차세대 AI 병목 현상

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 Next AI Bottleneck The Next AI Bottleneck The next AI bottleneck is not chips, memory, or megawatts. It is the time-to-power between a hyperscaler’s substation order and an energized site. Jun 1, 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 Webinar Today 2pm EST….. Welcome! You are invited to join a webinar: Tuttle Capital x Lead-Lag - HALO Webinar. After registering, you will receive a confirmation email about joining the webinar. Join Matt Tuttle of Tuttle Capital Management and Michael A. Gayed, CFA of Lead-Lag for a 60-minute educational discussion on the Heavy Assets, Low Obsolescence (HALO) lens. The conversation will explore the physical-infrastructure layer beneath the AI trade, including mines, railroads, energy companies, and power utilities, and how advisors can think about sector-level portfolio implications if the AI capex cycle broadens from compute toward infrastructure and materials. Topics will include: - What "heavy asset, low obsolescence" filters for as a framework - Why the AI capex cycle may rotate from compute toward infrastructure and materials - Sector-level implications for advisor portfolios - Live Q&A with Matt Tuttle, Michael Gayed, and RIA panelists CFP CE credit: 1.0 hour, Investment Planning, Intermediate. Date/time: Monday, June 1, 2026 at 2:00 PM ET. Registration: https://us06web.zoom.us/webinar/register/WN_FDWhyy9MTMy_wsGO2Tukyw This is an educational thought-leadership webinar. No specific product is being marketed. Disclaimer: This is sponsored advertising content for which Lead-Lag Publishing, LLC has been paid a fee. The information provided is solely the creation of Tuttle Capital Management. Lead-Lag Publishing, LLC does not guarantee accuracy or completeness. All statements are the sole opinion of Tuttle Capital Management. Past performance does not guarantee future results. Investing involves risk including possible loss of principal. Nothing in this webinar should be construed as a recommendation to buy or sell any security or as personalized investment advice. Zoom H.E.A.T.    183 GW FIRM U.S. DATACENTER POWER COMMITMENTS — ~22% OF 2025 PEAK LOAD 68% DATACENTER SHARE OF PROJECTED U.S. POWER DEMAND GROWTH THROUGH 2035 $1.1T PROJECTED U.S. INVESTOR-OWNED UTILITY CAPEX 2025–2029 On May 29, 2026, the smartest people in finance were arguing about Nvidia’s next print. The wrong fight. The binding constraint on AI buildout is no longer compute. It is the service order from the utility. Specifically, the time it takes to go from a signed hyperscaler land deal to an energized site capable of pulling 500 megawatts through a substation that does not yet exist. There are 220 gigawatts of U.S. datacenter power demand currently in the pipeline. Of that, 183 gigawatts are already backed by firm commercial commitments — power purchase agreements, signed interconnection requests, or construction-stage utility contracts. That is roughly 22% of 2025 U.S. peak demand. It is not an Nvidia problem. It is a substation problem. The marginal buyer of grid infrastructure has changed. This is no longer ESG-mandate capital underwriting decarbonization stories. It is hyperscaler uptime capital paying for speed to power. Hyperscalers do not need a cleaner narrative. They need energized sites, backup generation, switchgear, transformers, smart meters, and utility partners that can move faster than the model-training cycle. The AI trade has moved from megawatts to delivery. THE THESIS The bottleneck has moved again. For two years, the binding constraint on AI buildout was compute. Then it was high-bandwidth memory. Then networking. Then power generation. Each migration created a wave of mispricing — and a window where the consensus narrative lagged the physical reality by twelve to eighteen months. We are now in the next window. The constraint has migrated from generating electricity to delivering it. Specifically, from the substation to the rack. The U.S. has 183 gigawatts of firm datacenter power commitments already in the queue. Datacenters are projected to consume 68% of total U.S. power demand growth through 2035. In the three grid regions where most hyperscaler buildout is happening — PJM, ERCOT, and MISO — datacenters will drive 63% of growth. You cannot energize 183 gigawatts of new datacenter load without rebuilding the distribution layer. You cannot rebuild the distribution layer without transformers, switchgear, advanced conductors, smart meters, behind-the-meter generation, electrical balance-of-system, and the boring industrial firms that make them. The companies that compress the months between a service request and an energized site are the ones who get paid first. Most of these companies are not in the AI ETFs. They are not in the Magnificent Seven discussion. They sit in indexes still labeled clean energy, utilities, or electrical equipment — categories that systematically lag the cross-category convergence happening underneath them. Solar installations fell 16% year-over-year