With a h/t to the late historian Barbara Tuchman, from whom I borrowed the central idea in this piece, I'm pleased to share my latest long-form note to Capital Allocation Pro subscribers published initially on 8/6/26.
"Long" in this context is measured by modern (overly hurried?) metrics, with the note's main body comprising ~1,600 words. I'm unsure if it'll win me new friends--it's quite critical of certain players on the world stage--but I stand by what I've written and hope readers find the note useful even if they disagree with some or all (!) of it.
Feedback welcomed and encouraged, publicly or via email to ds@hedgeyeam.com.
Key Takeaways
In a note posted in early 2025, I discussed critically a half-dozen investment-related follies, defined as policies pursued against the interests of those pursuing them.[1] Roughly 18 months on, another crop of follies has emerged: proof, if any were needed, that folly is a renewable resource.
• Two of the “fresh follies” discussed below are matters of state: bombing Iran to bend its policies to America’s will, and betting allies will remain loyal no matter how they're treated.
• One is a matter of monetary policy: a new Fed chair pledging 2% inflation without specifying key metrics for gauging fulfillment of that pledge nor the tools to be used in fulfilling it.
• The other two are matters of technology and capital: US AI labs betting closed weights will beat an open-source Chinese onslaught, and hyperscalers betting a $500 billion-a-year habit will pay for itself on schedule.
Fresh Follies
As before, each folly discussed below meets a three-part test crafted by historian Barbara Tuchman in her 1984 gem The March of Folly:
- The policy must’ve been perceived as foolish by fair-minded observers in its own time;
- A feasible alternative course of action must’ve been available; and
- The policy must’ve been the handiwork of a group rather than just one person.
Tuchman's definition is worth restating, since it gets loosened with each retelling: folly isn't just a policy that fails. It's one whose failure was visible before the fact, for which an alternative existed, and which persisted anyway thanks to institutional momentum rather than any one person's bad judgment. The five initiatives assessed critically below aren't merely differences of opinion about odds. Each is a case where the pursuer's own stated goals arguably would be better served by a different course, and where the reasons for persisting are structural — bureaucratic, financial, reputational — rather than superior information the rest of us lack.
Folly #1 – Bombs Away
The theory, stated plainly: hit Iran hard enough and two things happen at once. One, the Strait of Hormuz (SoH) reopens for good. Two, Tehran quietly shelves its nuclear ambitions. Six months into a campaign that began with the killing of Iran's Supreme Leader, neither has happened. Iran keeps reopening hostilities after every truce and memorandum of understanding; Washington keeps needing fresh waves of strikes just to keep tankers moving; and Tehran's foreign ministry has repeatedly knee-capped further talks, most recently after multiple consecutive nights of bombing. Even the administration's own vice president has conceded the obvious: that cheap drones and mines make free passage through SoH a problem airpower alone can't solve.
The folly isn't using force; deterrence often calls for it. It's the assumption that more rubble equals more persuasion, applied to a regime whose survival instinct outweighs its balance sheet, and to a nuclear program that has already proven hard to “obliterate”. A lower-intensity mix of sanctions, verification, and actual diplomacy has been available the entire time — the vice president himself recently said so, out loud, on television. That the louder option keeps winning anyway, at the direct expense of the freedom-of-navigation and non-proliferation goals it's supposed to serve, is folly by Tuchman's own definition, no annotation required.
Folly #2 – Messing With Middle Powers
Vice President Vance's foreign-policy theory, road-tested since his chest-thumping speech at the Munich Security Conference in early 2025, boils down to this: America’s longstanding allies including Canada, Japan and NATO’s European member-states willabsorb thinner troop commitments, vaguer Article 5 assurances, tariff threats, and public scolding about their domestic politics and keep deferring to Washington anyway, because where else, really, would they go?
The answer, increasingly, is elsewhere. Canada and the five Nordic states held a “middle powers” summit in March 2026 explicitly aimed at needing Washington less — the first US-aligned bloc since 1945 to organize formally without the US in the room, which is either a coincidence or a message, and probably not the former. Germany and France have stepped up defense-industrial autonomy. NATO's own internal vocabulary has drifted, tellingly, from “burden-sharing” to “burden-shifting.”
