Summary

Walker’s answer to everyone asking about beaten-down SaaS: “I don’t think they’re that cheap,” and stock compensation is a real problem. A $1B company paying $100M/year in stock comp runs 10% dilution — doable while growing — but after a 90% drawdown “you can’t dilute your shareholders 50% per year,” while employees whose grants are 90% underwater ask why they should take a pay cut. The software index is down ~20% YTD, but that only looks cheap “on a chart.”
The terminal value of software is zero — “actually probably negative” — because when the business dies there are no hard assets, just leases to break. What ingrained this: customers of a company he was invested in called AI “an accelerant” for its product in October, then after the investment went against him said “AI has gotten so much better, we don’t even use that company anymore.” He guarantees some SaaS names down 70% will be five-baggers, and thinks AI will zero out a lot of others, as it did to Chegg — “a compounder business” — “in about two years.”
The borrowed framing that captures his view: SaaS is “a seven-foot bar.” Unless a generalist value investor has a specific, well-supported reason why AI won’t displace a given company, he passes — and “1x revenue vs comps at 3x” is not that edge, because relative valuation is not compelling when terminal value is zero.
His proposed AI hedge: a small slice of the portfolio in LEAPS on Microsoft, Meta, Amazon, and Google — insurance against AI capturing all the alpha and killing fundamental investing. The tell is Meta’s newly granted five-year options: stock at ~600, lowest strike ~1,200 and highest 3,000, implying that Walker sees a path to the stock reaching 4,000 or 5,000 — a $10 trillion-plus market cap. With implied vol in the high 30s against potential “winner take all or five winners take most,” the options “actually kind of feel too low to me.”
Pattern recognition is why investors allegedly improve with age — and also his “least favorite investor” failure mode. One 25–30-year veteran compared both a homebuilder and a SaaS name to the same anonymized “Procter & Gamble back in 2001” win: “he’s got a hammer and everything is a nail to him.” The unresolved question: “where’s the line between laziness and experience?”
His contrarian case for why AI might not kill fundamental investors: markets are meta-games that stuff patterns down the pattern-recognizer’s throat. He thinks 2010s companies spun off “the crappiest of the businesses” to indiscriminate spin-off buyers, while the market pumped 2025 DATs until it “kill[ed] the whole DAT game.” Companies could game AI factor-buying — e.g., distressed names running dividend-growth schemes and issuing equity into the bid. The other escape hatch: alpha that lives outside filings — conference checks and shaking the CEO’s hand.
The therapy-session close: he flagged Avis (CAR) at $95 on March 16 — aggressive Pentwater buying, 60% of stock locked up, 50% of free float short, “potential for a short squeeze” — and passed. On April 13 the stock is $371, a four-bagger in under a month, and since he says he would have played a squeeze with options, “probably would have been a 20-bagger.” His takeaway: “give yourself the grace… you can’t swing at everything” — but trying to turn himself into a robot is when he’s performed worst.

Deep dive

1. SaaS is a seven-foot bar — the fire he won’t run into

Walker’s self-described instinct is “running into the fire” when panic hits, and with the software index down ~20% YTD everyone wants to talk SaaS — but this time the instinct isn’t firing: “I don’t think they’re that cheap,” only cheap in the sense that “they were trading for 100 last year and they’re trading for 40 now.”
The stock-comp mechanism, spelled out: a $1B company paying $100M/year in stock comp is 10% annualized dilution — tolerable if growing. Let the stock fall 90% and it’s a $100M company paying $100M in stock comp: “you can’t dilute your shareholders 50% per year.” Employees are 90% underwater on old grants and ask why they should be underwater on prior grants or take a pay cut going forward. “It’s untenable on so many different levels.”
Terminal value is the deeper problem: because software is “all code and contracts,” it’s zero — “actually probably negative because you have to break the leases, you have to fire everyone” — unlike an office building you can convert. The anecdote “really ingrained in my brain”: in October, customers of a company he was invested in called AI “an accelerant” for its product; after the investment went against him, the same customers said “AI has gotten so much better, we don’t even use that company anymore.”
A listener comment on a payments episode crystallized it: “this is a seven-foot bar. Why would I try to step over a seven-foot bar?” He guarantees some software stocks down 70% will be five-baggers over five years — and thinks AI will zero out a lot of others, as it did to Chegg, the “compounder business” AI “zeroed out in about two years.” The edge he wants is an “X, Y, and Z reason why this specific software company will not be displaced by AI” — not “comps trade at 3x revenue, this one trades at 1x.” He’ll keep doing SaaS work anyway, to be ready when events fall into his lap 18 months out.

