Where the Empires Go

So where does this actually go?

I ended the last part with that question hanging, and I want to be honest before I try to answer it: I don’t know. Nobody does. Anyone who tells you they can call the next twenty years of this is selling something. But “I don’t know” is not the same as “anything could happen.” There’s a disciplined way to think about an uncertain future and a lazy way, and the gap between them is most of what I care about here. So let me walk through the method I actually trust, in order, and then point it at the kings.

Start with the base rate, not the story

Here’s the single most useful habit I’ve stolen from people who forecast for a living: before you reason about a specific case, look at what usually happens to cases like it. The outside view first, the story second. It feels backwards, because the story is vivid and the base rate is boring. But the boring number is almost always the better starting bet.

So what’s the base rate for a dominant company staying dominant?

It’s grim, for the company. The average tenure of a firm in the S&P 500 has collapsed: around 61 years back in 1958, down to 24 by 2016, and on track for roughly 12 by 2027.1 Only about 10 to 12% of the companies that were in the 1955 Fortune 500 are still on the list at all.1 Sit with that. The default outcome for a giant is not “reigns forever.” It’s gone in a couple of decades, eaten or merged or shrunk into irrelevance.

Which means the prior is correction. If you know nothing else about a specific empire, the smart bet is that today’s titan is tomorrow’s cautionary slide in someone’s deck. Entrenchment is the claim you have to argue up from there. It isn’t the natural resting state; it’s the exception you have to justify. A lot of the doom writing about permanent tech overlords skips this step entirely. It starts from “they’re huge now” and slides straight to “they’ll own us forever,” without ever asking what the survival curve looks like. The survival curve says most of them won’t.

That’s firms. States are a different animal, and this is where the method earns its keep.

The one transition that doesn’t open a window

For a consolidated state, the base rate flips. Personalist leaders run a long time, sure, but that’s not the interesting part. The interesting part is how they leave. And here’s the finding that genuinely changed how I think about this: when a strongman dies in office, the apparatus he built usually survives him.2 Death, the one transition you’d assume is the great reset, is the transition that most reliably doesn’t open a window. The machine keeps running. The Party, the security services, the inner circle… they close ranks and carry on, because their survival was never actually tied to his pulse.

So the cold version of forecasting succession for these people is almost insultingly simple: print their ages.

Murdoch, born 1931. Trump, born 1946. Xi, born 1953. Ambani, born 1957.

Succession is a mortality table. That’s it. But the table reads completely differently depending on whether you’re looking at a firm or a state. For a firm with no obvious heir, the founder’s death is a genuine shock to the share price, a real reset. For a consolidated state, the same death is a transfer, not a reset. Key-man risk is loud for the corporate cases and quiet for the entrenched political ones. The mortality table is identical; what it predicts is the opposite. That asymmetry is the whole reason I keep insisting on separating the company kings from the country kings, even though the three-lever pattern looks the same across both.

What actually predicts entrenchment vs correction

Base rates give you the starting odds. Leading indicators tell you which way a specific case is drifting off them. Two of them do most of the work, and the nice thing is they’re both observable in public, today, before the outcome lands.

The first: are the term and age limits coming off? When a rule that exists specifically to force turnover gets quietly deleted, that’s the clearest single tell that a case is arguing its way up from the correction base rate toward entrenchment. We’ve seen this twice in this series already. China’s legislature voting 2,958 to 2 to scrap the presidential term limit.3 The motorsport federation from Part 1 voting away its own term limits a few days before I’m writing this. The mechanism is always the same: the rule that was supposed to end the reign gets removed by the people the reign benefits.

The second, and this is the one I find more useful as an investor: is the cashflow hostage to a price the king can’t set? A rule-book king whose money depends on some exogenous number (an oil price, an ad market, a media-rights bubble) is one shock away from a correction no matter how captured his institutions are. Captured courts don’t help you when the underlying revenue evaporates. The clearest live example is the Gulf. The IMF puts Saudi Arabia’s fiscal break-even at roughly $90 a barrel, and Bloomberg Economics puts it closer to $111 once you fold in the sovereign fund’s mega-project spending.4 When Brent trades well under that, the most powerful chequebook on earth starts writing smaller cheques. We’ve already watched it happen: the Saudi fund is winding down its golf-league funding and pivoting capital back home.5 All the rule-book power in the world doesn’t change a break-even price. That’s the tell that the money leg is fragile even when the power leg looks bulletproof.

