1000억 달러의 해자

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 $100 Billion Moat The $100 Billion Moat Why no one can build the next credible Neocloud. Jun 17, 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. The AI Buildout Has a Physical Layer News vs. Noise: What’s Moving Markets Today ETF News In Case You Missed It H.E.A.T.   ~$100B Chamath estimate: cost of 1 GW of compute (2026) 684% Nebius Q1 2026 revenue growth YoY $27B Potential Meta–Nebius contract value 46% CAGR Neocloud market to 2031   On March 3rd, 2026, Chamath Palihapitiya posted a number on X that most investors scrolled past without understanding what they had just read. He was describing a data center project in Arizona. When the project started, one gigawatt of compute cost four to five billion dollars. Today, Chamath said, that same gigawatt costs one hundred billion dollars. The capital intensity went up twenty times in a single investment cycle. Most people read that as a warning that AI infrastructure is expensive. They are reading it backwards. The correct interpretation is this: any company that built its neocloud before that inflection happened is now sitting behind a capital barrier that essentially no new competitor can cross. Not because building a data center is technically impossible. Because gigawatt-scale AI infrastructure now requires simultaneously locking in four things that are all scarce at the same moment: secured grid power, Nvidia GPU allocations, anchor customer contracts, and the capital to carry it. The companies that crossed that threshold before the cost exploded did not just get lucky. They have a structural advantage that compounds with every passing GPU generation. This is the neocloud trade. But it is not a blanket trade on everything that calls itself a GPU cloud. It is a contracted utilization trade — and the distinction matters enormously.  THE MARKET NOBODY BUILT FOR EVERYTHING Hyperscalers — AWS, Azure, Google Cloud — are extraordinary businesses built to serve everything: storage, databases, enterprise software, networking, video delivery. Their GPU pricing reflects that full-stack overhead. Every H100 they rent has to amortize the fixed cost of maintaining infrastructure their customers use for non-AI workloads. It has to satisfy the margin expectations of a public company with a ten-figure cloud franchise to defend. Neoclouds have none of that overhead. They build for one workload: AI compute. The result is a meaningful and durable cost advantage on identical Nvidia silicon. Public list price data shows the spread is real across every measured configuration — and the advantage is not simply cheaper list pricing. It is cluster design, power efficiency, utilization economics, and the anchor-customer financing that lets a neocloud amortize its infrastructure over committed revenue rather than spot demand. The market sizing reflects this. Industry estimates put the neocloud sector at $35 billion in 2026, growing to $236 billion by 2031 at a 46% compound annual rate — faster than cloud computing grew in its own first decade. Gartner projects neoclouds capturing 20% of the $267 billion AI cloud market by 2030.   CONTRACT QUALITY: THE METRIC THAT MATTERS The neocloud trade splits cleanly into two completely different equity stories: Contracted AI Factory: Take-or-pay capacity agreements. Customer prepayments. Secured grid power. Nvidia roadmap relationships. These operators are AI infrastructure landlords. Depreciation is covered by committed revenue. Uncontracted GPU Rental: Same hardware. No anchor customers. No power lock. No Nvidia roadmap partnership. Renting H100s into a softening spot market as Blackwell supply improves. This is the Lucent risk: revenue looks fine until the customers stop re-signing. The market has not yet fully priced this distinction. It will.  THE REBUILDING OF A SEARCH ENGINE EMPIRE  Of all the publicly traded proxies for the contracted neocloud trade, Nebius Group is the most interesting — and arguably the most mispriced. Nebius is the infrastructure arm of the former Yandex, Russia's Google equivalent. When Russia invaded Ukraine in 2022, founder Arkady Volozh separated the international infrastructure assets from the Russian operation and relisted the rebuilt company on Nasdaq in October 2024. The team that built Yandex already knew how to run internet-scale compute at the lowest possible cost — exactly the operational DNA a neocloud requires. In Q1 2026, Nebius reported revenue of $399 million, up 684% year-over-year. Then, in March 2026, the company announced the most significant enterprise infrastructure contract of the year.   