cap, reprices to somewhere between 2-7x free cash flow. At the midpoint, that’s $14 trillion. A 75% drawdown that makes 2008 look like a rounding error.

Andrej Karpathy dropped an automation exposure analysis this week scoring 342 U.S. occupations on a 0-10 scale. The data point that inverts every previous technological disruption: jobs under $35K average 3.4 exposure. Jobs over $100K average 6.7. AI is coming for the expensive workers first. Financial advisors, lawyers, software engineers, mathematicians — the knowledge workers who power the tax bases of every major American city.

Meanwhile, Elon Musk admitted what nobody in the AI race has been willing to say: “xAI was not built right.” Nine of eleven co-founders gone. Rebuilding from the foundations. The man spending billions on AI infrastructure just confessed the whole thing needs a do-over. If the people building AI can’t keep up with AI’s pace, what chance does the average corporation have?

The clock speed mismatch between machines and institutions isn’t a feature gap. It’s a civilizational fault line. And we’re standing on it.

The Death of Terminal Value — And Everything Built on Top of It

Chamath Palihapitiya’s argument sounds like a thought exercise until you follow the chain reaction.

Start with the premise: AI lowers the cost of disruption and raises the pace of innovation so relentlessly that no company can project free cash flow beyond five years with any credibility. In the time you use AI to disrupt an incumbent, someone is using a better model to disrupt you. The cycle compresses until the market stops paying for year seven, because year seven is unknowable.

Now follow the dominoes.

Terminal value isn’t an obscure line item. It’s 60-80% of most equity valuations. Every pension fund, every 401k, every endowment, every index fund — they’re all pricing in the assumption that Apple will still be Apple in 2035, that JPMorgan’s moat still holds in 2040. Remove that assumption and you don’t get a correction. You get a repricing of the entire financial operating system.

Stocks drop 50-75%. That’s not a number on a screen — it’s retirement savings halved for anyone in equities. The average American has zero savings and can’t survive two weeks without a paycheck. Their 401k was the backstop. Now the backstop is gone.

But it gets worse. Mortgages are 10-30 year instruments underwritten against two things: property value and W-2 stability. If Karpathy’s automation scorecard is right, the knowledge workers earning $100K+ are the most exposed to displacement. Those are the people who qualify for mortgages. If the market consensus shifts to “your job at Cisco is reliable for 10 months, not 10 years,” the entire consumer credit apparatus seizes. Not because of subprime lending this time — because of employment confidence. Different cause, same transmission mechanism as 2008.

Consumer spending takes a dive. Not because people choose to save — because they’re terrified. Credit card balances that felt manageable when your job was “safe” become existential when your industry scores a 7 on Karpathy’s oblivion chart. Disposable income gets hoarded for the rainy day that’s now weeks away, not years.

Here’s the part Chamath says quietly that nobody else is willing to say out loud: without terminal value, there is no venture capital business. VC returns depend on future exits at valuations that assume the acquired company has durable cash flows extending years into the future. If those future cash flows are unpriceable, the exit multiples collapse. No exits, no fund returns. No fund returns, no new funds. No new funds, no new company formation at scale. The innovation engine that’s supposed to build the AI future runs out of fuel.

The uncomfortable question: If AI spending depends on capital markets that depend on terminal value that AI itself is destroying — who funds the revolution that’s eating the funding mechanism?

Machines Run at Clock Speed. Humans Don’t. That’s the Whole Problem.

Codie Sanchez posted this weekend about the need for change agents inside companies to keep up with AI’s pace. She’s right about the diagnosis. But the prescription assumes something the data doesn’t support: that most organizations — and most people — can actually change fast enough.

Andrej Karpathy’s Autoresearch framework demonstrated AI systems executing complex research end-to-end without human intervention. Not in a lab. In production. Shelly Palmer’s newsletter on recursive self-improvement documented that AI companies are now using AI to build AI products — and the release cycles that used to take months now take weeks. Anthropic reportedly built its Cowork product entirely in Claude in a week and a half.

