1An executive order turned psychedelic therapy into a listed equity trade with an FDA date attached
Felix, on Felix & Friends, built the day's only stock pitch on a policy change rather than on a product. An April 2026 US executive order fast-tracks psychedelic-assisted therapy and attaches $50m of research funding; RFK Jr is pushing an alternative-medicine agenda at HHS; bipartisan bills on mental health and veteran care are moving alongside. What lifts this above a policy story is that the commercial proof already exists, and it belongs to a company he no longer owns. Johnson & Johnson's Spravato nasal spray, at $590 a dose taken twice weekly, reached roughly $2bn of annual revenue in the first quarter of 2026, up about 50% year on year. Insurers pay for it. That was the part in doubt, and it is no longer in doubt.
The three names are all pre-revenue. (Compass Pathways) has a single drug, COMP360, with Phase 3 complete and an FDA decision expected late 2026 or early 2027. runs a five-drug platform, is backed by Peter Thiel, merged with Beckley in December 2025 — and holds a stake in while licensing its drug in some territories, so the first two names are not independent bets on the same outcome. (GH Research) has one compound, 5-MeO-DMT, whose commercial argument is a thirty-minute clinical session against six to eight hours for COMP360; if the drug works, the shorter chair time is what determines how fast clinics can actually run it.
The caveat comes from the person making the call, and it is severe. His own filter is profit per sale, a patent life of fifteen years or more, and a drug that is already selling — and all three names fail the third test outright. He calls the basket "a calculated speculation," not an investment, and sizes it at 1-3% of a portfolio across all three combined. He names the ways it breaks: FDA approval is not certain, capital raises will dilute holders before any commercial launch, the political sponsorship leaves if the personnel do, and Johnson & Johnson has an incumbent's reasons to push back on the regulatory path. Two further things to price in. The roughly $400bn market size he cites is an addressable-population figure drawn from the global burden of mental illness, not anybody's revenue forecast. And the video carries a paid affiliate promotion for a trading product, with framing — a billion-dollar market, get in now — that runs considerably hotter than the risk list underneath it.
The consequence is a position whose outcome is set by a regulatory calendar rather than by a business. There is no quarter to check against, no revenue line to model. There is a decision date, and it lands late this year or early next.
2The bubble argument moved from price to accounting
The week's most interesting bear case, on the Excess Returns weekly recap with Jack Forehand and Matt Ziggler, was not about valuation at all. Edward Chancellor's argument is an accounting one: GPU depreciation schedules have been stretched from three to three-and-a-half years out to six and a half. That lifts reported profits now and defers the charge onto later income statements. If he is right, the correction does not arrive as a price event you could time — it arrives as an earnings revision two or three years out, in companies that look profitable today.
His historical case is the 1845 railway mania, where capital spending reached 10% of UK GDP, a larger share than AI spending today, and three separate lines were built to duplicate the London-to-Peterborough route. The railway index eventually lost 60%. The detail worth keeping is what did well instead: canal stocks, the supposedly obsolete loser, outperformed railways from 1845 to 1850. He pairs it with a dotcom parallel — WorldCom's claim that internet traffic was doubling every two months, against an actual doubling every six — as an example of a demand statistic that went unchecked while capital was committed against it.
Andy Constant, formerly of Bridgewater and Brevan Howard, frames a bubble as a regime rather than a date: "something new" (ChatGPT in January 2023), then escalation events (the Fed easing during the spring-2023 banking stress, setting its inflation mandate aside), then a peaking phase — which he says is where we are now, and which "can last quite some time."
Jeremy Grantham, in the same recap, declines to confirm it. His own top signal is what he calls the scream from the stomach: at the tops of 1929, 1972, 2000 and late 2021, the prior year's market leaders began falling while the index kept climbing. Asked whether he sees it today: "Absolutely not." High-beta names rolled over briefly ahead of first-quarter earnings and came back hard after the beats. So one guest places us in the peaking phase and another says his own peak indicator is not flashing, in the same episode — and what would settle it between them is concrete and watchable: leaders declining while the index rises.
Grantham's other two arguments cut in opposite directions. He thinks mean reversion still works on the bottom 90% of the market but that a "novel elite" of monopolistic megacaps has escaped it, on winner-take-all software economics and permissive antitrust — with the uncomfortable corollary that monopoly margins rise while economic growth slows, which is the pattern US growth has shown across the last twenty years against the twenty before. And he thinks AI eventually becomes a cost of doing business rather than an edge: two to three years of advantage for early adopters, then margins normalise, the way prime mini-computers did for asset managers in the 1970s.
