메모리 세금의 종말 — 그리고 월가가 아직 가격에 반영하지 못한 것

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 Memory Tax Is Over — And Wall Street Hasn't Priced It In The Memory Tax Is Over — And Wall Street Hasn't Priced It In Mar 30, 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 $4 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 The Short Clock Is Geopolitical. The Long Clock Is Disclosure. Both Are Ticking. Most defense funds only make sense while the war is on. UFOD is built around two separate reasons to own it; near-term defense spending and a multi-year push in Congress around Unidentified Anomalous Phenomena. We've put them together in a single portfolio. See the UFOD holdings: thetruthisoutthereufod.com Distributor: Foreside Fund Services | Investing involves risk including possible loss of principle. H.E.A.T. Google’s TurboQuant doesn’t just compress AI models. It rewrites the economics of the entire semiconductor stack — and hands a death warrant to the companies that built their thesis on AI’s insatiable memory appetite. THE SETUP Every great trade starts with a structural inefficiency the market hasn’t caught up to yet. In the AI boom of 2023–2025, one truth became gospel: AI eats memory. The bigger the model, the more high-bandwidth memory you need. Companies like Micron, SK Hynix, and Samsung poured billions into HBM3E production. Nvidia built its empire partly on the assumption that the memory-to-compute shuttle would never slow down. Wall Street rewarded everyone who fed the beast. But here’s the thing about assumptions that become gospel: they’re the most dangerous when they’re about to break. “Google just quietly dropped a bomb. TurboQuant compresses frontier AI model weights by 60–80% — without meaningful accuracy loss. The memory wall just got a wrecking ball.” THE CORE MECHANISM Here’s what’s actually happening under the hood. Right now, every AI inference workload runs on models stored in 16-bit or 32-bit floating point precision. Those models are enormous — a GPT-4 class system needs hundreds of gigabytes just to sit in memory, before it does a single computation. That data gets shuttled endlessly between HBM stacks and compute cores. The bottleneck isn’t compute — it’s bandwidth. It’s memory capacity. It’s the tax you pay every single inference cycle. TurboQuant attacks the tax at the source. Using advanced quantization techniques, it compresses weights to 4-bit and below with algorithmic compensation that preserves accuracy. The result: a 70B parameter model that used to require 140GB of HBM now fits in 35GB. Not someday. Now. In production. At Google scale. The second-order effect is even more explosive: if you need less exotic HBM, commodity LPDDR5 and GDDR7 start doing work that yesterday required $800-per-chip HBM stacks. That flips the entire memory supply chain hierarchy. The premium tier shrinks. The commodity tier expands. Margins compress at the top, volume explodes at the bottom. KEY METRICS Memory footprint reduction HBM stacks needed per server Inference cost impact Edge deployment viability ~75% | 70B model: 140GB → ~35GB -60% | Direct hit to Micron HBM demand 2–4x | Cheaper per token, more throughput New | Frontier models now fit on-device WINNERS Google / Alphabet Owns the tech, cuts its own inference costs, widens the moat vs. OpenAI on cost-per-query economics Qualcomm Edge AI suddenly viable at scale — Snapdragon chips can now run models they couldn’t touch before MediaTek Same edge story; on-device inference becomes a real product category for budget silicon Apple On-device model compression accelerates Apple Intelligence — exactly what they need for privacy-first AI Cloud hyperscalers (broadly) Lower inference COGS = better margins or price war ammunition against each other LOSERS Micron Technology HBM demand narrative was their entire bull case — that thesis just got a quiet knife between the ribs SK Hynix World’s #1 HBM supplier; massive capex already committed to HBM4 — timing could not be worse Samsung (memory division) Already losing share in HBM; now the market it was chasing is structurally contracting Nvidia (long-term overhang) Not an immediate threat, but B200’s value prop leans on memory bandwidth dominance — worth watching Pure-play HBM capex plays Any company whose thesis was “sell shovels to the memory gold rush” needs a second look KEY TAKEAWAYS 1. This is a cost deflation event, not a demand destruction event. AI adoption accelerates when inference gets cheaper. The total addressable market expands. The spoils just get redistributed away from exotic memory suppliers. 