So the dashboard is short. Limits coming off pushes toward entrenchment. Cashflow chained to an exogenous price pushes toward correction. Read those two against each other for any given king and you’ve got a real, falsifiable lean instead of a vibe.

Scenarios, not a prediction

Here’s the part where the lazy forecaster picks a winner and the disciplined one refuses to. You don’t predict one future for a system this contingent. You lay out the handful of futures that are actually plausible, name the fork that decides between them, and then watch for the signposts that tell you which branch you’re on.

Two axes do it cleanly. Are the correction mechanisms alive or captured (courts, legislatures, press, antitrust, a real opposition)? And does the cashflow hold or crack (the money leg from the levers)? Cross them and you get four worlds:

                     CASHFLOW HOLDS        CASHFLOW CRACKS
                  ┌────────────────────┬────────────────────┐
  CORRECTORS      │     Entrenched     │     Hollow Crown    │
  CAPTURED        │     Sovereign      │  (power, no money)  │
                  ├────────────────────┼────────────────────┤
  CORRECTORS      │  Contested Throne  │ Creative Destruction│
  ALIVE           │ (rich, checkable)  │  (the old default)  │
                  └────────────────────┴────────────────────┘

Entrenched Sovereign is the danger case: the institutions are captured and the money keeps flowing, so nothing internal or external can dislodge the king. Hollow Crown is the Gulf-squeeze world, where the power is intact but the cheques are bouncing, so the empire shrinks from the wallet inward. Contested Throne is rich-but-checkable: the king has the money, but the courts and the ballot box still bite. And Creative Destruction is the historical default we started with: a better mousetrap, a cheaper substitute, and the giant just gets eaten.

The point of the grid isn’t to pick a quadrant. It’s that different kings sit in different quadrants, and the same king can slide between them as the signposts move. Musk and Tesla, with no obvious heir and live antitrust everywhere, lean toward the bottom row, the checkable, eat-able worlds. (Worth a caution: Tesla’s board publicly denied the reports that it was hunting for a successor, so don’t read “no one else can be Elon” as some board confession. It isn’t.6) The AI-compute chokepoint is the strongest entrenchment case in the whole catalogue. One foundry makes the overwhelming majority of the world’s leading-edge chips,7 which is about as locked-in as power gets. And yet the correctors are wide awake even there: antitrust scrutiny is already landing on the dominant chip designer.8 So the danger and the brake are firing at the same time. That’s not me hedging. It’s genuinely the honest read: the system is concentrating and self-correcting at once, and which force wins is unsettled.

The US disperses, China re-concentrates, India is the fight

Now point the method at the three big arenas, because they don’t behave the same way at all, and the difference is the most important thing in this whole series.

The United States is the slow, bottom-up corrector. For more than a century it has let private power pile up and then, eventually, partially, leaned on it: Standard Oil broken in 1911, the banks re-regulated after 1929, Wall Street half-releashed after 2008.9 The corrections are real, but they’re slow, lopsided, and they never quite stick. The pieces drift back together, the way I argued in Part 2. Power, when it gets checked here, tends to get pushed outward and dispersed, imperfectly, by courts and elections and the press.

China is the opposite, and you have to be careful how you say this. China also corrects concentrated private power, and hard: the 2020-22 crackdown on its own tech tycoons was real and brutal, a record IPO killed days before launch, multi-billion-dollar antitrust fines, a whole tutoring sector erased.10 Beijing frames every piece of that as legitimate governance: anti-monopoly enforcement, data sovereignty, cutting household costs, reining in financial risk.10 And there’s a serious version of that argument, Mancur Olson’s “stationary bandit” idea, where a durable authority that disciplines its own oligarchs is performing a real public function, not just settling scores.11 But here’s the structural difference scholars like Susan Shirk and Branko Milanovic point to: in China the corrector is the party-state, regulator and sovereign in one.12 So when it disciplines a rival concentration, the power doesn’t disperse outward the way it does in the US. It gets pulled up, into the Party.13 Two facts, on this reading, hold the whole contrast in one breath: the American correction split one giant into 34 companies; the Chinese correction removed the limit on one man.14

That’s the cleanest way I’ve found to answer “is concentration unavoidable?” Maybe it is. All three systems produce it. What’s not fixed is where the correction sends the power: out, in the US, or up, in China. Running the cycle again is not the same as the power actually spreading out.

