최대 AI 강세론자가 반도체에 대해 약세인가?

THE DAILY H.E.A.T. Torching Wall Street’s Obsolete Playbook. Delivered straight to your inbox 5 days a week. Subscribe Home Posts Is The Biggest AI Bull Bearish On Semiconductors? Is The Biggest AI Bull Bearish On Semiconductors? Leopold Aschenbrenner’s Q1 filing is not a clean semiconductor short. It is a relative-value map of the AI buildout: long powered capacity, storage, and compute—with substantial downside protection against the most crowded part of the stack. Jun 23, 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. Leopold Aschenbrenner’s Q1 2026 13F did not say “sell every chip stock.” It said something more interesting. At March 31, Situational Awareness LP held put options reported against approximately $8.46 billion of semiconductor-related underlying value—including SMH, Nvidia, ASML, Broadcom, AMD, Micron, Intel, TSMC, and Oracle. But a Form 13F does not disclose strikes, expirations, premiums, delta-adjusted exposure, written options, or the rest of the fund’s book. The SEC explicitly notes that short equity positions and written options are omitted from 13F reporting. What the filing reveals is a snapshot of long positions and long option positions, as of a date now 83 days in the past. The same filing showed large positions in Bloom Energy, CoreWeave, Applied Digital, Iris Energy, Core Scientific, SanDisk, and a basket of powered-land operators. That is not a clean “AI bull turns chip bear” story. It looks more like a relative-value trade: own the scarce physical capacity required to deploy AI, while purchasing downside protection against the part of the stack carrying the highest expectations. The signal is not that chips are finished. The signal is that chips without power are inventory. “The filing cannot establish the fund’s true net semiconductor exposure. What it can establish is a clear thesis about where physical AI capacity is scarce.” WHAT THE 13F ACTUALLY SHOWS Aschenbrenner’s Situational Awareness LP filed its Q1 2026 13F on May 18, covering positions as of March 31. The information table reports $13.68 billion in disclosed holdings. That figure does not equal AUM—it excludes foreign securities, private positions, cash, swaps, and omits short equity positions entirely. The put options reported against semiconductor names have underlying reference values of approximately: $2 billion against SMH (the semiconductor ETF), $1.6 billion against Nvidia, plus Broadcom, AMD, ASML, Intel, TSMC, and Oracle. The filing does not disclose whether these are outright bearish bets, portfolio hedges, spread legs, or convexity purchases against long positions elsewhere in the book. Alongside those puts, the filing shows extensive semiconductor exposure on the other side. Micron had approximately $422 million in call-underlying value plus $315 million in put-underlying value. TSMC had approximately $355 million in calls alongside puts. SanDisk had $724 million in common shares plus $389 million in call-underlying value. These mixed structures—long calls and long puts on the same name—suggest sophisticated positioning that a simple “short the sector” narrative does not capture. THE LONG SIDE: PHYSICAL AI CAPACITY Where the filing is unambiguous is the long book. Aschenbrenner’s disclosed longs map the physical layer of the AI buildout—not the GPU designers, but the power operators, the data center hosts, the storage infrastructure, and the hyperscale compute lessors. Company Category Q4 2025 Q1 2026 What It Represents Bloom Energy (BE) Power generation $1.07B (10.1M sh) $879M (6.5M sh) On-site fuel cell power for data centers. Large position—but reduced 36% QoQ. CoreWeave (CRWV) Hyperscale compute New position Increased GPU cloud capacity leased to AI labs. Tangible AI revenue, not optionality. Applied Digital (APLD) AI data center infra Held Increased ~19% AI-native facility operator. Real AI/HPC cash flows, growing. Iris Energy (IREN) Renewable HPC Held Increased ~34% Renewable-powered HPC hosting. AI revenue expanding vs. crypto. Core Scientific (CORZ) Data center hosting Held Decreased ~10% HPC hosting with AI colocation. Modest reduction. SanDisk (SNDK) Storage $0 $724M common + $389M calls Large new storage position. Storage is a bottleneck in AI inference pipelines. CleanSpark (CLSK) Powered-land optionality Small Increased ~648% Bitcoin mining