None of this proves transactional pressure can't work as short-term leverage, but the group-level response is the precise opposite of subordination, and it costs Washington exactly the influence — favorable tech terms, dollar invoicing, intelligence depth — that made the alliance worth having in the first place. A gentler, differentiated version of the same pressure was and is available. Choosing the version that visibly accelerates the very “de-Americanization” it claims to guard against is folly, pure and simple.
Folly #3 – Controlling the Uncontrollable
New Fed Chair Kevin Warsh has pledged to get inflation to 2% and keep it there, full stop. It's a fine applause line. It's also a promise made with roughly half the necessary tools and none of the fine print. The Fed can raise or lower short-term rates and fiddle with its balance sheet; it cannot un-invade a strait, un-impose a tariff, or make a barrel of oil cheaper by committee vote. Much of the inflation currently running well above the 2% target is arriving via tariff pass-through and Iran-conflict commodity costs, exactly the kind of supply-side shock a policy rate is bad at fixing, no matter how confidently Warsh juts his jaw at a press conference.
Worse, nobody — not Warsh, not the Fed, not Congress in decades “overseeing” it — has ever specified over what horizon 2% is supposed to hold. Trailing twelve months? A rolling five-year average? Some Platonic ideal measured at the end of time, once all data revisions are in? A promise with no stated deadline isn't a policy; it's a New Year's resolution that never has to be kept, because the year in question is TBD. Until someone nails down the clock as well as the number, “We will get inflation to 2%” is less a monetary-policy commitment than a mission statement — and mission statements, as any Fed watcher can tell you, have never once lowered anyone's grocery bill.
Folly #4 – Assuming This Time Is Different
US frontier labs built their business models on a premise that looked airtight two years ago: superior compute, proprietary weights, and tight product integration would let a handful of American firms capture most of the value in large language models, the way a handful of firms captured most of cloud computing a decade earlier. Chinese labs — DeepSeek, Alibaba's Qwen, Moonshot's Kimi, Zhipu's GLM, MiniMax — spent 2026 dismantling that premise, shipping open-weight, frontier-competitive coding and agentic models in rapid, sometimes near-simultaneous succession, at a fifth to a tenth of the token cost of closed US rivals. One recent release landed within a percentage point of a leading US lab's flagship model on a closely watched agentic benchmark, at roughly a fifth of the price. US developers' token traffic to Chinese open models jumped from roughly one-tenth a year ago to as much as 46% in a given week in July, as builders quietly route routine work to whatever's cheap and good enough, keeping the expensive stuff for special occasions.
The folly isn't chasing frontier capability. That's still legit. It's assuming capability leadership alone preserves pricing power once a free, self-hostable substitute covers most use cases that don't need the actual frontier. Per-token margins are being arbitraged in real time by rivals treating weights as a loss leader rather than the product itself. A cheaper, more open commodity tier alongside a genuinely premium frontier tier was and is available to anyone willing to admit the market is splitting in two. Most labs have chosen instead to defend an increasingly leaky moat, and the bill is already showing up in compressed token economics. Frontier lab investors beware …
Folly #5 – Hyper-Hasty Spending
Hyper-scalers are on pace to spend upward of $500 billion in 2026 on AI infrastructure — data centers, chips, power — roughly ten times what their Chinese rivals are spending, on the theory that scale of investment reliably becomes scale of return. That theory is getting squeezed from two directions at once: the open-source price war described in Folly #4, which compresses revenue per token even as capacity keeps expanding, and the still-open question of whether enterprise productivity gains are showing up fast or broadly enough to justify the outlay.
The more “good enough” open models substitute for premium closed ones, the harder it gets for hyper-scalers footing the capex bill to earn it back on anything but the thinnest sliver of genuinely frontier demand — a sliver that keeps getting thinner as the competition gets better at looking, and pricing, like a commodity.
None of this makes AI spending worthless — railroads and fiber-optic cable were wildly overbuilt too, and both eventually found ample cohorts of customers, if not the ones who paid for the tracks and cables initially. The folly, as in every prior infrastructure super-cycle, is assuming the payoff arrives on roughly the same schedule as the spending, rather than after the usual interval of price compression, write-downs, and consolidation that has followed every capital-intensive buildout in memory.
A more disciplined, staged approach tied to actual unit economics was and remains available; competitive fear of being out-invested by a rival has so far proven more controlling than disciplined capital allocation, which is exactly how super-cycles like the one critiqued here tend to end.
