모두가 잘못 읽고 있는 AI 지도

THE DAILY H.E.A.T. Torching Wall Street’s Obsolete Playbook. Delivered straight to your inbox 5 days a week. Subscribe Home Posts The AI Map Everyone's Reading Wrong The AI Map Everyone's Reading Wrong Silicon Valley just published the consensus blueprint for the AI economy. The fulcrums it points to are obvious. The ones it doesn't are where the money will actually be made. May 22, 2026 I’ve been a trader and investor for 44 years. I left Wall Street long ago—-once I understood that their obsolete advice is designed to profit them, not you. Today, my firm manages around $5 billion in ETFs, and I don’t answer to anybody. I tell the truth because trying to fool investors doesn’t help them, or me. In Daily H.E.A.T. , I show you how to Hedge against disaster, find your Edge, exploit Asymmetric opportunities, and ride major Themes before Wall Street catches on. Table of Contents H.E.A.T. News vs. Noise: What’s Moving Markets Today ETF News In Case You Missed It H.E.A.T. The Situation Three weeks ago, Chamath Palihapitiya published what may be the cleanest piece of investment cartography since the OSI stack diagrams of the early 1990s. A six-layer map of where value accrues in the AI economy — infrastructure, chips, data, models, execution, application — with a clean fork between physical and digital systems at the chip layer. The framework, dated May 1 and circulated again on X on May 15, is already being passed around every group chat from Sand Hill Road to Park Avenue. It will define how a generation of allocators thinks about this market. And that is precisely the problem. When a framework this clean becomes the consensus map, every investor reading it draws the same conclusions in the same places. They look at the chip layer and buy NVIDIA. They look at the model layer and buy the hyperscalers. They look at the application layer and chase whatever vertical SaaS name pitched them last quarter. They mistake the layers themselves for the fulcrums. They are missing where the real chokepoints sit. And the chokepoints are not in the layers. They are in the boundaries between them. The trade, in one line:  Don't buy the colored blocks. Buy the tollbooths between them. The Stack Itself Strip the consensus reading away and look at what the Social Capital diagram actually says. The chip layer doesn't print itself — it sits on top of a hardware supply chain that runs through one Dutch lithography company, four Japanese chemical suppliers, and a single mine in western North Carolina that produces the high-purity quartz no fab can operate without. Above the chips, the model layer is collapsing in price — a 1,500x cost decline in six years, by Chamath's own numbers. Intelligence is becoming free. The interesting question is no longer who builds the smartest model. It's who owns the layer between the model and the user, the layer Social Capital labels execution: orchestration, memory context, APIs, the agent runtime. That's where the fee tollbooth gets built. Below the chips, the framework forks. To one side, software AI compounds toward zero marginal cost. To the other, physical AI compounds toward physical constraint — batteries, actuators, sensor fusion, supply chains for materials that cannot be made into software no matter how clever the model. Every dollar in this stack still originates at the top, at the application layer, and flows down. But every dollar gets taxed at the fulcrums on the way. The question is not which layer to own. The question is which tollbooth. "Rockefeller had 90% of refining by 1880. Cisco had 85% of routing by 2000. The same pattern is forming now — and it isn't forming where the chart's color blocks tell you to look." Where the Fulcrums Actually Are There are six layers in the diagram. There are roughly four places in the diagram where the entire stack would collapse if a single supplier walked away. Those four places are not coextensive with the layers. That is the contrarian read. The first sits between infrastructure and chips: EUV lithography and the photoresists that feed it. ASML in Veldhoven is the only company on earth that builds extreme ultraviolet machines capable of printing sub-7nm geometries. Each machine is the size of a city bus and contains more than 100,000 components. Four Japanese chemical companies — JSR, Tokyo Ohka Kogyo, Shin-Etsu, and Fujifilm — supply the photoresist film without which the machines are inert. There is no second source for either. The U.S. CHIPS Act allocated $52 billion to domestic fabrication. None of it builds an EUV machine. The second sits inside the chip layer itself: the CUDA software moat. NVIDIA's silicon is reproducible — AMD makes competitive parts, the hyperscalers are designing their own. NVIDIA's developer ecosystem, twenty years of accumulated libraries, kernels, and tooling, is not. The diagram labels the chip layer as a single colored block. Inside that block, the real moat is software. The third sits between models and applications, in the layer Social Capital calls execution: the agent runtime. When intelligence is free, value migrates to whoever orchestrates intelligence into outcomes. Tool use, memory, API integration, the workflow scaffolding that turns a probabilistic model output into a deterministic business result. This is the layer where Palantir, ServiceNow, and a small set of vertical AI infrastructure names are quietly establishing the toll road. The fourth sits on the physical fork, where the diagram says actuation: batteries, motors, and the rare-earth supply chain that feeds both. The smartest physical AI model in the world is dead the moment its battery runs out. Humanoid robotics — the application the entire physical fork is being built toward — is a battery-density problem, a motor-torque problem, and a magnet-supply problem before it is a software problem. 