AI 군비 경쟁이 지구의 대기권을 막 떠났습니다

The AI Arms Race Just Left Earth’s Atmosphere 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 Arms Race Just Left Earth’s Atmosphere The AI Arms Race Just Left Earth’s Atmosphere Dec 22, 2025 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. ❝ Dividends Payable Today: BITK .10/Share MSTK .20/Share ❝ I’m hosting a webinar entitled “Why Covered Call ETFs Suck and What to Do Instead” (More Info Below) January 15 2-3pm. Sign Up Here Table of Contents --- H.E.A.T. The AI Arms Race Just Left Earth’s Atmosphere Winners, Watchlist, Losers – The Space‑Data‑Center … Likely Winners: “Picks & Shovels to Orbit” Watchlist: High Torque, High Fantasy Losers / Danger Zone: Where the Fantasy Collides W … Big Takeaways --- News vs. Noise: What’s Moving Markets Today AI Bubble Talk vs. What the Tape Is Actually Sayin … Step One: Remember What Actually Drives This Stuff The AI Bull Case (Why This Might Not Be Over) The Bear Case (Why This Can Still End Ugly) So What Was News Today vs. Just Noise? How I’d Translate This Into Positioning --- A Stock I’m Watching Why Covered Call ETFs Suck-And What To Do Instead --- H.E.A.T. The AI Arms Race Just Left Earth’s Atmosphere — # (# ) When Google starts talking about building AI data centers in orbit , you know the AI trade has officially gone from “big” to “are we all seeing this?” Project Suncatcher is exactly that: a plan to run AI workloads not in a warehouse next to a substation, but on a constellation of solar‑powered satellites. Think 80‑plus birds in sun‑synchronous orbit ~650km up, each stuffed with AI chips, working together to answer Gemini queries and beam the results back down. The pitch: No local permitting fights. No NIMBYs screaming about substations and water use. Near‑continuous solar power. The reality: you’ve just moved the AI infrastructure problem from an overloaded grid to an overloaded orbit . Low Earth orbit is already crowded with ~16,000 satellites and over a million debris fragments big enough to shred hardware. Pack 80+ Suncatcher satellites 100–200 meters apart in the single most congested orbital lane, and one bad collision or solar‑weather wobble can trigger a Kessler‑style chain reaction that trashes the whole cluster and litters space with shrapnel. Meanwhile, radiation quietly eats electronics, maintenance is a nightmare, and the unit cost of power in orbit still rhymes with “insane” once you factor in launch, replacement, and insurance. Even Microsoft’s underwater data‑center experiment ultimately came back to shore; space is another order of magnitude harder. Underneath the sci‑fi, though, is the real tell: ❝ A handful of companies have built an AI‑compute machine so power‑hungry that they are openly talking about leaving the planet to feed it. On Earth, AI data centers are already on a collision course with the grid and gas markets . TD Cowen estimates U.S. data centers could add 6+ Bcf/d of natural‑gas demand by 2030 , on top of LNG exports and industrial use, pushing long‑term gas toward the $4–5/mcf range and forcing a massive build‑out of gas plants and pipes. In orbit, the constraint is different: physics, debris, and regulation . Space is technically a global commons, but in practice it’s becoming the lightly regulated playground of Amazon/Blue Origin, SpaceX, and now Google‑plus‑whoever actually builds Suncatcher. If this idea ever scales, it drags a whole ecosystem of launch, satellite, debris‑removal, and space‑insurance players along with it. So no, you don’t buy Google because it might one day host Gemini in the sky. But you can start thinking about what a “space‑data‑center” narrative means for the space stock complex . Winners, Watchlist, Losers – The Space‑Data‑Center Angle Likely Winners: “Picks & Shovels to Orbit” These are the names that win whether Suncatcher works perfectly or flames out on re‑entry, because they get paid to launch, build, or operate the hardware. Launch & Space Infrastructure $RKLB (Rocket Lab) – Small‑launch plus spacecraft buses. Any shift toward more specialized AI constellations (compute, sensing, in‑orbit processing) is rocket fuel for launch cadence and satellite‑bus demand. Large Launch (SpaceX – private) – Not in your brokerage account, but they’re the quiet toll‑road behind almost every serious constellation. If “compute in