operator with multi-GW power assets. AI pivot is prospective—Texas AI campus energizes 1H 2027. Riot Platforms (RIOT) Powered-land optionality Held Increased ~86% Grid-scale power assets. AI conversion is longer-dated optionality, not current revenue. IREN / BTDR / Others Powered-land optionality Held Various increases Similar profile: valuable power and interconnection assets; AI revenue early or prospective. Two categories emerge clearly. The first—CoreWeave, Applied Digital, Iris Energy, Core Scientific—has real and growing AI cash flows today. The second—CleanSpark, Riot, and the Bitcoin mining operators—holds valuable power and land assets where AI conversion is the thesis, not yet the income statement. Aschenbrenner is buying the substations. But he is buying the option value of powered land, not necessarily current AI cash flow. The distinction matters for how quickly these positions can pay off.   THE BOTTLENECK IS MIGRATING—BUT CHIPS ARE STILL SELLING The hyperscalers are not stepping back from semiconductor procurement. Alphabet reported $35.7 billion in Q1 2026 capital expenditures, with approximately 60% of technical-infrastructure investment directed at servers. Nvidia has confirmed that AWS, Google Cloud, Microsoft, and Oracle will be among the first to deploy Vera Rubin systems. OpenAI and Nvidia have announced a planned 10-gigawatt deployment involving millions of GPUs. What changed is not the volume of chip orders. What changed is that chips alone no longer determine whether AI systems get deployed. Power interconnection, cooling, and permitting now co-determine deployment timelines. A data center full of Blackwells that cannot be energized is not a data center. It is a warehouse. This is entirely consistent with Aschenbrenner’s published thesis in Situational Awareness: The Decade Ahead. His 2024 analysis projected that AI chip demand by 2030 could represent multiples of current leading-edge capacity, and that many trillions of dollars would flow into GPU, data center, and power infrastructure before decade-end. His conclusion on NVDA and TSM at the time was that the buildout was “still not even close to fully priced in.” The put positions may express a view on current valuations, on timing, on cyclicality, or on capex digestion after an enormous spending surge. They do not, based on any public statement he has made, express a view that chip demand is about to disappear. Those are very different bets. SITUATIONAL AWARENESS LP: BY THE NUMBERS Situational Awareness LP — What Is Publicly Known Fund inception: September 2024 | Starting AUM: $225 million Press-reported AUM (WSJ, citing people familiar): above $20 billion Press-reported return through May 2026 (WSJ): approximately +270% after fees 13F information-table value as of March 31: $13.68 billion (not AUM; excludes foreign, private, cash, shorts) WSJ figures are press-reported, not audited public disclosures. Earlier-cited +100% YTD / 96pts alpha figures reflect prior public reporting and may be materially superseded.   WHAT THE FILING SIGNALS The following table reflects what can and cannot be concluded from the disclosed positions. Tickers appear in the downside-protection row because put positions were reported; that does not establish a directional view, time horizon, or net exposure. Signal Disclosed Names Correct Framing Large long exposure BE, CRWV, APLD, IREN, CORZ, SNDK Conviction longs across power generation, AI compute hosting, and storage. Mixed QoQ changes—SNDK is the largest new addition. Powered-land optionality CLSK, RIOT, and mining-adjacent names Valuable power and interconnection assets. AI conversion is a multi-year thesis, not current cash flow for most. Reported downside protection SMH, NVDA, AVGO, AMD, ASML, ORCL Put positions held as of March 31. The filing cannot establish strikes, delta, net exposure, or intent. Mixed option structures—do not classify MU, TSM Long calls and long puts reported on the same name. Sophisticated structures that resist simple directional labeling. Broader physical infrastructure Solaris Energy, Power Solutions Int’l Smaller positions suggesting continued tracking of the full power-delivery stack.   FIVE TAKEAWAYS 1. The filing cannot be read as “sell chips.” It can be read as: own physical AI capacity, purchase protection against the most-crowded layer of the stack. Those are different statements with different investment implications. 