And India is the contested middle, the one genuinely undecided case, which is exactly why it’s the interesting one to me. The signals point both ways, and I’m going to lay both out rather than pretend I know the answer.

Toward self-correction: promoter holdings across the Nifty-500 fell to a record low of about 49.5% in March 2025, so the founding families are, in aggregate, holding less of the market, not more.15 The big US fraud case against the Adani group has not been closed on the promoter’s terms. A federal judge refused to quickly wave away the government’s move to drop it, calling the request “terse, bland and conclusory,” and gave prosecutors until July 13th to respond, days from now as this posts.16 The Adanis deny all of it, the charges are allegations and not convictions, and I want to be clear I’m describing a live court process, not a verdict. And at home, India’s own market regulator, SEBI, finished its Hindenburg-driven investigation in September 2025 and cleared the group, holding that the allegations were “not established,” which the Adanis call vindication.17 Keep both of those in view at once: an open foreign indictment and a closed domestic clean chit. The Reliance-Disney streaming venture formally walked away from a cricket media-rights deal it couldn’t justify, and the cricket body reopened the bidding.18 And the markets and competition regulators have been visibly assertive. Those are real brakes, biting in real time.

Toward entrenchment: India’s demographic dividend is still ahead, with the working-age share peaking around 2030, and a tide like that tends to enrich whoever already holds the rule-book.19 The competition regulator has chosen a deliberately hands-off posture on AI, consciously not intervening so a domestic ecosystem can form, which is sensible industrial policy and also a way of letting concentration build unchecked. And if the 2029 election produces another dominant result, the entrenchment case gets a lot stronger. That last one is speculation, and I’m flagging it as speculation. I genuinely don’t know how 2029 breaks, and anyone who claims to is guessing with confidence.

One more thing on India, because it’s where the base-rate discipline gets tricky. You cannot just import the American survival curve here. Indian family-promoter firms persist far, far longer than that brutal S&P churn rate implies; the dynasties have outlasted everything. So if you naively applied the US base rate to India you’d massively over-predict correction. The outside view still rules, but you have to use the right outside view, the reference class that actually matches. That’s the part the lazy forecast always skips.

So what do you do about it?

Put the method together and the verdict comes out probabilistic, not apocalyptic. The base rate is correction. Most giants get eaten. Entrenchment is the harder claim, and it’s strongest exactly where the power is structural rather than personal: the AI-compute chokepoint, not any one charismatic founder. And even there the correctors are awake. The US pushes power out, slowly. China pulls it up, fast. India is genuinely up for grabs. Nobody is permanently winning yet, and the empires that look most permanent are often the ones quietly bleeding from the cashflow leg while you’re staring at the rule-book leg.

Which leaves the only question that matters for someone like me, and probably like you. You’re not going to buy a federation or chair a sovereign fund. You have some savings and no power. So if the world really is sorting itself into a handful of kingdoms that keep reconstituting after every correction, what, concretely, do you do with that as an investor? Do you own the rails the concentration enriches? Do you bet on the correction instead? Can you even tell the difference between a real moat and rent-seeking that’s already priced in?

That’s the last part. And it’s the one I’ve been most reluctant to write, because it’s the one where I have to put my own money where this whole argument has been pointing.

References

Footnotes

  1. Corporate-longevity figures. The ~61-year average tenure of the late 1950s traces to McKinsey / Richard Foster’s work; the fall to ~24 years by 2016 and the projected ~12 by 2027 are Innosight’s (2021). Only roughly 10-12% of the 1955 Fortune 500 remain. ↩ ↩2

  2. From the autocratic-regimes literature on how personalist rule outlives the ruler: New Data on Autocratic Regimes (Geddes, Wright & Frantz, Perspectives on Politics, 2014). Death in office is the transition least likely to end a consolidated regime. ↩

  3. China removes presidential term limits, enabling Xi Jinping to rule indefinitely (NPR, Mar 2018); the National People’s Congress vote was 2,958 to 2 with 3 abstentions, reversing the limit Deng Xiaoping had introduced to prevent over-concentration of power. ↩