orbit” becomes a real category, Falcon and Starship are the pipes. Satellite Manufacturing & Components $RDW (Redwire) / other space‑components names – Power systems, structures, and avionics all get leveraged if constellations start carrying heavier, hotter AI payloads that need better thermal management and power. Defense primes with space exposure – $LHX, $RTX, $NOC They already build hardened, radiation‑tolerant satellites for defense and intel customers. If commercial AI players wake up and realize “space is a war zone,” they end up talking to the same vendors DoD uses. Earth‑Observation / Space Data Platforms $PL (Planet Labs), $SPIR (Spire), $IRDM (Iridium) – They’re not “AI data centers in space” per se, but they already monetize constellations and space‑based data. As AI drives demand for more in‑orbit processing (edge inferencing on imagery, weather, RF), these platforms are natural partners or acquisition targets. Watchlist: High Torque, High Fantasy Names that could benefit if the “compute in orbit” meme really catches, but where the tech and economics are still speculative. Space‑based compute / comms hybrids $ASTS (AST SpaceMobile) and similar “cell tower in space” concepts: if they can move some AI inference closer to the edge (on‑satellite), they become part of a low‑latency, always‑on AI network. Big “if,” and funding risk is real. Space‑debris & in‑orbit servicing plays Early‑stage companies focused on debris removal, life extension, and on‑orbit servicing (some small‑caps and privately held names). If we start parking GPU farms in the busiest orbital lanes, somebody will get paid to clean up the mess and keep them running. Timing is the issue. Solar‑in‑Space / Power Beaming Concepts There are tiny public and many private efforts around orbital solar and wireless power transmission. If Suncatcher‑type ideas get traction, these graduate from science projects to “capex line item.” For now, they’re story stocks more than cash machines. Treat this whole bucket as venture‑capital lite : small position sizes, assume dilution, and be ready for binary outcomes. Losers / Danger Zone: Where the Fantasy Collides With Physics “AI in Space” pure‑story microcaps Anything that suddenly rebrands as “space‑AI” without real revenue, hardware, or backlog is the exact kind of stock that rides the first narrative spike and then vanishes when investors remember you can’t pay interest with PowerPoints. Over‑levered satellite builders without recurring revenue Launch cycles are lumpy, constellations get delayed, and now you’re adding extra complexity (radiation‑hardened AI hardware, thermal loads, higher insurance). If the balance sheet is already stretched, an AI‑in‑space pivot can be the straw that breaks it. Terrestrial data‑center names counting on unlimited grid expansion This one is more subtle. If space‑data‑center hype really picks up, some investors will start asking uncomfortable questions about the terrestrial names that have penciled in infinite power and water. You don’t short them because of Suncatcher, but you do mark them “prove‑it” on grid access and permits. Big Takeaways Space data centers are a symptom, not a solution. They’re what you get when a handful of AI platforms chase “infinite scale” into a world with very finite land, water, copper, and electrons. Whether Suncatcher works or not, it tells you how desperate the arms race has become. The real trade is the infrastructure around the madness. Launch, hardened satellites, space components, and debris mitigation are the toll roads. AI may or may not live in orbit; rockets and space‑hardware vendors will be busy either way. Don’t confuse sci‑fi with cash flow. Most of the value in the “AI goes to space” story will accrue to the same sort of businesses that win every capex arms race: disciplined picks‑and‑shovels with recurring revenue and sane balance sheets, not the first microcap that mentions GPUs and orbit in the same press release. You don’t have to believe we’ll all be chatting with models running above the Kármán line. You just have to recognize that AI is now stressing both the grid and the sky —and position yourself in the parts of the stack that get paid to build, launch, and maintain whatever crazy infrastructure comes next. --- News vs. Noise: What’s Moving Markets Today AI Bubble Talk vs. What