2. Power is joining compute as a binding constraint—not replacing it. Hyperscalers spent over $35 billion in Q1 capex, with the majority going to servers. The chips are still selling. The new scarcity is the watts to run them. 3. The powered-land thesis and the AI cash-flow thesis are not the same trade. CoreWeave, Applied Digital, and Iris Energy have AI revenue today. CleanSpark and Riot have power assets and a multi-year conversion roadmap. Sizing and time horizon should differ accordingly. 4. Categories are lying. Wall Street labels CLSK, RIOT, APLD, IREN, and CORZ as “crypto infrastructure.” The underlying cash-flow mix is shifting toward AI, HPC, and power. The index label is misdirecting capital away from the actual exposure. 5. A $20-billion fund has different portfolio-construction needs than a $225-million launch fund. The puts may reflect downside protection, liquidity management, relative-value positioning, or convexity against the long book. The 13F does not reveal which. Do not size a position based on what you assume he is doing.   CREDIBILITY FIREWALL SOURCED / DISCLOSED PRESS-REPORTED / DIRECTIONAL EDITORIAL VIEW Put option underlying values per Q1 2026 13F (SEC filing, filed May 18, positions as of March 31): ~$8.46B semiconductor-related. $2B SMH, $1.6B NVDA. $13.68B total information-table value. $879M Bloom Energy (6.5M shares, reduced from 10.1M in Q4). SanDisk: $724M common + $389M call-underlying. CoreWeave, APLD, IREN, CORZ: all confirmed in filing. CleanSpark increased ~648%, Riot ~86%. AUM above $20B and approximately +270% after fees through May 2026: WSJ reporting, citing people familiar with the fund. Not audited public disclosure. Alphabet Q1 2026 capex $35.7B with ~60% to servers: Alphabet investor relations. NVDA Vera Rubin / OpenAI 10GW deployment: company announcements. CleanSpark Texas AI campus 200MW+ energizing 1H 2027: CleanSpark investor relations. The “relative-value” and “bottleneck migration” framing is our analytical lens, not Aschenbrenner’s stated thesis. The two-category split (AI cash flows today vs. powered-land optionality) is editorial judgment. The conclusion that the puts may be hedges or convexity rather than directional shorts is our inference from 13F limitations—not disclosed by the fund. The AI Buildout Has a Physical Layer Many of today’s data centers are still using copper wiring. The same metal we’ve been using for a hundred years. At the speeds AI demands with data moving between thousands of GPUs, billions of times a second, copper doesn’t just slow down. It turns that data into heat. The more you push through it, the worse it gets. There’s no software for fix for that. So what’s the answer? Explore the Photonics Layer….. Tuttle Capital Pure Play Photonics ETF (FOTO) Distributor: Foreside Fund Services | Investing involves risk including possible loss of principle. News vs. Noise: What’s Moving Markets Today The news. The KOSPI's 9.99% collapse grabs the headline. The more durable story is that AI is finally being priced as a capital-intensive business. Companies are moving from tokenmaxxing to tokenminimizing because cheaper tokens have not produced cheaper AI bills — agentic workloads simply consume more of them. Alphabet has announced up to $84.75 billion of equity financing after raising more than $85 billion of debt over the previous year, then lost Noam Shazeer to OpenAI and John Jumper to Anthropic in consecutive weeks. SpaceX is taking its $20 billion bridge loan into the bond market. The AI story suddenly comes with debt, dilution, and talent leakage. That is not rotation. It is repricing. Fed-funds futures now imply roughly 50 basis points of tightening by December, with Thursday's PCE expected near 4%. The cost of capital is back, and equities are finally getting the memo. ([Reuters](https://www.reuters.com/world/asia-pacific/south-koreas-kospi-plunges-nearly-10-after-regulator-cautions-leveraged-etfs-2026-06-23/)) The noise. Hormuz is improving, not normalized. Tankers are moving again, but traffic remains well below the roughly 125 daily crossings seen before the war — Reuters counted 55 merchant ships on Saturday and only five on Sunday. The U.S. oil waiver expires August 21, Washington and Tehran are still publicly describing the nuclear commitments