SKYWAY SUMMIT
Investments · Topic Brief
Castle, Chapel, Stone
Michael Burry's SW50 Part 2 covers eight software names across two categories, but its real product is a framework: a five-tier quality classification (Fortress, Castle, Chapel, Stone, Wood), a strict intrinsic-value rule (P/IV15 with a 1.5-2.25x buy zone), and a stock-based-compensation accounting method (Tragic Algebra) that surfaces who is creating value for shareholders versus who is renting earnings out to employees. The framework is the durable insight; the eight names are how it gets applied.
Eight names, two categories, one full position
Of the eight names covered, Burry discloses owning one (Adobe), watching several, and is on record not owning the other seven. The matrix:
| Name | Tier | P/IV15 | Rank | Score | Burry's action |
|---|---|---|---|---|---|
| Productivity Tools | |||||
| ADBE Adobe | Chapel | 0.91x | 2 / 50 | 64.7 | FULL POSITION |
| INTU Intuit | Castle | 2.81x | 24 / 50 | 38.2 | WATCHING · entry ~$171 |
| ADSK Autodesk | Chapel | 3.58x | 22 / 50 | 38.8 | WAIT FOR DOWN CYCLE |
| U Unity Software | Castle | 2.13x | 35 / 50 | 14.9 | WATCHING · entry ~$13 |
| DOCU DocuSign | Stone | 4.26x | 47 / 50 | 6.3 | DO NOT OWN |
| Cybersecurity | |||||
| ZS Zscaler | Chapel | 5.85x | 38 / 50 | 13.7 | WATCH FROM AFAR |
| PANW Palo Alto Networks | Castle | 7.86x | 40 / 50 | 12.0 | KEPT WELL AWAY |
| CRWD CrowdStrike | Castle | 23.18x | 50 / 50 | 2.4 | DEAD LAST IN SW50 |
P/IV15 = price as multiple of Burry's intrinsic-value-at-15%-return per share. Rank = position in his SW50 composite score, 1 best of 50, 50 worst. Burry's working rule: intrinsic value sits at roughly 1.5-2.25x IV15, with 2x as the buyback ceiling.
Burry opens the piece with a trading update outside the SW50 topic: he rolled his SOXX and QQQ puts at a debit, increasing capital at risk in both, with the explicit framing "I do believe we are witnessing a blowoff top." This is not the article's argument, but it is a high-information disclosure from a source the Belief Ledger already tracks. Worth its own entry, not a footnote.
Five tiers: quality is competitive posture, not financial health
Burry's tier system is not a Morningstar-style quality score. It is a posture classification: where does this company sit, in this AI-era moment, relative to the threats actually pressing on its business? The five tiers run from impregnable to disintegrating:
The eight names covered occupy the middle three tiers: Castle (INTU, U, PANW, CRWD), Chapel (ADBE, ADSK, ZS), Stone (DOCU). Fortress and Wood are mapped in the SW50 universe but no name in this article inhabits either extreme.
On the attack. Walled-garden AI strategy with proprietary models kept in-house. Liability lock-in (regulatory, fiduciary). Customers buy confidence and accountability, not code.
INTU U PANW CRWD
Defensible incumbent. Genuine AI integration but real competition at the edges. Open-partnership AI strategy introduces strategic exposure even where the core moat holds.
ADBE ADSK ZS
Acute AI-era exposure. Core workflow is exactly the structured-input-to-structured-output that LLMs and Microsoft Copilot replicate at marginal cost approaching zero.
DOCU
Tier is posture, not size
PANW and CRWD are Castle despite being among the worst-ranked names in the SW50 by composite score (40/50 and 50/50). The tier reflects platformization moats and proprietary information advantages (PANW: 30 billion attacks blocked daily, 15 petabytes of endpoint data; CRWD: trillions of daily Falcon instances across ~30,000 customers), not valuation. Castle = the company is well-positioned to defend its turf. It does not mean the stock is buyable. The valuation answer is in P/IV15, separately.
The Stone classification for DocuSign is similarly posture-driven, not size-driven. DOCU has 67% market share in eSignature and 1,200+ clients spending $300k+ annually. None of that matters in Burry's read because the workflow itself (convert authenticated structured input to authenticated structured output, then store) is now natively replicable by Microsoft 365 + Purview + Copilot ("the Borg"), Adobe Acrobat Sign as a free Document Cloud attachment, and CLM-native vendors IronClad and SpotDraft pulling value-capture upstream. Market share is a lagging indicator of a moat already breached.