Corporations are quarterly organisms. Boards meet four times a year. Strategy cycles run 12-18 months. Budget approvals take weeks. AI improvement cycles are measured in days. That’s not a gap that hiring “change agents” can bridge. It’s a structural mismatch between the metabolism of the technology and the metabolism of the institutions deploying it.

The historical pattern is instructive — but not comforting. The internet was a distribution revolution. It changed howhumans connected to markets and information. But every transaction still had a human on both ends. AI is a substitution revolution. It doesn’t connect humans more efficiently. It replaces them in the chain entirely. The internet reshuffled the deck chairs. AI punches a hole in the hull.

This is where the agency question becomes existential. Adapting to AI disruption requires a specific psychological profile: risk tolerance, self-awareness, the willingness to bet on yourself when every institutional signal says “stay put.” That’s 10-15% of the population on a generous day. The other 85% aren’t going to retool. We already know this — retraining programs have a dismal track record across every previous technological transition. The people who can’t adapt won’t fall. They’ll freeze.

And frozen people don’t build companies. They vote.

The signal for your business: The Davos reality check on AI ROI made the same point from a different angle this week — tools don’t pay off until the work itself changes. You can’t bolt AI onto a quarterly planning cycle and expect machine-speed results. The companies that win will be the ones that redesign the factory, not just swap the motor. The question is whether your organization can redesign itself faster than the next model release makes the redesign obsolete.

The Casino Economy Is Already Here

Combine these two forces — collapsing terminal value and a clock speed mismatch between AI and human institutions — and you get something the financial system has no model for: permanent, structural volatility.

Not the cyclical kind that markets have always lived with. The kind where the rules get rewritten mid-hand. Where a 30-year mortgage, a pension obligation, and a career at a Fortune 500 company all become bets on a future that nobody can credibly model beyond five years.

Vinod Khosla predicted this week that 80% of all jobs will be done by AI by 2030 and $15 trillion in U.S. labor GDP is going away. Sequoia’s Sonya Huang declared 2026 the year of AGI. Karpathy’s occupational analysis puts hard numbers behind what Khosla and Huang are describing in broad strokes. The convergence of these predictions — from a VC, a researcher, and a top-tier investment firm — isn’t consensus. It’s a fire alarm.

The self-defeating mechanism is real. Chamath acknowledges it: if markets crash, AI capex dries up, disruption slows, moats restabilize. The doom loop has a built-in governor. But the more likely outcome isn’t apocalypse or recovery. It’s oscillation. Shorter cycles. Fatter tails. Higher volatility. Five years of confidence followed by a crisis of faith, followed by recovery, followed by another crisis. The VIX doesn’t just spike — it moves to a permanently elevated baseline.

Howard Lindzon has a name for where this ends up: the Degenerate Economy. The lines between investing, trading, and gambling have already blurred beyond recognition. You can parlay an Nvidia position with a basketball bet on Robinhood. Retail investors now account for 25% of all stock market trading volume — nearly double their share from a few years ago. 0DTE options, memecoins, prediction markets, sports betting — all flowing through the same digital wallet on the same phone. Lindzon’s framing is “speculation as entertainment” and “investing as a sport.” But add Chamath’s terminal value thesis and it becomes something darker: when the future is unpriceable and long-duration bets stop making sense, everything becomes a short-duration bet. The whole economy becomes a casino not because people want to gamble, but because gambling is the only rational response to a world where nobody can see past five years. More cortisol in the system. More dopamine hits chased. The degenerate economy isn’t a subculture anymore. It’s the default setting.

That creates a two-tier economy. People who can ride the volatility — cash-rich, asset-light, adaptable, high-agency — versus people locked into long-duration commitments: mortgaged homeowners, pensioners, traditional employees, anyone whose financial architecture assumes a stable ten-year horizon. The first group gets richer. The second group gets crushed. The wealth gap Elon-as-trillionaire-vs-the-average-American-at-$50K isn’t a prediction. It’s math.