THE META DEAL  |  SOURCED: NASDAQ PRESS RELEASE, MARCH 2026 Structure: $12 billion in committed dedicated GPU capacity beginning early 2027, plus up to $15 billion in additional capacity tied to Meta purchasing Nebius's unsold third-party compute. Total potential contract value: up to approximately $27 billion. Platform: One of the first large-scale deployments of Nvidia's Vera Rubin architecture — the generation after Blackwell. Nvidia relationship: Nvidia made a $2 billion equity investment in Nebius and is partnering to support deployment of more than 5GW of Nvidia systems by 2030. Post-contract guidance: $7 billion to $9 billion in annualized recurring revenue for 2026 — 540% year-over-year growth. The $12 billion committed tranche is what matters structurally. It is not a letter of intent or a capacity option. It is contracted infrastructure spend with a defined start date. The additional $15 billion is tied to Meta's actual consumption of third-party capacity — real optionality, not a phantom number. The take-or-pay structure is exactly the contract quality that separates Tier 1 neoclouds from the rest of the market.  WHY GPU ALLOCATION IS A STRATEGIC RELATIONSHIP, NOT A TRANSACTION  Nvidia does not sell to the highest bidder. GPU allocation is a relationship asset earned through proof of operational capability. Nvidia allocates its most advanced silicon to operators who have demonstrated they can deploy at scale, on time, with the power and organizational infrastructure to make the hardware productive. The operators who delivered Blackwell deployments on schedule receive Vera Rubin allocations. The operators still ramping their first H100 cluster go to the back of a queue that stretches years. This is why the Nvidia equity stake in Nebius matters more than it might appear. It is not a financial investment. It is a signal of roadmap alignment — a statement that Nebius is inside the allocation relationship in a way that smaller operators are not. The moat compounds with each generation for the operators inside that relationship. It does not compound permanently. The AMD wildcards — Meta's 6GW Instinct commitment and OpenAI's separate 6GW AMD agreement — are early evidence that hyperscalers are beginning to build Nvidia alternatives. That does not threaten the 2026 neocloud thesis. It is the 2028 question that every holder of Tier 1 neocloud equity should be tracking.   “The bottleneck migrated from silicon to capital. The next migration is to power. Follow it.”  WINNERS, LOSERS & THE TIERED LANDSCAPE   TIER TICKER VERDICT WHY IT WINS / LOSES KEY RISK 1 NBIS WINNER — Contracted Neocloud $12B committed Meta capacity + up to $15B option; 8x YoY revenue growth; one of first large-scale Vera Rubin deployments; Nvidia $2B equity stake & 5GW partnership Russia-origin headline risk; single anchor customer concentration 1 CRWV WINNER — Contracted Neocloud Largest public GPU cloud; IPO'd March 2025; long-term hyperscaler contracts; first-mover AI cloud platform $30–35B projected 2026 capex; high leverage; $66.8B backlog execution risk 1B NVDA WINNER — GPU Allocator Every neocloud GPU hour = Nvidia silicon; allocation relationships weaponized as strategic moat across generations Already priced for perfection; AMD/custom silicon long-cycle displacement risk 1B ETN / VRT / PWR WINNER — Power Delivery Gigawatt-scale AI factories require massive electrical infrastructure; power delivery is the next confirmed bottleneck after silicon Utility permitting delays; grid interconnection backlogs 2 SMCI / DELL WINNER — Rack Integration AI server / rack-scale integration beneficiary; SMCI planning $7B equity raise to meet $39B in AI server orders SMCI accounting history; customer concentration; capital dilution 2 DLR / EQIX WINNER — Data Center RE Powered shell capacity in constrained markets; landlord economics on AI infrastructure land rush Hyperscaler build-own pressure; rate sensitivity on long-duration assets Pressure AMZN / MSFT / GOOGL PRESSURE Full-stack overhead creates structural GPU pricing premium vs. neoclouds; enterprise budget migration accelerating Won't disappear; bundled services and enterprise lock-in remain powerful Hedge AMD CHALLENGER HEDGE Meta 6GW Instinct GPU partnership + OpenAI 6GW agreement signal hyperscalers actively diversifying from Nvidia dependency; not core to today's neocloud stack but the right 2027–2028 watch Neocloud operators are Nvidia-first today; AMD software ecosystem still narrower Avoid Uncontracted spot GPU clouds AVOID Same hardware as Tier 1 neoclouds; none of the contracted utilization, anchor customers, or Nvidia roadmap alignment. This is where the Lucent risk lives. Depreciation vehicles dressed as AI infrastructure plays  PRESSURE POINTS PRESSURE POINT WHO FEELS IT TIMELINE MAGNITUDE