2. Edge AI just became real. When frontier-class models compress to 35GB, they fit on high-end consumer and enterprise hardware. The next platform battleground is on-device, and Qualcomm and Apple are positioned best. 3. Watch Micron’s next earnings call closely. Management will almost certainly downplay TurboQuant’s impact. That’s your tell. When execs minimize a structural threat, it’s usually because they’ve already seen the internal models and don’t like what they show. 4. The moat is in the algorithm, not the hardware. Google didn’t build a better chip. They built smarter software that makes expensive chips less necessary. This is the pattern of every great tech disruption — the incumbent’s advantage erodes from the software layer down. 5. Don’t confuse short-term and long-term effects. HBM stocks won’t crater tomorrow — existing contracts, data center build-outs, and product cycles provide a cushion. But the 2027–2028 forward estimates for HBM ASP and volume are quietly wrong, and the market will figure that out. News vs. Noise: What’s Moving Markets Today This came from the Speaker of Iran’s parliament over the weekend….. He’s probably not wrong, it also shows the Iranians understand that the stock market is a pressure point for Trump. Very few areas of the market look anywhere near decent at this point. There are of course the oil companies….. I think you ought to own oil companies, and we recommend an allocation to our wealth management clients, but I’d be real careful about chasing if you don’t own them. Gold and silver miners were the strength on Friday. Still like them here, but any long term disruption to oil prices will impact their bottom line. May eventually need to shift to the actual metals if this continues….. Even with what we wrote above, the memory companies look ok, except for MU. Of course if this all continues they will eventually get everything. I would be very careful and nimble here, memory could end up having a big problem based on what I stated above. Big question is will lower memory costs lead to more consumption? Photonics are also holding up….. While copper is not…. I still think it’s optical AND copper, not OR. So some of these names maybe candidates to snap back. Cyber security names have sold off hard on new model releases from Anthropic and AWS. I still think AI= more need for security, not less. Be watching these names for a potential bounce…. MPWR is a below the radar screen name I’m watching. Every AI server NVIDIA builds needs to be powered. MPWR has quietly captured 30-35% of that market — making it the single largest supplier of power management chips to AI data centers. NVIDIA is rolling out a new 800-volt DC power standard for its next-gen AI infrastructure. MPWR is the frontrunner to own that transition. It’s been crushed along with everything else…… Two dumb articles today….. wsj.com Battered by Stock Losses, Investors Find Little Relief in Bonds www.wsj.com/finance/investing/battered-by-stock-losses-investors-find-little-relief-in-bonds-af3f8a14?mod=hp_lead_pos1 Readers of this newsletter know bonds are not a hedge. The definition of insanity is doing the same thing over and over again and expecting different results. wsj.com Three Reasons the Stock Market Can Endure the War www.wsj.com/finance/stocks/three-reasons-the-stock-market-can-endure-the-war-23c5f966?mod=hp_lead_pos2 As long as we don’t have stagflation then stocks can endure the war just fine. Stagflation is a game changer. ETF News MEMY Holdings Update: We sold Symbotic $SYM ( ▼ 5.97% ) and Alibaba $BABA ( ▼ 0.58% ) and we added Western Digital $WDC ( ▼ 8.6% ) and Constellation Energy $CEG ( ▼ 0.96% ) All 5% positions. For a full list of MEMY holdings, visit: https://incomeblastetfs.com/etf/memy Distributor: Foreside Fund Services, LLC A Stock I’m Watching Today’s stock is Lockheed Martin $LMT ( ▼ 2.8% ) It’s bumping right up against it’s 50 day moving average here and could go either way. A move back above could be a long entry, a further break below could be a short if you are so inclined. I like the defense contractors in my secret gap theme, but that could take a bit to play out. I also think a long drawn out war helps them, but even if the war ends they have to restock. Defense contractors seem like the only win/win in this situation, but the chart has to confirm the thesis. In Case You Missed It I had the pleasure of talking to Dividend Degenerates on why I like put spreads better than covered calls for income….. 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 Mar 17, 2026 • 8 min read Why I Look at What Analysts Can't Cover Mar 23, 2026 • 14 min read 45 Days. That's All That Stands Between Your AI Portfolio and a Supply Crisis Nobody Sees Coming. Mar 20, 2026 • 13 min read Bezos Just Filed for a 51,600‐Satellite “AI Data Center” Network... Here’s the Real Trade View more 2