in 2025 under the OBBBA. The narrative is dead. The cash flows are not. U.S. investor-owned utility capex is projected to hit $1.1 trillion over 2025–2029, with distribution — the part that touches the datacenter — growing fastest. The U.S. datacenter electrical-equipment market alone is forecast to expand from roughly $20 billion in 2024 to $65 billion by 2030. That money has to go somewhere. It is going to a small group of suppliers with the products, the certifications, and the scale to deliver inside the timeline. THE MECHANISM IN ONE SENTENCE When AI load growth overwhelms utility interconnection timelines, the companies that shorten time-to-power — distribution hardware, electrical equipment, behind-the-meter generation, and grid-edge intelligence — get repriced from clean-energy laggards to AI-infrastructure suppliers. The index labels are the last thing to change. Wall Street thinks this is clean energy. Hyperscalers think it is uptime. The market is still pricing the label — the cash flows are already pricing the load. TIERED WINNERS Four tiers. Tier 1A is the immediate behind-the-meter bypass trade. Tier 1B is the distribution hardware bucket Wall Street is still filing under “industrials” rather than “AI infrastructure.” Tier 2 is grid-edge intelligence. Tier 3 is contracted-cash-flow renewables. Tier 4 is the contrarian solar survivor trade. Names are starting points for analysis, not portfolio recommendations. TIER 1A  —  BEHIND-THE-METER BYPASS COMPANY TICKER MECHANISM WHY IT MATTERS NOW Generac GNRC Behind-the-meter standby & prime power Cleanest “bypass the interconnection queue” name; hyperscalers building on-site generation to avoid five-to-seven year utility waits EnerSys ENS Industrial batteries, datacenter UPS Rack-density-driven; battery backup load scales directly with thermal envelope, not exposed to solar ITC risk TIER 1B  —  DISTRIBUTION HARDWARE & ELECTRICAL EQUIPMENT The category Wall Street is still filing under “industrials.” U.S. datacenter electrical equipment is projected to triple from $20 billion to $65 billion by 2030. Transformers, switchgear, transfer switches, panelboards, and power distribution units are the physical chokepoint between a utility substation and an energized rack. Lead times are running 18–24 months. COMPANY TICKER MECHANISM WHY IT MATTERS NOW Eaton ETN Switchgear, power distribution, datacenter electrical Largest pure-play exposure to datacenter electrical buildout; bridges rack-to-grid Hubbell HUBB Utility distribution components, grid hardware Direct exposure to distribution-side capex, the fastest-growing line in $1.1T utility spend GE Vernova GEV Grid equipment, transformers, gas turbines Transformer order book runs into 2028; gas turbine backlog at multi-year highs nVent Electric NVT Electrical enclosures, connection & protection Underfollowed datacenter electrical exposure; smaller cap than ETN with cleaner exposure ratio Vertiv VRT Datacenter thermal & power infrastructure The bridge between datacenter power architecture and utility interconnection; consensus already crowded but cash flows still compounding Grid-construction services also matter here even if they fall outside a tight “hardware” bucket: Quanta Services (PWR), MasTec (MTZ), and MYR Group (MYRG) provide the labor and EPC capacity for the transmission and distribution buildout. Treat them as adjacent exposures, not core names. TIER 2  —  GRID-EDGE INTELLIGENCE COMPANY TICKER MECHANISM WHY IT MATTERS NOW Itron ITRI Smart meters, distribution sensors Distribution capex is the fastest-growing utility line; ITRI sits on the data layer Ormat ORA Geothermal baseload 24/7 carbon-free power matches hyperscaler PPA structure; strategic but slower-build than Tier 1A Landis+Gyr LAND.SW Smart metering, grid software European grid modernization plus U.S. utility pull-through; Swiss listing limits U.S. liquidity Tantalus Systems GRID CN Smart grid for mid-size utilities Underfollowed Canadian small-cap; thinly traded — size positions accordingly TIER 3  —  CONTRACTED-CASH-FLOW RENEWABLES COMPANY TICKER MECHANISM WHY IT MATTERS NOW Brookfield Renewable BEP Scaled renewable IPP Long-dated contracted cash flows; PPA repricing tailwind from datacenter demand HA Sustainable Infra HASI Specialty lender to clean energy projects High rates hurt the multiple; the loan book reprices to a higher coupon — sequencing matters TIER 4  —  CONTRARIAN SOLAR SURVIVOR TRADE COMPANY TICKER MECHANISM WHY IT MATTERS NOW Shoals SHLS Electrical balance-of-system for utility solar Closer to utility-scale electrical infrastructure than to module economics; pre-July 4 safe-harbor rush pulls demand forward SolarEdge SEDG Power optimizers, inverters Special-situation recovery, not an AI-infrastructure name; size accordingly PRESSURE POINTS Five timing and margin risks worth tracking. These are not thesis-killers — they are calendar items that create entry windows and air pockets for active managers. PRESSURE POINT READ-THROUGH July 4, 2026 48E/45Y safe-harbor cliff Utility-scale wind and solar