P/IV15: intrinsic value at 15% return, with a hard buy rule
The valuation metric is Burry's own and the rule is strict: intrinsic value per share sits in the 1.5-2.25x IV15 range, with 2x as the working buyback ceiling. Buybacks above that destroy intrinsic value per share. Below 1.5x is the Fat Pitch zone. The eight names spread across the entire scale, with CRWD off the visible end:
Tragic Algebra: what's left after the SBC charge
The accounting framework underneath the verdicts is Burry's Tragic Algebra: a multi-year pooled calculation of cumulative GAAP net income minus cumulative true stock-based-compensation cost (his omega, calculated at ~2.5x reported GAAP SBC historically), expressed as a retention ratio (ΔE). 100% means the company kept every reported dollar of earnings for shareholders. Below that, SBC is renting earnings out to employees. Negative or not-calculable means SBC dilution swamps income entirely.
For four of the eight, Burry's retention ratio cannot be computed because earnings are negative. The absolute owners'-earnings number tells the story instead:
- Unity (U) · SBC 15% rev, falling from 30% historicalNegative across the six-year pool
- Zscaler (ZS) · SBC 25% rev, flatOwners' earnings ~-$200M on $3.33B
- Palo Alto Networks (PANW) · SBC 13.1% rev, fallingOwners' earnings ~-$700M on $11.3B
- CrowdStrike (CRWD) · SBC 22.2% rev, no buybacksOwners' earnings ~-$1.4B on $4.93B
CRWD's shares-outstanding count has risen ~25% over the last five years (a 4.5% annual dilution rate), which on Burry's discounted-cash-flow math "simply devastates intrinsic value per share." PANW's SBC is the cleanest of the cyber names in percentage terms but the largest in absolute dollars; the company's "adjusted FCF" of ~37% margin disappears entirely once GAAP SBC is recognized.
Ballard's Test: the philosophical anchor under the framework
"An entity does not possess capacity for understanding until reason is demonstrated in the absence of language."
Ballard's Test, reintroduced by Burry as the test for whether AI threatens a business model
Burry's read of where LLMs land today: they are not reason engines, they are language-output engines. The implication for software: businesses whose value rests on converting human language to structured output (DocuSign signature, basic accounting categorization, simple contract drafting) are exposed because that is literally what an LLM does natively. Businesses whose value rests on expert judgment with liability backstops (Intuit's accountants signing returns, Autodesk's licensed engineer of record) survive because the AI cannot assume the liability and the customer is buying the liability shield, not the software. This is the rule Burry uses to sort the entire universe.
The cleanest fit. Adobe at $247.60 trades at 0.91x its $262 IV15, the only one of the eight in the Fat Pitch zone. Owners' earnings ~$7.7B on revenue of $26.7B. Sales multiple at 14-year lows: the market is pricing Adobe at pre-subscription-era multiples, fully unwinding the subscription premium it earned over two decades.
The bear case is that Adobe is exactly the kind of "structured input to structured output" tool a frontier autonomous model replaces. Burry's response is direct: human creative input is not structured, and LLMs will never beat the creative instinct of a talented human being. The defensive layer is concrete: 850M monthly active users growing 17% YoY, 99 of the Fortune 100 as customers, "chain-of-IP custody" over the licensed brand libraries any AI agent would need before doing creative work, and 85% of 2026 Sundance films still made with Creative Cloud. The Figma deal was blocked in 2023 but Burry now reads the miss as lucky: Figma is the part of Adobe's stack most exposed to agentic AI, and the market has already punished Adobe for losing it.
Capital return is the kicker. $25B buyback authorization against a $108B market cap, ~23% of the float, at a price below IV15. The CEO transition is messy, near-term overhang, but Burry's view is "much of this is already in the share price." He owns a full position.
Intuit is the highest-quality name among the productivity tools but Burry does not own it. The reason is discipline: at 2.81x his $114 IV15, the stock is well above the buyback rule and not yet in his strike zone. Burry would find Intuit "interesting in the area of 1.5x IV15," which implies an entry near $171, ~46% below current.
What earns Castle: Intuit is on the attack, not defending. TurboTax Live up 38% with 36% net-new through the 600-store local channel. AI agents already serving 3M customers with 85% repeat engagement. In January 2026 alone the AI accounting agent categorized 237M transactions; the AI business-tax agent finds an average $1,000 in incremental deductions per customer; the AI stock-basis agent lowers taxable income by an average of $12,000 versus filings done without it. Intuit Accountant Suite has signed five top-20 accounting firms. CEO Goodarzi's framing is sharp: customers buy confidence, not code, which is why they spend at least 7x more on accounting and tax experts than on software alone.