the Tape Is Actually Saying After looking absolutely wrecked for a few days, the AI complex just did the thing bears hate most: AI‑heavy leaders snapped back hard. Bitcoin bounced with them. $QQQ undercut its 50‑day moving average and then ripped back above it – a classic “undercut & rally” pattern. In other words: the part of the market everyone just finished writing eulogies for is suddenly acting like it still wants higher. So is this just another dead‑cat bounce in an AI bubble… or a resumption of a still‑alive bull leg? You can’t answer that just staring at today’s candles – you have to zoom out to liquidity and the AI build‑out itself. Step One: Remember What Actually Drives This Stuff The uncomfortable truth: big tech, AI, and Bitcoin trade off global liquidity first, stories second. The dot‑com bubble didn’t burst because people suddenly decided Pets.com was dumb; it burst when the Fed pulled back the Y2K liquidity hose and tightened into overvaluation. Since then, the NASDAQ and global QE have moved like a braid – when central‑bank balance sheets grow, growth/tech runs; when they shrink, it bleeds. 2025 gave us a preview of that again: temporary liquidity scares (shutdown drama, debt‑refi delays, the October BTC market‑maker mess, MSCI/MSTR noise) opened up those “alligator jaws” between Bitcoin and the NASDAQ. The point: as long as global liquidity is expanding or at least not aggressively contracting , it’s hard to kill an AI/crypto bull outright – you usually just shake out the over‑levered tourists. Today’s AI snapback and QQQ undercut‑and‑rally are exactly what you’d expect if the bigger liquidity regime is still friendly. The AI Bull Case (Why This Might Not Be Over) The bullish argument isn’t “AI is cool.” It’s six concrete points: 1. Capabilities are compounding stupidly fast. Empirical work on “agentic” systems shows the length/complexity of tasks AI agents can complete has been doubling roughly every 6–8 months since 2019 – a new Moore’s law for usefulness , not just FLOPs. In some dimensions the doubling time is closer to 3–6 months. 2. Usage is following – this isn’t just slideware. Google has reported 14× growth in AI token usage in ~8 months , implying real‑world demand is exploding, not fading. Multiple buy‑side deep‑dives estimate 15–25% of S&P 500 market cap already reflects concrete AI earnings expectations, not 30‑year fairy‑tale DCFs. 3. Demand is still chasing capacity, not the other way around. Street work pegs AI + cloud capex up ~60% YoY into 2025 , with 2026 guidance still being revised up , because hyperscalers and GPU clouds say they’re still capacity‑constrained, not sitting on empty racks. Q3 ’25 big‑tech capex was up ~75% YoY – this doesn’t look like the late‑stage fiber glut where capacity sat dark for years. So while Twitter screams “overbuild,” the people actually buying capacity are still complaining there isn’t enough. 4. Global liquidity still has their back. Despite the “QT” headlines, central banks have quietly pivoted back toward QE‑lite : reserve‑management T‑bill buying, facilities, swap lines – i.e., more base money sloshing around. With public debt at nosebleed levels, the IMF/BIS crowd openly admits policymakers are drifting toward financial repression – lower real rates, structurally friendlier liquidity, not a Volcker rerun. The dot‑com top in 2000 came after a clear tightening regime. We’re not there. 5. Valuations are rich, not 1999‑insane. $NVDA at ~40–45× trailing and low‑20s forward P/E with triple‑digit growth is frothy, but nowhere near Cisco at 200–400× earnings in 2000. $MSFT, $GOOGL, $META, $AMZN all throw off real free cash flow that can fund AI capex internally. You’re not betting on eyeballs and “new economy” nonsense – you’re betting on real businesses stretching their moats. 6. Overbuilding can pull forward monetization. The 1990s “wasted” fiber is what made broadband/cloud profitable in the 2010s. Similarly, “too many GPUs” today means: cheaper, more abundant compute later, more experiments that wouldn’t otherwise get funded, and more pressure on management to force AI into every profit center to pay for the racks. That’s exactly how real S‑curves work: you overspend before you know the full ROI, then scramble to monetize the sunk cost. The Bear Case (Why This Can Still End Ugly) The bears aren’t hallucinating either. Their arguments: 1. Concentration is extreme. Top‑10 S&P names sit near 40% of the index , a 145‑year high. One company flirts with 16% of U.S. GDP in market cap. Every prior episode of this kind of dominance (railroads, telcos, dot‑coms) eventually mean‑reverted with big drawdowns. 