May 2026 Random Ramblings

Summary

• Walker’s market read is a show tune: nothing dents this tape. He keeps hearing Crazy Ex-Girlfriend’s “We’ll Never Have Problems Again” as the market screams through all-time highs despite “the war with Iran and energy prices skyrocketing,” deficits, and rising rates — and it’s hyper-concentrated: the S&P is up ~10% YTD, with roughly five percentage points attributed to a group he calls five stocks (he names Google, Nvidia, Micron, and AMD) and more than all of it from the top 20.
• Memory valuations look “really really stretched” even granting the supply crunch. Names at ~20x tangible book will “probably earn 150 over the next 2 years” on a book of 50 — and still trade at 4-5x two-years-forward tangible book after the profit cycle. “It’s just not sustainable”: supply comes from Chinese fabs, the oligopoly breaking, hyperscalers building their own memory, or engineering around it. He emphasizes that he is an AI optimist, not an AI skeptic; his concern is valuation and cyclicality.
• The “cyclical to structural” chorus from sharp VCs is his loudest warning sign. It echoes 2006-07’s “macro cycles are dead” — which he says may have avoided a recession only to produce “basically a depression instead” — and “the marginal price can’t be 5,000 times the marginal cost forever… We’re going to have a cycle, and I feel pretty confident in saying that.”
• On AI and investing careers: “It feels very much like we’re cooked.” An investor’s job can be framed as pattern recognition or analysis, and AI beats both — “I can read four 10-Ks in a day, maybe. AI can read 400 10-Ks in 4 seconds.” He acknowledges quantitative investing has already been beaten by machine learning for decades; his concern is qualitative investing’s edge diminishing. His one counterpoint: relentless AI optimization may make markets more fragile, so alpha could migrate to whoever can “wait for the fragility, wait for the left tail.”
• Your 2016-2023 letters could predict your AI take — and that lack of flexibility bothers him, including his own. He notes that many famous short sellers are short CoreWeave over depreciation, economics, and circularity arguments; tech-focused investors often embraced “AGI within 4 years” and cycles-over. He wonders why macro investors did not pivot to revolutionary AI views, or tech investors to an overhyped-AI view. He audits himself: early to the tools, yet he shorted no SaaS, bought no semis, and made no money on power trades despite a history in bankrupt power companies.
• MicroStrategy is now effectively the preferred-equity market — and he’s positioned long Bitcoin / short MicroStrategy. MSTR was 8% of equity issuance in 2025, 10% so far in 2026, and 60% of preferred issuance this year (he thinks 33% last year); if Bitcoin trades from 75,000 to 50,000 against $10B+ of 10% preferreds, the “perpetual motion machine” inverts and could “blow up the preferred equity class.”
• A CFO moving from a ~$500M company to the same role at a ~$100M company is a “super unique” signal. Via Artem Fokin: the move would probably mean less pay and can read like “calling your shot” on the smaller company’s equity — though it could also mean he was overmatched or fleeing “a very rocky ride”; “probably some combination of all three.”

Deep dive

1. Nothing dents this tape — “We’ll Never Have Problems Again”

• Walker’s framing device is a song from Crazy Ex-Girlfriend sung by a toxic on-again couple — “We’ll Never Have Problems Again” — because nothing seems to matter to this market: “the war with Iran and energy prices skyrocketing,” deficits, rising rates, and still “the stock market just seems to go up and up and up” through all-time highs. He sees risks and thinks market-wide valuations are more stressed, though not every individual valuation is stressed.
• His energy counter to the worriers: the Strait of Hormuz deadline keeps slipping (“if this isn’t open by the beginning of April, it’s going to be a disaster” — then end of April, then mid-May), and oil can adjust — at 100 “there is a lot of energy that can come online that isn’t economic at 60, 70,” plus demand destruction. His honest close: “I just don’t know, man.”
• The concentration numbers: the S&P is up ~10% YTD, with roughly half the gain coming from a group he calls five stocks; the four he names are Google, Nvidia, Micron, and AMD. Expanding to 20 names, “more than all of the S&P’s gains have come from these 20 stocks.” Underexposed? “You’re underperforming basically.” Meanwhile some consumer staples sit near “global financial crisis levels,” at ~7x free cash flow, with no earnings cliff.
• On the space frenzy — everything space-related up ~50% in ten days as the SpaceX IPO approached — he’d dismissed front-running it (“the markets aren’t that easy”), and concedes: “turns out I guess the markets are that easy.” Not dot-com extremes, but “buy memory” and watch it rise 20% a week — perhaps conservatively — or buy any semiconductor stock while detailed value work goes ignored: “very frustrating” for an active investor.