The walled-garden AI strategy is what most distinguishes Intuit. Intuit's Financial Large Language Model is proprietary; OpenAI, Anthropic, and Microsoft are partner channels, not architectural dependencies. The liability lock-in is the moat no outside AI agent can assume. Burry's only complaint is capital allocation: Intuit has been buying back stock above intrinsic value, which "hurts intrinsic value per share." A new $8B buyback authorization is "not below intrinsic value in the low $300s."
Autodesk has the best Tragic Algebra score of the four productivity names with positive earnings at 103.5%, second-best in the entire SW50 behind only Veeva. Eleven years of disciplined buybacks more than offset SBC ($8.3B in buybacks vs $5.1B in SBC, against $4.3B cumulative GAAP net income). The buyback mechanic is Buffett-style: a programmatic grid that deploys more capital at lower prices. Q4 2026 alone deployed $333M at an average $303, the lowest quarterly buyback price of the fiscal year. SBC is committed to fall below 10% of revenue in FY27.
The real differentiator is the professional-liability backstop. Civil 3D controls how roads are built; Revit controls how buildings are built. The licensed engineer signs the drawing, and the engineer is personally liable for civil and criminal penalties if the bridge falls down. Mythos-class frontier agents can do design work faster, but the licensed engineer is locked in until insurance starts accepting agent output. As long as the engineer is locked in, the engineer uses Autodesk. Add a decade of proprietary 2D/3D multimodal model development for AI-native design.
Burry's verdict: "I do not own Autodesk. I am a long-time admirer of the company, and I have owned it in the distant past." Reasoning: high SBC, "the valuation is not only rich but not close. I hope to revisit this in the next down cycle. With data centers and industrial backlog, this is not that." The data-center build-out tailwind is real (more construction = more Autodesk seats, and hyperscalers deploy more seats per project than typical builders) but the cyclical pullback Burry wants is not on offer yet.
Unity gets a Castle classification despite an "atrocious shareholder history" and a six-year Tragic Algebra picture that Burry calls "ugly" (cumulative GAAP net loss -$3.63B, cumulative GAAP SBC $2.66B, with true SBC materially higher because of the dramatic stock decline from $130 in 2021 to $15 last summer). ΔE is not calculable. So why Castle?
The Castle classification reflects an inflection. CEO Matt Bromberg (brought in from EA two years ago) has rebuilt the strategy around three integrated pieces: the Unity game engine (~70% market share in mobile game creation, leader in automotive interactive software, half of all video games use Unity, customers include Disney, BMW, US Air Force); Vector, the in-house ML ad-targeting product launched mid-2025 that displaced and shut down the $4.4B IronSource acquisition this April; and Unity AI, a 3D world model trained on 20 years of proprietary game-making data. Q1 2026 revenue $508M (+17%), gross margin in the high-80s, SBC falling to 15% of revenue from 23% a year ago, with GAAP profitability projected later this year. Strategic revenue grew 35%, Vector +80%, Vector is now 80% of strategic Grow revenue, adjusted EBITDA margin hit 27% (best in two years).
Burry's verdict: "I do not own Unity. I am comfortable at the Castle level for Unity, which does not help the atrocious shareholder history much. Unity is inflecting though, and has a plausible, aggressive path forward, though still with uncertainty and a poor track record on profit. At 2.13x P/IV15 and a mid-30s rank within SW50. If it gets back down to 1x P/IV15, I will be waiting." Entry at 1x IV15 implies ~$13, roughly 52% below current.
DocuSign is the worst-rated productivity name on every measure that matters in Burry's framework. Down 85% from the 2021 peak but still 4.26x its $11.37 IV15. Owners' earnings $195M on $3.4B revenue. The Tragic Algebra is "charitably the most despicable in the productivity category": $2.39B in SBC since 2023 swamps $1.35B in cumulative GAAP net income, producing a retention ratio of negative $9.50 per GAAP dollar. The author's letter grade for shareholder treatment: F.