2. The AI hedge: LEAPS on Microsoft, Meta, Amazon, Google

The fear motivating the trade: AI is “the first time ever where… this is a tool that displaces people’s brains” — lawyers call it “the ultimate junior,” and juniors may eventually be followed by seniors. He worries that in “three, five years… maybe it’s a month away,” AI captures all the alpha, while his own portfolio has little AI exposure. So: should everyone hold a small position in long-dated calls on Microsoft, Meta, Amazon, and Google?
The Meta option grant is the anchor. With the stock at ~600, the lowest strike on newly granted five-year options is ~1,200 — a double in five years, “a very nice return for the stock,” leaves the options “basically worth zero” — and the highest strikes are 3,000, which Walker reads as a path to the stock reaching 4,000 or 5,000 — a $10 trillion-plus market cap. Farthest Meta LEAPS run only to December 2028, short of the January 2031 option expiry, but he thinks the acceleration could start being reflected in the next 18 months.
Name-by-name logic: Microsoft has OpenAI, Azure, and a stock dinged by the SaaS selloff — “you almost get like two ways to win”; Meta aligns with the grant and just rolled out its AI tool; Amazon has AWS and its talked-up custom-chip NVIDIA competitor; Google has GCP plus AI exposure — “I believe they have an investment in Anthropic. I can’t remember that for sure.”
On pricing: Meta, Microsoft, and Amazon LEAPS have volatility in the high 30s, he believes, which is “not low,” but against “a chance you just tip into a winner take all or, you know, five winners take most” — with Meta’s option grant underwriting at least a double — “it actually kind of feels too low to me, to be honest with you.” Repeatedly hedged as a thought experiment, not advice.

3. Pattern recognition: the line between experience and laziness

Why investors allegedly improve with age: accumulated patterns. Good earnings but the stock rises less than expected — the young see a buying opportunity, the old reread the 10-Q; big synergy targets on a merger — the young call it “the merger for the ages,” the old dig in; a puked-up spin-off becomes a buy signal after you’ve seen it two or three times.
The counter, from his own conversations: his “least favorite investors” are older ones using patterns as a crutch. One 25–30-year veteran compared a homebuilder to his (anonymized) “Procter & Gamble back in 2001” win — plausible enough — then compared a SaaS company to the exact same trade: “this guy is… he’s got a hammer and everything is a nail to him. He’s living in the past.”
He concedes the ambiguity — “maybe I’m being too dismissive of their pattern” — but the open question stands: “where’s the line between laziness and experience?”

4. Will the market stuff patterns down AI’s throat?

The bridge to AI: an LLM predicts by using past patterns — exactly the skill above — so it could replicate investors. But markets adapt. He thinks that in the 2010s, companies noticed event-driven investors “indiscriminately buying spin-offs” and “started spinning off the crappiest of the businesses… giving their investors a chance to dump their bags”; in the late ’60s, issuing stock to buy companies and create earnings growth got bid up; in 2025, DATs traded at huge premiums (MicroStrategy once ~2.5x NAV), and the market started pumping DATs until it “kill[ed] the whole DAT game.”
His scenario: as AI takes over more investing, companies game its factors — even distressed names running “dividend games,” issuing equity into a dividend-growth bid, “trading way above fundamental value because AIs have pattern recognition with dividend growth.” Self-aware caveat: “Maybe I’m telling myself a story… maybe me applying my lazy investor pattern recognition logic to AI is just silly.”
The second way AI misses: alpha outside filings — conference checks, and the classic judgment call where AI assesses a company as “50/50” on fraud but the investor who met the CEO and his wife “living in a shack” concludes 90/10 it’s not.
His own practice, no skeptic: increasing AI use for summarizing tertiary names — competitors, non-critical suppliers — though “in my core names, I’m still not really doing that.” Once comfortable, “I feel like you get the Limitless drug” from reviewing and ingesting much more information — with the acknowledged risk of getting too comfortable.

5. The Avis miss — therapy for a four-bagger that got away

The setup he published March 16: Avis (CAR) at ~$95, “really aggressive buying from Pentwater, who I really respect,” 60% of the stock locked up between Pentwater and the largest shareholder, 50% of free float sold short — “the potential for a short squeeze here.” He passed, having “gotten burned in rental cars before.”
Checked live on April 13: $371 — a four-bagger in under a month. And since he says he would have played a short squeeze with options, “it probably wouldn’t have been a four-bagger, probably would have been a 20-bagger.”
The closing psychology: “you need to give yourself the grace… you can’t swing at everything” — yet trying to turn himself into a robot fails too: “when I’ve tried to turn myself into a robot, it’s actually probably when I’ve performed the worst.” The fine line is learning from the miss without “letting it tear you apart” — and he admits, “I have no answers.”

Why Claude Feels Different (And What That Means for AI) | The a16z Show SpaceX's Financials Leaked: Is it Worth $2TN | Meta Debuts Muse Spark: Are They Back in the AI Race?