  4. Saudi Arabia’s path forward amid lower oil prices (IMF / Bloomberg Economics, Dec 2025); the IMF puts the fiscal break-even near $90/bbl, and Bloomberg Economics near $111/bbl once PIF mega-project spending is included. ↩

  5. Saudi PIF to end funding for LIV Golf (CNBC, Apr 2026); the fund is ending LIV funding after the 2026 season (~$5.3bn cumulative since 2021) and pivoting capital back toward domestic investment. ↩

  6. Tesla’s board reportedly sought a successor while Musk wheeled around Washington (TechCrunch / Washington Post, May 2025); chair Robyn Denholm publicly denied the successor-search reporting, so it should not be treated as a board admission. ↩

  7. TSMC 2025 update: riding the AI wave amid global expansion (SemiWiki, 2025); TSMC’s market share at the leading edge (3nm and 2nm) is upwards of 90%. ↩

  8. Nvidia gets DOJ subpoena in escalating antitrust probe (Bloomberg Law, Sept 2024); note Nvidia said it had not in fact been subpoenaed (CNBC, Sept 2024). ↩

  9. The US “concentrate, correct, drift back” cycle: Standard Oil (Wikipedia) on the 1911 dissolution; Glass-Steagall legislation (Wikipedia) on the 1933 banking response (repealed 1999); and the post-2008 Dodd-Frank rollback ($50bn to $250bn SIFI threshold, 2018) (Yale Journal on Regulation, 2018). ↩

  10. The speech that scuppered the world’s largest IPO (Fortune / Reuters, Nov 2020) on Ant’s ~$34.4bn IPO suspended days before launch; Alibaba fined $2.8bn in landmark antitrust ruling (Caixin, Apr 2021); Double Reduction Policy (Wikipedia) on the 2021 for-profit tutoring ban. Beijing’s stated rationale for each: anti-monopoly enforcement, data sovereignty, household-cost and inequality reduction, and financial-risk control. ↩ ↩2

  11. Mancur Olson’s “stationary bandit” argument, that a durable, settled authority disciplining rival power can function as governance rather than mere predation: The Rise and Decline of Nations (Mancur Olson, Wikipedia). ↩

  12. China in Xi’s “New Era”: The Return to Personalistic Rule (Susan L. Shirk, Journal of Democracy, 2018); the framing that China’s correction of private power pushes authority up into the Party rather than dispersing it is Shirk’s, and is contested. ↩

  13. Capitalism, Alone: The Future of the System That Rules the World (Branko Milanovic, Harvard University Press, 2019), on the distinction between liberal and political capitalism, and the warning that the US risk is drifting toward the latter, with wealth and political power fusing in one elite. ↩

  14. The two corrections side by side: the 1911 Standard Oil dissolution into successor companies, Standard Oil (Wikipedia), versus the 2018 removal of China’s presidential term limit, 2018 amendment to the Constitution of China (Wikipedia). Successor-company counts vary by source (commonly cited as 33-34). ↩

  15. Promoter holdings in Nifty-500 drop to record low of 49.5% in March 2025 (Tribune India / Motilal Oswal, 2025). ↩

  16. US federal judge questions DOJ decision to drop Adani charges (Al Jazeera, Jun 2026); Judge Garaufis declined to quickly grant the dismissal, called the request “terse, bland and conclusory,” and set a 13 July 2026 deadline. The US (EDNY) indictment charged Gautam and Sagar Adani with securities-fraud and wire-fraud conspiracy and securities fraud (not FCPA), all of which they deny; see DOJ EDNY indictment (US DOJ, Nov 2024). The SEC civil side settled for ~$18m with no admission: Adani settlement (CNBC, May 2026). ↩

  17. SEBI clears Adani Group, dismisses Hindenburg allegations (Business Standard, 18 Sept 2025); India’s market regulator found the show-cause allegations “not established” and disposed of the proceedings. ↩

  18. ICC faces major crisis as JioStar steps away from its billion-dollar India media-rights deal (iSportConnect, 2026); JioStar told the ICC it could not continue the ~$3bn deal and the ICC reopened bidding. ↩

  19. Imminent end of ‘demographic dividend’: India’s working-age share set to fall by 2036 (ThePrint, citing official projections); the working-age share is projected to peak around 2030. ↩