The Dollar Net Retention picture summarizes the operational story: peak 125% in Q1 FY22, trough 98% in Q4 FY24, latest 102% in Q4 FY26, still well below the 112-117% pre-COVID baseline. 102% implies a shrinking business once mix is normalized. eSignature has stalled and 67% market share is now a lagging indicator: competitive entry from Microsoft Word + 365 Copilot + Purview (the inevitable Borg), Adobe Acrobat Sign distributed free through Document Cloud, and CLM-native incumbents IronClad and SpotDraft moving the value capture upstack to contract drafting where signature becomes a free feature.
The defensive bet is Intelligent Agreement Management: $350M ARR after 18 months, ~10% of revenue, and ahead of plan on renewals. The thesis is that 200M agreements stored in Navigator becomes a moat no agent can replicate. Burry's response: not settled science whether new AI-native agents need that historical context to perform. The $2.6B buyback authorized against a $9.6B market cap is value-destructive at 4x IV15.
Cybersecurity gets a different reframe than productivity tools, and it inverts the obvious read. Customers are scrambling, revenues are strong, stocks have been on a tear. But the entire architecture the incumbents are priced for, what Burry calls One Lake One Life (OLOL), is itself the vulnerability:
OLOL (One Lake One Life). The flywheel that powered the last decade of cloud-native security: aggregate all customer data into one cloud, train one set of detection models on the lake, deliver protection to all customers simultaneously. More customers, more data, more customers, more data. This is what every public security platform Burry covers actually runs: ZS, PANW, CRWD all OLOL.
The problem. "In a Mythos & Friends world, aggregated data becomes the fatted calf, high value and relatively weak." If the one lake is breached, all the data is revealed. Frontier AI models are finding security vulnerabilities at machine speed, multiplying known-but-unpatched vulnerabilities by as much as 10x.
SDDD (Segregated and Deception-Defended Data). Burry's preferred new paradigm: more sequestered or isolated data lakes paired to deception, diversion, and misdirection strategies. "The first legacy company to wholesale go SDDD will secure a lead position against frontier AI model threats and also shuffle the industry deck." None of the three covered has done that yet. Acquisitions and platform tweaks are pointing in that direction, slowly.
Three structural points compound: (1) none of the three cybersecurity firms make money on Burry's owners-earnings basis; their reported FCF is fully gobbled by GAAP SBC. (2) Cybersecurity stocks rallied to all-time highs after Anthropic's Glasswing project announcement on May 22 (the project gives fifty critical-security-infrastructure organizations early Mythos access), but Burry reads Glasswing as "letting the wolf in the hen house" since it presupposes Mythos remaining unavailable to attackers. Wholesale devastation would have awaited if Anthropic had released Mythos into the wild. (3) The Innovator's Dilemma is real here: legacy firms addicted to OLOL cash flows and margins may be slow to acquire or iterate toward SDDD.
Zscaler runs the cloud-based Zero Trust Exchange across ~160 data centers and ~9,400 enterprise customers, mediating 500 billion transactions per day. Cloud-based and OLOL by construction, but Burry credits it for being focused on data traffic and connections rather than storing precious customer data: "ameliorates the fatted-calf aspect to an extent." Q3 revenue $850M (high-teens organically), AI Protect bookings crossed $100M trailing-12-months as a new product line, and a historic deal closed with a giant healthcare system across 2,000 sites displacing a major firewall incumbent and a legacy VPN. The Symmetry Systems acquisition adds identity-to-data mapping; CEO Chaudhry mentioned Mythos six times on the earnings call. CEO Chaudhry also pointed at the future: "Today, users are the weakest link in cybersecurity. But soon, AI agents will be the weakest link... a single compromised agent can move from discovery to data theft in minutes."
What keeps Zscaler out of Castle: less scale than Cloudflare and PANW, plus a still-OLOL data architecture. Burry's tease: "If Zscaler is moving away from its OLOL with a more segregated and deception-defended data (SDDD) infrastructure, then it could climb to Castle even Fortress." But the ten-year Tragic Algebra "is U-G-L-Y," with SBC at 25% of revenue (vs PANW 13%) and no buyback discipline; ΔE is not calculable. Burry: "I do not own Zscaler. At 5.85x P/IV15 after the post-earnings drop, my interest in Chaudhry's implementation with the Mythos and Symmetry interaction is closer to actionable than it was a week ago at $185, but still academic at this price. I am not shorting the stock here, but will watch from afar. I see a decent company with decent competitive advantage but indecent SBC and IV15 multiple."