Capital moat (gigawatt-scale buildout) New neocloud entrants Now — structural Existential GPU allocation scarcity Operators without Nvidia roadmap relationships 2026–2027 Critical Power / grid interconnection constraints All data center operators at scale 2026–2030 High Hyperscaler GPU margin compression AWS / Azure / GCP AI divisions 2026–2027 budget cycles High Vera Rubin transition timing Operators not yet in Nvidia partnership 2027 Medium AMD / custom silicon displacement Nvidia's allocation leverage, long-cycle 2027–2029 Low–Medium Uncontracted GPU cloud implosion Spot GPU renters without anchor demand 2026–2027 High within cohort  CREDIBILITY FIREWALL What is sourced and confirmed versus where we are drawing conclusions. SOURCED / REPORTED DIRECTIONAL INFERENCE Nebius Q1 2026 revenue: $399M, up 684% YoY; guided to $7B–$9B annualized run-rate by year-end 2026 (Yahoo Finance, Nebius earnings release) Nebius may reach $10B+ ARR ahead of guidance if Meta deployment ramps early Meta–Nebius: $12B committed dedicated GPU capacity beginning early 2027 + up to $15B additional capacity option; total potential value up to ~$27B (Nasdaq press release, March 2026) The deal's take-or-pay structure likely de-risks Nebius's balance sheet for 2+ GPU generations Nebius is among the first large-scale Nvidia Vera Rubin platform deployments; Nvidia made a $2B equity investment in Nebius and is partnering to deploy 5GW+ of Nvidia systems by 2030 (Nvidia Newsroom) Implies unusually strong Nvidia roadmap alignment — but formal 'priority allocation' ahead of all competitors is an inference, not a confirmed term Chamath Palihapitiya stated 1 GW compute cost rose from ~$4–5B to ~$100B in one investment cycle (X, March 2026); noted as dramatic illustration of capital intensity, not independently audited infrastructure cost breakdown Gigawatt-scale AI factories are now strategic-capital projects potentially running into the tens of billions when GPUs, power, networking, cooling, land, financing, and refresh cycles are included CoreWeave IPO priced March 2025, trades as CRWV; projected $30–35B of 2026 capex, more than 2x 2025; $66.8B contracted backlog (Reuters, February 2026) High capex + high leverage = rate-sensitive equity; backlog quality and execution pace are the key variables AMD and Meta announced partnership for up to 6GW of AMD Instinct GPU deployments, first 1GW in 2H 2026; AMD also has a separate 6GW agreement with OpenAI (AMD newsroom, February 2026) Hyperscalers are actively building Nvidia alternatives; neoclouds remain Nvidia-first today but the AMD supply chain is a legitimate 2027–2028 watch Neocloud market: $35B in 2026, projected $236B by 2031 at 46% CAGR; Gartner projects 20% AI cloud market share by 2030 (industry estimates) Market capture projections assume contracted utilization models dominate; uncontracted spot GPU capacity likely does not survive to 2031 at scale  BEAR CASE SPOTLIGHT The most important risk in the neocloud thesis is not AMD, not hyperscaler retaliation, and not GPU pricing compression. It is contract quality — and the investor's inability to distinguish it from the outside. The neocloud trade only works at the Tier 1 end of the spectrum: contracted utilization, take-or-pay capacity agreements, customer prepayments, secured power, and Nvidia roadmap relationships. Nebius and CoreWeave have varying degrees of all of those. But a significant number of companies marketing themselves as neoclouds have none of them. They are renting uncontracted H100s into a spot market that has already started to soften as Blackwell supply improves. Same hardware. Completely different equity story. The Lucent parallel from our Cheap Tokens issue is precise: in 2000, Lucent's revenue looked extraordinary right up until the moment its customers stopped buying. Uncontracted GPU clouds are financing depreciation today and calling it growth. The second risk is hardware cycle timing. Neoclouds are financing gigawatt-scale infrastructure against long-term contracts that may or may not survive a silicon transition. Operators who locked in ten-year power agreements for Blackwell-era deployments could find themselves holding partially stranded assets when Vera Rubin's successor arrives and customers demand the new architecture. The contracted neocloud landlord weathers that transition because their customers re-sign. The uncontracted spot GPU operator does not. The third risk is the AMD wildcard. Neoclouds are structurally Nvidia-first today. But Meta's 6GW AMD Instinct commitment and OpenAI's separate 6GW AMD agreement are not small bets. If AMD's software ecosystem reaches parity with CUDA for the dominant inference workloads — a 2027–2028 question, not a 2026 one — Nvidia's allocation leverage over neoclouds weakens. That does not break the thesis. But it means the moat is not permanent by default. It has to be renewed every hardware generation. FIVE TAKEAWAYS  1. The neocloud moat is not GPUs. It is contracted utilization, secured power, and Nvidia roadmap alignment — and it has to be renewed every generation. The companies that have all four — committed capacity agreements, take-or-pay customers, grid power, and confirmed next-generation silicon access — are AI infrastructure landlords. The companies that have only the hardware are depreciation vehicles. 