The tape carried a cruder version of the same worry. Bloomberg's language, via Yahoo Finance, was "borderline mania" and "bubble-like euphoria" in semiconductors — the price-based framing that the accounting argument is trying to replace.
3The index barely moved while single names swung more than 20% on their own news
Futures eased off record levels: the S&P 500 down 0.56%, the Dow 0.71%, the Nasdaq 0.62%, with the VIX up 6.78% to 18.43. Underneath that calm, the moves were enormous and idiosyncratic. fell 5-6% pre-market on Samsung supply and labour disruption. jumped 24% on an AI optical-interconnect partnership. rose 22.93%, 13.24% after an earnings beat, 5.05%, and 12% on a first-quarter beat. crossed $5.5trn of market value with earnings due Wednesday.
The move comes with a warning attached: the same feed carried it as both +24% and −22% within the day's headlines, so the direction is unresolved in the source material and only the magnitude is reliable. Worth knowing before treating a percentage as a fact.
One structural item arrived as a news headline rather than a market move. CNBC led with AI's electricity appetite reviving nuclear power, describing it as a forsaken energy source coming back. That is the same capital-spending story presenting as a power-procurement problem instead of a chip-supply one — and power plants have build times that GPUs do not.
The shape of the day is the point. An index that barely moves while its components swing 20% is not a market with a single view on AI. It is a supply chain in which each link is being repriced on its own news, in both directions, at once.
4Private credit's risk moved from the banking system to the retail saver
Mark Rubenstein, in the same Excess Returns recap, makes the case that private credit is not the next global financial crisis: it is tiny in the context of the overall financial system, and redemption gates function as designed when the money behind them is institutional. What has changed is who is behind them. Retail and mass-affluent investors are increasingly locked in through their advisers, and a gate that reads as prudent fund structure to an institution reads as a frozen account to a saver who thought the money was reachable. He points to Blue Owl's share-price collapse and its founder's loan-collateral problems as the visible stress, and to the knock-on for middle-market companies that depend on this channel for financing. His conclusion is conditional rather than alarmist: not systemic on its own, dangerous in combination with another shock.
His other argument is a rotation one — long Europe and long financials against the US market, on the view that the structural divergence between the two has been overdone and that European innovation is undercovered. His examples are DeepMind in the UK sitting inside , XTX in London competing with Jane Street, and Revolut likely to list in the US while remaining UK-domiciled. On the sector itself his line is that growth is the enemy of financials: what you want from a bank is utility-like behaviour, not growth-chasing.
The consequence is that the interesting question about private credit is no longer the size of the asset class. It is the liquidity expectation of the people who now own it, and that is not something a spread tells you.
What the sources recommended
Three buys, all from one video, all pre-revenue, and the person making them explicitly calls the basket a speculation rather than an investment.
The buys. , and — the psychedelic-therapy basket, capped at 1-3% of a portfolio in total across all three, with FDA approval of COMP360 as the load-bearing catalyst and dilution, political change and incumbent pushback named as the ways it fails. The internal detail matters more than the count: holds a stake in and licenses its drug in some territories, so three tickers deliver less diversification than three tickers imply. Conviction runs highest on , which has the completed Phase 3 and the nearest decision date, and lowest on , whose case rests on a shorter treatment session for a compound still to be proven.
The disposals. One, and it was a disclosure rather than a call. Felix sold after a 45% gain on this same thesis and said directly that it is not a reason for anyone else to sell — a psychedelic drug at $2bn of revenue barely moves a company of that size. He was not exiting the theme; he was exiting the wrong-sized vehicle for it. That is why the name sits in the record as informational rather than on the exit list.
The depreciation argument gets its first read on Wednesday night
The week's scheduled catalysts are corporate from end to end — no FOMC meeting, no CPI print, no payrolls report — so earnings are the only data the market gets. and report Tuesday, and Wednesday, and Thursday, then on 25 May, on 26 May, and on 27 May, and and on 28 May.
Wednesday after the close is where two of the arguments above get marked at once. Capital-spending commentary tests the supply-chain read from the third claim; the useful-life assumptions behind the depreciation line test Chancellor's from the second. Both live in the same call, and neither has been settled by anything the tape has done so far.
Two unscheduled items sit alongside. Trump meets his security advisers on Iran on Tuesday, and Iran's stock market reopens the same day.