PANW just hit a new all-time high. $205B market cap on $11.3B revenue, trading at 7.86x its $31.62 IV15. The company offers the most complete cybersecurity platform at scale: 65,000+ firewall customers, 1,550 Prisma platform customers (up 35% YoY, 110 net-new last quarter outside seasonally-strong Q4), 600+ XSIAM customers, three big-ticket acquisitions this year (CyberArk for $25B identity, Chronosphere for observability, Koi for Agentic Endpoint Security), and a pending Portkey acquisition for AI-gateway security. The Castle classification reflects platformization moat plus information advantage: 30 billion attacks blocked daily across 15 petabytes of transaction/endpoint data. Customers who have platformed expand PANW spend ~20% with low-single-digit churn. The flywheel works at scale, "Wall Street loves this dearly, too much even."
What keeps Burry off: the M&A treadmill is steadily eroding ROIC (organic growth fell from 29% in 2022 to 14% this fiscal year), and the dilution from acquisitions plus stock-based compensation is brutal. Owners' earnings ~-$700M on $11.3B revenue. SBC at 13.1% of revenue is best of the three cyber names and falling, but with $25B paid for CyberArk on top of CyberArk's own legacy SBC, the dilution math gets worse before better. PANW is still squarely OLOL/SASE in architecture; XSIAM AgentiX (Agentic remediation that runs inside customer IT environments) is a shift toward SDDD but Burry sees it as a first step, not the model change. Burry: "I do not own Palo Alto Networks. The premium valuation and brutal dilution both from acquisitions and stock-based compensation keep me well away... priced with maximum market confidence even as the security architecture becomes structurally more exposed than ever. Maybe at the sound of galloping horses, I'll turn back to see what is left."
CRWD is the maximum-extreme case in the entire piece. Stock at $649, all-time high $671.55 set May 26, market cap $163B, on $4.93B revenue. P/IV15 of 23.18x against the $27.84 IV15, by a wide margin the most expensive name. Dead last (50/50) in the SW50 composite score. Burry: "Not last among cybersecurity names. Last, period."
The platform story is genuinely strong on the surface. Unlike PANW, CRWD was built as a platform from day one: no legacy product, no acquisitions, just the cloud-based Falcon platform with one agent, one cloud, the Threat Graph. The company rode KISS (Keep It Simple, Stupid) to trillions of daily instances and ~30,000 customers. The Falcon platform has 29 modules across endpoint, identity, cloud, security operations, data, and AI/agentic security. New module activation is instantaneous and frictionless via Falcon Flex, the consumption-based pricing model. Q4 revenue $1.166B (+22% organic), AI-native module is the fastest-growing piece.
The framework problem: this is OLOL with "unparalleled elegance." Management presents it as a moat against all new threats; Burry sees it as a moat only against other OLOL competitors. CRWD has the cleanest cloud-native architecture and therefore the most concentrated fatted-calf risk. Add the financial picture: owners' earnings of approximately -$1.4B on $4.93B revenue, GAAP SBC of $1.097B (22.2% of revenue), and no buybacks at all. Shares outstanding up ~25% over five years, a 4.5% annual dilution rate, which on a discounted-cash-flow basis "simply devastates intrinsic value per share." Burry does not state his position explicitly here, but with rank 50/50 in his universe and a 23x P/IV15, the inference is clear: not owned, not buyable, the canonical example of what the framework flags as the maximum mismatch between market valuation and intrinsic value.
A complementary lens, not a primary screen
Software does not sit in Skyway's confirmed edge zone, which is structural scarcity plus Western policy catalyst. Burry's universe is fundamental value in the AI-transition era, a different selection function. The lens is useful as a complement: a way to read what the source diet keeps surfacing about AI-resistance hierarchies, and a discipline check for any software name that creeps into the watch list on a Burry endorsement alone.
The cybersecurity half maps more interestingly to Skyway themes than productivity tools do. The OLOL-to-SDDD paradigm shift is itself a structural-scarcity-plus-Western-policy story in miniature: a structural vulnerability (centralized data aggregation), a clear policy/regulatory direction (sovereign data, deception architectures), and a winner-take-most opportunity for whichever incumbent or startup wholesales-pivots first. None of the three covered names occupies that position today, which is exactly Burry's framework verdict.
What is genuinely transferable to Skyway is the method, not the picks. Three specific framework imports worth proposing for adoption (full belief-ledger paste blocks are filed separately):
1. Ballard's-Test classification for AI exposure. Sort any name in the universe by whether its core workflow is language-output (LLM-exposed) or expert-judgment-with-liability (AI-resistant). This applies cleanly to names Skyway already covers and produces testable predictions about which workflows survive.