1 EUV monopoly (ASML) 4 Photoresist suppliers (Japan) 1,500x Model cost decline / 6 years 6 Layers in the stack The Fork That Matters Wall Street is going to mis-price one side of this fork badly, and it isn't the side most people think. On the digital fork — agents, models, applications — every dollar of value can in principle be replicated at near-zero marginal cost. That's a feature for users and a problem for investors. The economics look like 1999-vintage commercial internet: explosive top-line growth, brutal compression as competition arrives, a handful of platform winners and a long tail of road-kill. Buy the platforms. Avoid the road-kill. Standard playbook. On the physical fork, the economics are inverted. Every robot consumes physical resources — batteries, magnets, rare earths, steel, copper, the power grid itself — that cannot be replicated by software. Demand for these inputs is convex in robot deployment. One million humanoid units in the field doesn't require one million batteries. It requires one million batteries plus replacements plus the energy storage to charge them plus the grid capacity to deliver that energy plus the upstream mining to produce the lithium and nickel and cobalt and neodymium. Software AI gets cheaper as it scales. Physical AI gets more expensive as it scales — and the cost flows directly to the companies that own the constrained physical inputs. That is the inversion the diagram doesn't show, and it's the most important investment fact of the next decade. Spotlight: The Four Hidden Chokepoints 1. EUV lithography (ASML) + photoresist (JSR, TOK, Shin-Etsu, Fujifilm). No second source. Every advanced chip in the world depends on a single Dutch supplier and four Japanese chemical companies. 2. CUDA software moat (NVIDIA). Silicon is replicable. Twenty years of developer tooling is not. The moat is inside the chip layer, not the chip layer itself. 3. Agent runtime / execution layer. When intelligence is free, the orchestration tollbooth captures the rent. Palantir, ServiceNow, and a thin tier of vertical names are positioning here. 4. Actuation supply chain. Batteries, rare-earth magnets, precision motors. The physical fork's pricing power compounds with every humanoid deployed. Winners — Three Tiers The framework names six layers. The investable opportunity is narrower. Three tiers below, ranked by the durability of pricing power — not by market cap, headline AUM exposure, or how loudly the name was pitched at the last conference. Tier Name / Ticker Why it sits on a fulcrum Tier 1 — Fulcrum monopolies ASML (ASML) NVIDIA (NVDA) TSMC (TSM) Single-supplier monopolies on the underlying enabling technology. ASML on EUV; NVIDIA on the developer ecosystem; TSMC on advanced-node manufacturing capacity. These are the names that survive when every layer above them gets re-priced. Tier 2 — Layer leaders Applied Materials (AMAT) Lam Research (LRCX) KLA Corp (KLAC) Palantir (PLTR) Deposition, etch, inspection — the equipment without which TSMC cannot turn a wafer. On the execution layer, Palantir is the most established orchestration platform with both commercial and defense distribution. Tier 3 — Physical-fork derivatives MP Materials (MP) Rockwell Automation (ROK) ABB Ltd (ABBNY) Fanuc (FANUY) Rare-earth processing capacity outside China (MP) plus the three Western names with the deepest installed base in industrial actuation. These compound on humanoid and robotics deployment regardless of which model wins. Pressure Points Not every name in this stack benefits. Some sit on the wrong side of the fulcrums. Three structural pressures below. Where the pressure sits Why Pure-play foundation-model labs without distribution Model prices fell 1,500x in six years. There is no obvious mechanism by which they stop falling. A model lab without a captive application moat or proprietary data is selling a commoditizing input. Workflow SaaS without proprietary data or distribution Seat-based application pricing is contested the moment a competent agent can do the same workflow at one-hundredth the cost. The names that survive have a proprietary distribution chokepoint, proprietary training data, or both. The rest are competing on UI polish against a deflationary intelligence layer. Legacy industrial automation without an AI integration path On the physical fork, the actuation layer compounds for whoever can attach to modern AI models. Industrial names that miss the integration window become hardware vendors competing on price with Chinese alternatives. Credibility Firewall Distinguishing sourced fact from editorial inference. As always. Confirmed Directional • Social Capital published the six-layer AI stack framework on May 1, 2026, and circulated it on X on May 15, 2026. • ASML is the sole supplier of EUV lithography systems globally. • JSR, Tokyo Ohka Kogyo, Shin-Etsu, and Fujifilm dominate the global photoresist market. • Spruce Pine, NC is the primary global source of the high-purity quartz used in semiconductor crucibles. • NVIDIA's CUDA platform is the dominant developer