Every previous technological revolution concentrated wealth before it distributed it. The Gilded Age preceded the New Deal. The robber barons needed thousands of workers to generate their fortunes. Jensen Huang needs a few thousand engineers and a lot of electricity. The ratio of capital owner to labor has never been this lopsided, and AI only tilts it further.

What this tells you: Every long-duration financial instrument in your life — your mortgage, your pension, your career plan, your kid’s college savings — was built for a world with predictable terminal value. We may be entering one without it. The institutions that adapt fastest to shorter time horizons survive. The ones that don’t become the next pension crisis, the next municipal bankruptcy, the next “how did nobody see this coming?” headline. You’ve heard this movie quote before. This is the part where Michael Burry stares at the screen and says: “I may have been early, but I’m not wrong.”

What This Means For You

Three tectonic forces are moving simultaneously: AI is compressing the time horizon over which any business can project durable cash flows, it’s displacing the highest-paid workers first rather than the cheapest, and the financial system’s load-bearing assumption — that the future is priceable — is cracking under the weight of both.

Stress-test your business against a five-year ceiling, not a ten-year horizon. If your strategy, your valuation, or your competitive position depends on assumptions about what your market looks like in 2033, you’re building on sand. The companies that survive oscillation are the ones that can generate returns on compressed timelines.

Invest in organizational clock speed, not just AI tools. The gap isn’t technological — it’s metabolic. If your company makes decisions on a quarterly cycle and your disruption arrives on a weekly one, no amount of AI spending closes the gap. Redesign the decision architecture first.

Audit your exposure to knowledge-worker displacement. If your business model, your tax base, your real estate portfolio, or your customer base depends on highly paid knowledge workers maintaining stable employment, Karpathy’s scorecard just handed you a risk map. Read it.

The financial system was built for a world where the future was predictable enough to bet on. We may be entering one where it’s not. That’s not a market correction. It’s a regime change.

Three Questions We Think You Should Be Asking Yourself

If terminal value becomes unpriceable, what happens to every financial instrument in your life that assumes it isn’t? Your mortgage. Your pension. Your 401k allocation. Your company’s valuation. Every one of these is a bet that the future looks roughly like the present, extended forward. Chamath’s argument isn’t that the future will be worse. It’s that it will be unknowable on the timescales our financial system requires. That’s a different kind of problem — and it has no precedent in modern markets.

Does your organization have the metabolic rate to survive what’s coming — or are you a quarterly organism in a weekly world? AI improvement cycles are measured in days. Corporate planning cycles are measured in quarters. Budget approvals take weeks. The companies that die won’t be the ones that chose the wrong AI model. They’ll be the ones that couldn’t change the tires while the car was moving. Be honest about which category you’re in.

What happens to democracy when 85% of the population is economically frozen in a casino they didn’t choose to enter? The agency required to thrive in an AI-disrupted economy — the ability to bet on yourself, to retool, to build something from nothing — isn’t a skill most people have. Retraining programs have a dismal track record. The people who can’t adapt won’t disappear. They’ll vote. And people who lose faith in systems don’t build new ones. They burn the old ones down.

“What happens if AI makes every moat temporary?”

— Chamath Palihapitiya

— Harry and Anthony

Sources

• Chamath Palihapitiya — “The Collapse of Terminal Value”
• Codie Sanchez on change agents
• Andrej Karpathy — Autoresearch and occupational automation analysis
• Vinod Khosla — 80% of jobs replaced by AI by 2030
• Sonya Huang / Sequoia — “2026: This Is AGI”
• Elon Musk — “xAI was not built right”
• Howard Lindzon — “The Degenerate Economy”
• Shelly Palmer — Recursive Self-Improvement
• Tomasz Tunguz — “You Are Responsible for Your Agent”
• Every.to — “What Comes After LinkedIn”
• The Davos reality check on AI ROI