projects that fail to begin construction by July 4, 2026 face a much narrower tax-credit path, effectively forcing developers into a compressed placed-in-service deadline by end of 2027. Pulls demand forward into Q2; creates a possible air pocket in 2H26 for SEDG and SHLS. Wind turbine Section 232 follow-through Pending tariff action or import restrictions on imported turbines could raise balance-of-system costs and squeeze installer margins; neutral-to-positive for domestic suppliers. USMCA renegotiation deadline (Jul 1, 2026) Cross-border component flows at risk. LAND.SW and GRID CN have lower direct U.S. exposure; pure-domestic names are insulated. Higher-for-longer rates Project finance economics and HASI book value most exposed to multiple compression. Partially offset by spread expansion on new originations. Midterm elections (Nov 3, 2026) Further policy whipsaw on remaining IRA provisions. Adds volatility, does not change the datacenter-driven demand floor. CREDIBILITY FIREWALL SOURCED / REPORTED DIRECTIONAL 220 GW of U.S. datacenter power demand currently in pipeline; 183 GW backed by firm commercial commitments — ~22% of 2025 peak load (Wood Mackenzie) The “clean energy” sector label is misallocating capital relative to underlying AI-infrastructure exposure Datacenters projected to drive 68% of total U.S. power demand growth through 2035 Distribution-layer suppliers and electrical equipment makers will be repriced as AI infrastructure within 12–24 months 63% of datacenter growth concentrated in PJM, ERCOT, and MISO regions The four-tier screen is a starting point for analysis, not a portfolio. Position sizing and entry are reader-specific. $1.1T projected U.S. investor-owned utility capex 2025–2029; distribution is 32% of 2025E mix (EEI) OBBBA-driven solar weakness is a setup, not a sector death sentence, for survivors with scale U.S. datacenter electrical equipment market projected to grow from ~$20B to $65B by 2030 (Wood Mackenzie) Behind-the-meter generation and electrical equipment compound fastest in a slow-interconnection environment Utility-scale solar installations −16% YoY in 2025 (Wood Mackenzie) Transformer and switchgear lead times — currently 18–24 months — will not normalize before 2028; the supplier cohort with capacity now is structurally advantaged BEAR CASE THE CLEANEST PUSHBACK The speed-to-power trade is already in the price. GNRC traded near $279 as of May 28. ENS is at $228. ORA is at $137. ETN and VRT have been crowded longs for eighteen months. These are not unknown names. The easy money may already be gone. There are four rebuttals worth taking seriously. First, timing. If hyperscaler capex grows more slowly than expected — and Microsoft, Meta, and Google have all signaled some moderation — the second-derivative trade falls first. Datacenter electrical equipment names have higher beta to hyperscaler capex revisions than the hyperscalers themselves. Second, policy. The OBBBA-driven solar slowdown is real. If Q3 and Q4 2026 deliver a sharper-than-expected utility-scale installation air pocket, SEDG and SHLS face genuine earnings risk independent of the long-term thesis. Third, rates. HASI specifically — and the entire long-dated renewable IPP space, BEP included — derates if 10-year yields hold above 4.5% through 2027. Fourth, phantom load and cost allocation. Not every datacenter interconnection request becomes energized load. The gap between 220 GW of pipeline and 183 GW of firm commitments is itself the signal: some announced load is duplicative, speculative, or contingent on power pricing, permitting, and utility cost-allocation rules. State regulators in Virginia, Ohio, and Georgia are already pushing back on who pays for substation upgrades. If tariffs, deposits, or cost-sharing rules tighten further, the demand pipeline shrinks from headline gigawatts to financeable gigawatts. The thesis survives, but the multiple expansion narrows to suppliers with confirmed orders rather than the broader category. The thesis does not require any of these to be wrong. It requires only that the distribution-layer names — ITRI, ETN, HUBB, GEV, NVT, GNRC, ENS — continue to be classified outside the AI infrastructure bucket while their cash flows say otherwise. That is a categorization arbitrage, not a growth bet. FIVE TAKEAWAYS 1. The bottleneck has migrated from generation to delivery. The binding constraint on AI buildout is no longer chips, memory, or even raw megawatts — it is the time-to-power between a hyperscaler service request and an energized site. That puts distribution capex, 32% of $1.1 trillion in utility spending through 2029, at the center of the next leg. 2. Distribution hardware is the missing picks-and-shovels screen. Switchgear, transformers, transfer switches, and electrical balance-of-system are projected to grow from a $20 billion to a $65 billion U.S. datacenter market by 2030. ETN, HUBB, GEV, NVT, and VRT sit at the center of that buildout and remain mostly bucketed as “industrials” rather than “AI infrastructure.” 