2. Nebius is the most interesting public proxy for this trade, and it is not priced like one. Q1 2026 revenue up 684% year-over-year. A $12B committed Meta capacity agreement, with up to $15B in additional options. One of the first large-scale Nvidia Vera Rubin deployments, backed by a $2B Nvidia equity stake. The discount exists because the company was incorporated out of Russian internet infrastructure. That headline risk is real. So is the operational DNA. 3. CoreWeave is the better-known name, but $30 to $35 billion in projected 2026 capex makes it a leveraged bet on backlog execution. The $66.8B contracted backlog is real. So is the financing structure required to deploy it. CoreWeave is not wrong. It is rate-sensitive. The spread between its cost of capital and its contract yields is the whole equity story. 4. The bottleneck has already migrated from silicon to power and capital — and the next migration is to grid interconnection. In 2023, the constraint was H100 availability. In 2026, the constraint is financing a gigawatt of infrastructure that starts at tens of billions before the first GPU is racked. By 2027, the constraint will be high-voltage grid capacity and transmission infrastructure. Follow the bottleneck: ETN, VRT, PWR, and GEV are not AI plays in the traditional sense, but they are AI infrastructure plays in the correct sense. 5. Not all neoclouds are the same trade. Know which tier you own. Contracted AI factory with Nvidia roadmap alignment: structurally sound. Uncontracted spot GPU rental capacity: avoid. The market has not yet fully priced the distinction. It will. 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 This is going to be interesting. We know Trump wants rates cut, we know how he treated Powell when he didn’t play ball, and we know that Trump is telling everyone higher inflation is transitory……. Trump Picked Warsh to Cut Rates. His Committee Is Talking About Hikes. At his first meeting this week, the new Fed chairman confronts an economic turn that began before he arrived www.wsj.com/economy/central-banking/trump-picked-warsh-to-cut-rates-his-committee-is-talking-about-hikes-3814bcaf?mod=hp_lead_pos10 Higher rates is one thing that can mess with this rally. Tech sold off yesterday, could have been profit taking after a huge up day, or could have been this…. MSFT Falls 2% On Tuesday — Microsoft Reportedly Walks Away From $3B Oracle Cloud Deal Microsoft has reportedly abandoned talks with Oracle to lease $3 billion in AI computing power after a stalemate over federal security regulations. Yahoo Finance Reports are now coming out that this may not be accurate, and this morning the NASDAQ 100 has retraced almost half yesterday’s losses. $SPCX ( ▼ 4.95% ) is up again pre market. Yesterday it took all the oxygen away from the other space stocks. So far today those are looking green. As I talked about in yesterday’s webinar, the space trade is still one of my favorites as we aren’t even in the early innings of what will end up in space. 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 $SPCI ( ▲ 0.97% ) Holdings Update We added $SPCX ( ▼ 4.95% ) to SPCI Yesterday: For a full list of holdings, visit: https://www.incomeblastetfs.com/etf/spci Distributor: Foreside Fund Services, LLC A Stock I’m Watching $SATS ( ▼ 7.66% ) From Oppenheimer….. Forced selling is done, not fundamental. SATS sold off as arbitrageurs who had built synthetic SPCX longs (using SATS as the primary SpaceX proxy) unwound en masse post-IPO. That exit was purely mechanical not a reassessment of SATS's intrinsic value leaving the stock trading at an estimated ~50% discount to its residual SpaceX NAV exposure. In Case You Missed It 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 10, 2026 • 1 min read Carnegie owned the rail, not the railroad The same pattern keeps repeating. Jun 11, 2026 • 4 min read Restructuring the AI memory trade This time it’s built differently. Jun 15, 2026 • 12 min read Washington Just Made AI Sovereignty Investable How a simple decision could open up the floodgates to AI investments across the globe. View more 2