2. Tragic Algebra screen for SBC-heavy names. Where current Skyway coverage relies on FCF, Burry's ΔE measure exposes companies that pad reported FCF by adding back SBC. CRWV in particular bears this scrutiny ($25B debt plus high SBC norms in AI infrastructure).
3. P/IV15 valuation rule, adapted. Skyway's reverse-DCF engine (07_Analytics/scripts/reverse_dcf.py) already solves for implied growth at a stated WACC. Adding an IV15 calculation (intrinsic value at 15% return assumption) and a 1.5-2.25x rule is a natural extension of Component A.
4. OLOL/SDDD architecture lens for any data-aggregation business. Centralized data lakes were the moat. In a Mythos-class threat environment, they are the vulnerability. This generalizes beyond cybersecurity to any business model premised on data-aggregation flywheels.
AP-4 (no own thesis) is the first risk. "Burry owns ADBE" is not a thesis. ADBE does not fit any confirmed Skyway edge zone, the catalyst (subscription premium unwinding, CEO transition resolving) is not a Skyway-pattern catalyst, and the vehicle question is unanswered. Buying ADBE because Burry buys ADBE would trigger the rule.
AP-7 (right thesis, wrong vehicle) is the second. If a software name does enter the watchlist on framework grounds, none of the eight belong on Skyway's 5x stock CFD. ADBE has 52-week range volatility; DOCU is too small-cap; INTU and U at a deep pullback target require patient cash equity; the three cyber names trade at extreme volatility and platform-risk multiples. Cash equity only for any software adoption.
Risks to Burry's read
Three lines of attack a sceptic could mount:
Castle moats may not survive if liability frameworks adapt. Burry's whole Castle thesis rests on AI being unable to assume liability. If insurance regimes evolve to accept agent output (he names this explicitly for Autodesk's engineering license) or if regulatory frameworks adapt to allow LLM-signed work, the Castle becomes a Chapel becomes a Stone. The pathway is slow but the direction is one-way.
Tragic Algebra punishes optionality. SBC-heavy companies are often growth companies that retain talent through equity. The ΔE metric correctly identifies dilution, but for a name like CRWV (which Skyway already covers) it would flag heavily despite the AI-infrastructure thesis being structurally early. Burry uses this lens for mature software; importing it indiscriminately to early-stage names could over-penalize the right businesses.
P/IV15 hard rule is sensitive to the 15% return assumption. The 1.5-2.25x band assumes a 15% required return. A regime where the cost of capital sustains lower (8-10% returns acceptable) compresses the intrinsic-value band upward, making Burry's "expensive" stocks look fairer. This is a real risk for the framework, not a defect, but should be made explicit before adoption.
OLOL may persist as the dominant architecture. Burry assumes Mythos-class AI threats force SDDD adoption. If Glasswing-style defensive partnerships hold up and the cyber incumbents harden their lakes faster than attackers can target them, OLOL extends its run and the legacy cyber names compound for several more years.
The eight names are out of Skyway's confirmed edge zone, so this is not an entry-screen source. The value is the framework itself: the five-tier posture system (Fortress, Castle, Chapel, Stone, Wood), Ballard's-Test classification of AI exposure, Tragic Algebra accounting (ΔE with omega-true-SBC at ~2.5x reported GAAP), the P/IV15 valuation rule with a 1.5-2.25x band, and the OLOL-to-SDDD architecture lens for data-aggregation businesses. Burry owns one full position (Adobe), watches several at deeper-discount entry points, and is on record not owning the seven richer names. The macro disclosure (rolled SOXX+QQQ puts, "blowoff top") is the highest-information item in the piece and warrants its own ledger entry separate from this intake.
Source ageing. Article published 28 May 2026, four days old at intake. All prices and multiples cited are as of source date; verify before any action. Burry's disclosed positions (long ADBE, short SOXX/QQQ) are point-in-time and subject to change without notice. P/IV15 is Burry's own metric and not independently auditable from the source.
Cross-references. Complementary to Anatomy of the AI Bezzle (Burry, 22 May 2026) which argues the macro AI-spend cycle is unsustainable. The two pieces are consistent: this is the long-side of Burry's AI-era software view (one ADBE long, framework for sorting the rest); the AI Bezzle brief is the short-side of the AI-infrastructure spend.