3. Behind-the-meter is the speed bypass. Hyperscalers facing five-to-seven-year interconnection queues are increasingly building on-site generation. GNRC and ENS are the picks-and-shovels names that benefit regardless of who wins the cloud war. 4. Pipeline is not load. The 37-gigawatt gap between 220 GW of announced demand and 183 GW of firm commitments is the structural caveat. Not every interconnection request becomes energized load. The trade is in suppliers with confirmed orders, not in the headline-gigawatt narrative. 5. Categories are lying — again. Wall Street’s sector buckets systematically lag the cross-category convergence happening underneath them. Defense became AI infrastructure. Memory became AI infrastructure. Power semiconductors became AI infrastructure. The grid is next. Wall Street still files these stocks under “clean energy.” The cash flows file them under AI. The relabeling is where the money is. 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 Lot of news last week, all of it pretty much plays right into our investment thesis—-own the bottlenecks and balance out with the HALO (heavy asset low obsolescence) names (these include energy stocks)…… This is news, there’s a reason we just launched a photonics ETF…. Nvidia is investing billions into this emerging technology that could change the AI industry Photonics is considered to be a more efficient alternative to the current process of transferring data using electricity, which could be crucial to the AI boom. CNBC • Kai Nicol-Schwarz There are a bunch of things you need to be laser focused on. First is whatever Jensen says or does, see above, and second is whatever Trump says or does, see below. This is news, I keep telling you buy the dip on any headlines of an end to this war….. Trump ends Iran meeting without announcing ‘final determination’ on deal Trump's Truth Social statement left unclear which of his conditions are already part of a deal that negotiators are working on to pause the U.S.-Iran war. CNBC • Kevin Breuninger This could also be news. ServiceNow’s stock soars to a historic month as AI fears fade across software ServiceNow shares gained over 40% this month, leading an industrywide relief rally among enterprise software companies. MarketWatch • Christine Ji Select software names are starting to look interesting. Remember we gave you CRWD a while back as we thought AI would need more cybersecurity not less…. We gave you NOW last week, and if you are following along we just added TTD. I don’t fully trust software names, and some of them are walking dead, but I think you can sprinkle the right ones in. This is news for anyone who thinks the European aerospace and defense trade is over…. Europe Is Facing Up to a Painful Vulnerability to Drone Warfare www.bloomberg.com/news/articles/2026-05-30/europe-is-facing-up-to-a-painful-vulnerability-to-drone-warfare?srnd=homepage-americas When I launched a UFO disclosure ETF everyone thought I was crazy, the narrative is now moving very quickly towards my thesis…. I still might be crazy, but I’m also right, and early. This is mostly noise. Nothing goes up in a straight line forever, pullbacks are healthy, and if you are in the right theme, they are a buying opportunity….. Space stocks tumble after a Blue Origin rocket explodes and SpaceX’s valuation gets a reality check The red-hot space sector was feeling some heat on Friday, cooling from some of the spectacular gains seen in May. MarketWatch • William Gavin,Barbara Kollmeyer ETF News $MEMY Holdings Update: We replaced $ITRI ( ▲ 2.8% ) and added $TTD ( ▲ 7.7% ) All 5% positions. For a full list of MEMY holdings, visit: https://incomeblastetfs.com/etf/memy Distributor: Foreside Fund Services, LLC Tuttle's New Photonics ETF Targets AI Infras’ Next Bottleneck - Tuttle Capital Pure Play Photonics ETF (B The FOTO ETF targets photonics stocks poised to benefit from the AI infrastructure boom and hyperscaler data center spending. Benzinga • Chandrima Sanyal A Stock I’m Watching $FNV ( ▼ 1.09% ) I started talking about gold towards the end of last week. The time to buy the miners is when nobody wants them and this chart just had an undercut and rally at the 200 day. FNV is one of my favorite miners because it’s a royalty company. 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 27, 2026 • 11 min read Did The Memory Cartel Just Lose Its Pricing Power? Three companies have set the marginal price of memory for twenty years. Last week, a single Corsair stick with a Chinese DRAM chip inside showed where the first leak opened — not in HBM, but in the consumer DDR5 floor the entire memory trade has been resting on. May 26, 2026 • 10 min read Your Portfolio Isn't Up. The Ruler Shrank. Wall Street keeps repricing assets in a currency that no longer holds its shape — and the trades that win from here aren't the ones that benefit from inflation. They're the ones that get paid in something other than the dollar. May 27, 2026 • 1 min read The builders got famous. The suppliers got rich. The suppliers got rich. Every time. View more 5