software environment for GPU compute. • Per Social Capital, the cost of running a frontier model has fallen approximately 1,500x in six years. • The designation of EUV, CUDA, agent runtimes, and actuation as the four primary fulcrums is editorial. • Tier rankings of named companies are this newsletter's assessment, not consensus rankings. • The claim that boundary layers compound pricing power faster than layer-center incumbents is a thesis, not a measured fact. • Forward share, margin, or deployment estimates for any named company should be treated as analytical, not predictive. • The physical-fork inversion (cost rises with scale) is a structural argument; actual outcomes will depend on battery chemistry, magnet supply, and grid build-out pace. Why Now — Catalyst Chamath's stack diagram is not the cause of anything. It is a marker — a signal that the institutional capital community has reached enough consensus on the shape of the AI economy to publish maps of it. That consensus formation is itself the catalyst. Maps create positioning. Positioning creates flows. Flows create the divergence between names that sit on the labeled layers and names that sit on the unlabeled fulcrums between them. This is the moment the consensus map gets drawn. Six months from now, the trade is to be early on the names that consensus missed. Twelve months from now, it's too late. The Bear Case The honest pushback on this thesis sits in three places. First, photoresist substitution. Self-aligned quadruple patterning and nano-imprint lithography are both maturing. If Canon's nano-imprint platform reaches commercial parity, ASML's monopoly compresses. Second, agent-runtime commoditization. The execution layer could itself be commoditized by open-source orchestration frameworks faster than the toll road gets built. LangChain, LlamaIndex, and a half-dozen others are well-funded. Third, humanoid robotics may simply not happen on the timeline anyone is modeling. If commercial humanoid deployments slip from 2027 to 2031, every name on the physical fork de-rates. Battery and magnet demand follows units, not narrative. None of these break the framework. All of them compress the returns. Position size accordingly. What Would Change the Story — Next 90 Days Three watch items convert the bear case into a tracking discipline. If any of them moves materially, the thesis re-prices and so does the positioning. 1. Nano-imprint lithography acceleration. Watch for Canon's NIL platform showing commercial-volume production at any leading-edge node, or any foundry announcement of pilot adoption. The signal is volume, not press release. 2. Agent-runtime commoditization. Watch enterprise software earnings for evidence that open-source orchestration is taking paid orchestration share in production deployments — not pilots. The names with exposure here will flag it in guidance before they flag it in revenue. 3. Humanoid deployment slippage. Watch unit deliveries from Tesla, Figure, Agility, and the Chinese leaders versus their stated 2026 and 2027 guidance. A six-month slip is noise. A multi-year slip re-prices the physical fork. Five Takeaways 1.  The fulcrums are not the layers. The Social Capital diagram colors in six layers. The pricing power lives at four boundary points between them — EUV/photoresist, CUDA, agent runtime, and actuation supply chain. Own boundaries, not blocks. 2.  Intelligence is becoming free; orchestration is not. A 1,500x decline in model cost is the most important data point in the framework. It moves the rent from the model layer to the execution layer. Position there before consensus does. 3.  The physical fork inverts standard tech economics. Software AI scales toward zero marginal cost. Physical AI scales toward physical constraint. Batteries, magnets, and rare earths get more valuable per unit of AI deployed, not less. 4.  NVIDIA's moat is software, not silicon. The chip layer in the diagram is one color block. Inside it, CUDA is the moat. Competing silicon arrives every quarter. The developer ecosystem does not get replicated on that cadence. 5.  Consensus maps create the inefficiency. The moment a clean framework circulates institutionally is the moment positioning flows toward the labeled layers and away from the unlabeled fulcrums. The trade is in the unlabeled space. Every era of computing is won by whoever sits on the chokepoint. The chokepoints in this era are now visible. The crowd is looking at the wrong parts of the map. The relative trade is the trade — concentration in the tollbooths, weight away from the commoditizing middle. Space Exposure Designed For Potential Weekly Payouts The space economy is no longer just a government program — it’s becoming part of a global industry that is gradually advancing. But why view long-term growth as the only potential path forward? SPCI seeks to offer approximately 100% concentrated exposure to the space industry while striving to generate weekly income through a disciplined put credit spread options strategy. Gain exposure to the “final frontier” and a potential weekly distribution while you wait for the theme to play out. Discover SPCI. Distributor: Foreside Fund Services | Investing involves risk including possible loss of principle. News vs. Noise: What’s Moving Markets Today $WMT ( ▼ 0.88% ) announced earnings yesterday and the market didn’t take to kindly to them….. What was interesting was the why. Management provided insight into the impact