differently, and WTI around $74 Monday was a rational unwind of the worst-case outcome. It was not proof that safe passage, insurance, and final treaty terms have been fixed. A chokepoint that can be reclosed between negotiating sessions is not resolved. ([Reuters](https://www.reuters.com/world/middle-east/shipping-slows-after-iran-says-it-has-again-shut-strait-hormuz-2026-06-21/)) Where Does the Money Go When AI Hits a Wall? When capital chases a tech theme, it tends to pile into the most obvious layer and miss the one underneath. AI spending is now bumping hard against memory. Hyperscalers — the big cloud builders like Amazon, Google, and Microsoft — have shifted memory from 8% of their build budgets to an estimated 30% in a single cycle. That capital has to go somewhere. If the constraint is memory, and the build can't move without it, shouldn't an investor own the layer AI runs on? View HBMX fund holdings → Distributor: Foreside Fund Services | Investing involves risk including possible loss of principal. <Link = http://www.hbmxetf.com/> ETF News Tuttle Capital Adds SpaceX to SPCI, UFOD and HALX Following Historic IPO The Tuttle Capital Space Industry Income Blast ETF, UFO Disclosure ETF and Heavy Assets Low Obsolescence ETF Add Positions in Space Exploration Technologies Corp. (SPCX)Tuttle Capital Management announces that the Tuttle Capital Space Industry Income Blast ETF (CBOE BZX: SPCI), the Tuttle Capital UFO Disclosure ETF (CBOE BZX: UFOD), and the Tuttle Capital Heavy Assets Low Obsolescence ETF (CBOE BZX: HALX) have added positions in Space Exploration Technologies Corp. (NASDAQ: SPCX) following... Yahoo Finance A Wall Street investor built a fund around UFO disclosure | Reality Check Ross Coulthart explores untold stories using a fact-based approach to address various topics, from unidentified aerial phenomena to other mysteries that often go unreported. In this episode of &#82… NewsNation A Stock I’m Watching $CRWV ( ▼ 5.01% ) This is one of the only green names on my screen this morning :( CoreWeave is being valued like a cloud company, but it should be analyzed like a structured-finance vehicle attached to a fleet of GPUs. The demand is real: first-quarter revenue doubled to $2.08 billion, backlog reached $99.4 billion, 98% of revenue came from committed take-or-pay contracts, Meta added a $21 billion commitment through 2032, and Jane Street signed a $6 billion compute deal and bought $1 billion of stock at $109. But only 36% of CoreWeave’s $98.8 billion in remaining performance obligations is expected to become revenue during the next two years, while the capital has to be spent now; two customers still accounted for 65% of quarterly revenue. In the first quarter, customer cash helped produce $3 billion of operating cash flow, but CoreWeave used $7.7 billion in investing activities, paid $536 million of net interest, and ended March with $25.1 billion of debt, $10.1 billion of lease liabilities, and another $40.7 billion of signed leases that have not even commenced. Management’s 56% adjusted EBITDA margin is the number bulls love. The number I care about is $1.1 billion of quarterly depreciation on property and equipment: adjusted operating margin was 1%, and GAAP operating margin was negative 7%. The stock works if CoreWeave can keep converting take-or-pay contracts into roughly 5.9% customer-backed project debt instead of relying on unsecured dollar bonds costing 9.625%. That financing spread—not the backlog headline—is the thesis. Monday’s $111.29 close was almost exactly Jane Street’s entry. The demand is real. So is the bill required to deliver it. In Case You Missed It Great conversation on wide ranging topics with Kenny Polcari… AI vs. inflation: why Wall Street is ignoring soaring rates Does AI hype have Wall Street ignoring inflation? Trader Talk host Kenny Polcari discusses the growing concerns with NYSE Senior Market Strategist Michael Reinking, and Tuttle Capital Management CEO &amp; CIO, Matthew Tuttle Yahoo Finance 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 Jun 12, 2026 • 4 min read Don’t forget this AI Bottleneck The AI trade might be missing its biggest constraint. Jun 22, 2026 • 4 min read Where the income comes from How the strategy seeks to generate income. Jun 9, 2026 • 1 min read The GPU trade is closing. What comes after it splits three ways. View more 2