2. Memory at 20x tangible book: supply always comes online

• The arithmetic he can’t get past: memory players at ~20x tangible book; on a book of 50 they’ll “probably earn 150 over the next 2 years” — yet still trade at 4-5x two-years-forward tangible book after that huge profit cycle. “It’s just not sustainable.” He stresses that he is an AI optimist, not an AI skeptic; the valuations and cycle dynamics are what concern him.
• The supply-response menu, as he lists it: Chinese fabs; the memory oligopoly (“eventually they always break and they always bring supply online”); hyperscalers building memory themselves because “it’s a bottleneck. It always gets built”; or “engineering around memory in some way.”
• Even if AI demand does not decline, he says its rate of change must eventually slow; otherwise capital will overbuild the industry, and the marginal price cannot remain 5,000 times marginal cost forever. Eventually, “we’re going to have a cycle.”
• What worries him most is who’s saying cycles are dead — “really sharp VCs and market historians… who I’ve got a lot of respect for” declaring memory and semis have gone “from cyclical to structural… permanent demand forever.” The precedent: 2006-07’s “macro cycles are dead.” He says maybe those forecasters were right about avoiding a recession, because the result was “basically a depression instead of a recession.” His categorical call: “We’re going to have a cycle, and I feel pretty confident in saying that.”

3. “We are cooked” — unless AI optimization makes markets fragile

• His decomposition: an investor’s job can be framed as pattern recognition or analysis — “two sides of the same sword” — and AI beats both: “I can read four 10-Ks in a day, maybe. AI can read 400 10-Ks in 4 seconds,” with “all the investing experience in history” backtested near-instantly. Today’s hallucinations aside, “the AI we’re dealing with 3 years from now, 5 years from now, I don’t know. It feels very much like we’re cooked.”
• He acknowledges the pushback that quantitative investing has already been beaten by machine learning for decades. His worry is that AI increasingly comes for qualitative investing too, diminishing the edge and leaving less alpha for everyone.
• The Buffett angle, as he recalls it: “if you’re an investor with 60 IQ points, you should go and sell 30 IQ points because being too smart can be a liability” — so do swarms of 200-IQ AIs create opportunity? He doubts that humans can simply set traps for them: “AI is probably going to have seen all your tricks.”
• His live hypothesis: optimization breeds fragility. LTCM is his classic example: he recalls its blowup being associated with what they called a “17-standard-deviation move,” described as once every 10,000 years. In markets, he says, similar moves can arrive every three or five years — “I don’t know” — because markets have fat tails. If AI compresses short-term alpha, the residual alpha may belong to whoever “can wait out markets and wait for the fragility, wait for the left tail.”

4. Flexibility: your old letters could predict your AI take

• The observation: reading an investor’s letters from 2016 to 2023, he says he could often guess their AI stance. He notes that many famous short sellers are short CoreWeave on arguments about depreciation, economics, and circularity — “Nvidia invests into CoreWeave and then CoreWeave buys the Nvidia GPUs” — while investors deep in the tech weeds often argued that AI was the future, cycles were over, and there would be “AGI within 4 years.”
• He finds the lack of cross-pollination strange: shouldn’t macro investors from 2016-20 have recognized revolutionary AI in 2023, or tech-forward investors in Google, Facebook, and similar companies have identified AI as overhyped? He says he does not know of examples.
• Projecting onto himself: “reasonably early” to AI tools such as Claude and Cowork, yet “you didn’t short any SaaS companies,” didn’t buy semis, and — despite a history in bankrupt power companies — “didn’t make any money buying power trades.” He wonders whether he was too wedded to his earlier view that all power was a commodity to recognize power becoming a bottleneck.

5. MicroStrategy is effectively the preferred-equity market — and a CFO calling his shot

• Disclosure first: he has “a little bit of the long Bitcoin short MicroStrategy trade on.” The scale stats: MSTR was 8% of equity issuance in 2025, 10% so far in 2026, and 60% of preferred-equity issuance this year versus what he thinks was 33% last year — “MicroStrategy is the preferred equity market right now.”
• The failure mode: the “perpetual motion machine” (issue stock above NAV, buy Bitcoin, NAV rises, repeat) inverts if Bitcoin goes from 75,000 to 50,000 against $10B+ of 10% preferreds — “that cycle does not look good on the other side” — and if MSTR remains 60% of preferred issuance for another 18-24 months, a negative event could “blow up the preferred equity class.”
• His homework assignment: read MSTR’s Q1 call and slide deck — a sophisticated reader will first think “this is a really interesting way of thinking about it,” then realize they are issuing preferred equity at 10% to buy Bitcoin while assuming Bitcoin rises 20% per year: “It’s free. It’s just so crazy what they’re doing.”
• Via friend Artem Fokin, unnamed companies: a ~$100M company hired a CFO from a ~$500M company into the same role — “super unique,” since moving from larger to smaller would make more sense if the person were a chief accounting officer becoming CFO, i.e. receiving a promotion. The move would probably mean less pay and could read like “really calling your shot” on the smaller company’s equity; the counters — being overmatched, not “long for” the bigger company, or fleeing “a very rocky ride” — remain possible.
• Walker holds no position in either company. He says the signal could reflect the management team, the company joined, or the company left — probably some combination — and is increasingly focused on such unusual examples.