of higher energy prices, inflation, and the state of the consumer. We have a large fuel business and we see that in the most recent period, the number of gallons that customers fill up with when they come to our fuel stations fell below 10 for the first time since 2022, that's an indication of stress. And so certainly, as you look at quarter-over-quarter incremental pressure, that's one of the areas that I would call out. I'll point to just gallons consumed as an example. Like in our Sam's business....in the month of May, our gallons are up 12%. If you look across the industry, they're down 5%... that tells you that customers are coming to us looking for value. What's important to note about that is that a fuel member spends 1.6x more in the rest of the basket than a non-fuel member. And so it just shows the importance of engagement and the importance of leaning in, in these periods where wallets are stretched to provide these price points for customers that they find attractive. On inflation, like-for-like inflation was a little more than 1% in the quarter, but through the quarter, obviously, we saw fuel prices go up and as you think about a category like food, it's heavily dependent upon fertilizer and nitrogen and phosphates are heavily dependent upon the Strait of Hormuz and the closure there. So I think it's possible that if fuel prices persist at this level, you may see some upward pressure on average unit retail prices. Egg deflation for us, lastly, contributed probably to about almost a 100 basis points of deflation in that like-for-like inflation number, meaning it would have been higher. We'll begin lapping periods as we go through the year where eggs weren't as high. So you're right that, that could also put some upward pressure on the year-over-year inflation number that's printed. John David Rainey, CFO Meanwhile, NVDA’s earnings reinforce that we are still int he early innings of the AI Capex super cycle. Traders are looking at the issues WMT flagged as temporary, and this as a multi year trend. We will see, but in the meantime they are not making this hard…. What Iran Tells Us About UFO Disclosure When governments confront unknown threats in their airspace, defense budgets surge and the same aerospace and surveillance companies move hardest. On March 2nd, Northrop jumped 6% and Lockheed 3.3% on the Iran news — and President Trump has since ordered the formal release of government UAP files, with the Pentagon confirming compliance. So if a conventional conflict can move these stocks this fast, what happens when the bigger story breaks? See the UFOD holdings: [thetruthisoutthereufod.com ETF News Meet the Investor Betting UFO Disclosure Will Move Markets — Liberation Times | Reimagining Old News Matthew Tuttle is no stranger to unconventional investments. But with the Tuttle Capital UFO Disclosure ETF – ticker symbol UFOD – the chief executive and chief investment officer of Connecticut-based Tuttle Capital Management may have made the defining bet of his career: that UFO disclosure will Liberation Times | Reimagining Old News A Stock I’m Watching $INFQ ( ▲ 11.22% ) Usually my favorite quantum stock. Up big again this morning on the news about US Government grants. In Case You Missed It Great talk on with the Acquirers Podcast on markets, value investing, inverse Cramer, and Michael Gayed joins me to talk about taking income from your portfolio and how to get more than 4%…… The H.E.A.T. (Hedge, Edge, Asymmetry and Theme) Formula is designed to empower investors to spot opportunities, think independently, make smarter (often contrarian) moves, and build real wealth. The views and opinions expressed herein are those of the Chief Executive Officer and Portfolio Manager for Tuttle Capital Management (TCM) and are subject to change without notice. The data and information provided is derived from sources deemed to be reliable but we cannot guarantee its accuracy. Investing in securities is subject to risk including the possible loss of principal. Trade notifications are for informational purposes only. TCM offers fully transparent ETFs and provides trade information for all actively managed ETFs. TCM's statements are not an endorsement of any company or a recommendation to buy, sell or hold any security. Trade notification files are not provided until full trade execution at the end of a trading day. The time stamp of the email is the time of file upload and not necessarily the exact time of the trades. TCM is not a commodity trading advisor and content provided regarding commodity interests is for informational purposes only and should not be construed as a recommendation. Investment recommendations for any securities or product may be made only after a comprehensive suitability review of the investor’s financial situation.© 2026 Tuttle Capital Management, LLC (TCM). TCM is a SEC-Registered Investment Adviser. All rights reserved. Keep Reading May 21, 2026 • 2 min read Of the four pillars, one is doing the heavy lifting right now. This is the most important moment for Themes I've seen in years. May 21, 2026 • 11 min read The Software Industry Just Repriced Itself — And Wall Street Hasn’t Noticed A $500 billion subscription model is being dismantled in public. The companies breaking it on purpose are the ones to own. The ones pretending it isn’t happening are the ones being marked down by their own customers. May 6, 2026 • 4 min read What They Do > What They Say View more 4