2. Capex is enormous vs. current returns. Big tech is spending hundreds of billions per year on AI infra. Some analyses estimate AI capex/revenue ratios north of 10–15:1 in parts of the stack. GPU rental prices have already fallen double‑digits YoY – an early sign that supply is finally catching up. In the 1990s, telcos spent ~$500B on fiber, left 85–95% of it dark, and vaporized ~$7T in equity when the math caught up. AI doesn’t have to rhyme perfectly… but the scale is at least in the same league. 3. The macro engine is narrow. Strip out data‑center/AI‑linked investment, and some estimates put H1 ’25 U.S. real GDP close to flat . That makes the cycle fragile: if AI capex slows, GDP and earnings both feel it. 4. Project failure rates are ugly. Survey work suggests 40–80% of enterprise AI projects are being abandoned or failing to reach production. That doesn’t kill the long‑term story, but it does argue the ramp to clean, recurring AI revenue is bumpier and slower than bulls want to admit. 5. All big tech bubbles look invincible right before they don’t. Railways, 1920s radio, 1990s telco, dot‑com – they all had a very convincing “this time is different” productivity story… right up until a 50–80% drawdown reset expectations. Under that lens, the risk isn’t that AI is fake – it’s that the market has pulled forward a decade of earnings into a few years of price and will need a long, sideways, volatile “lost decade” to grow into the multiple once liquidity stops helping. So What Was News Today vs. Just Noise ? Noise: The last few days of AI puke. Today’s AI face‑rip. QQQ’s cute undercut‑and‑rally at the 50‑day. All of that is positioning + sentiment around one simple question: “Do we still have liquidity and a story?” For now, the tape says yes . News: The structural link between global liquidity and high‑beta/AI is still intact. The bull case (capabilities, demand, liquidity, profits) can reasonably support another 1–2 years of build‑out. The bear case (concentration, capex scale, narrow macro, high failure rates) is getting louder – which tells you the eventual come‑to‑Jesus moment, when it comes, could be violent. How I’d Translate This Into Positioning Not advice, just how I’m thinking about it: Treat this AI snapback as confirmation the bull isn’t dead yet. As long as $QQQ holds that 50‑day undercut‑and‑rally and liquidity doesn’t reverse, the path of least resistance in AI/semis is still up or sideways, not straight down. Barbell the exposure: Core in profitable, self‑funded AI winners ($NVDA, $MSFT, $GOOGL, $META‑type names, plus “physical AI” plays in power, gas, grid). Only small, speculative sleeves in levered GPU clouds, story stocks, and pre‑revenue AI flyers. Respect the bubble math and the liquidity reality. You can believe we’re still in the build‑out phase and that the longer capex runs ahead of earnings, the worse the eventual hangover will be. Both can be true. For now, the market just told you loud and clear: the AI story isn’t over. It’s just moving from “straight up” to “who actually earns their cost of capital in a world where money isn’t free.” That’s where the real stock‑picking starts. --- A Stock I’m Watching Today’s stock is Generac (GNRC)…. Generac is one of the cleaner “picks-and-shovels” plays on grid fragility + AI-driven load growth because it sells the unsexy but mission-critical layer: backup generation, storage, and power management across residential, commercial/industrial, and data & telecom end-markets. What’s changed is that management is explicitly calling out the potential acceleration in hyperscale + edge data centers requiring significant backup power capacity , and they’ve already begun initial shipments of large‑megawatt generators to data-center customers —i.e., this isn’t just a narrative, it’s starting to show up in real deployments. The near-term bear case remains that parts of the core residential business are “storm/outage beta” (a lower-outage environment can pressure home standby/portable demand), which is exactly why the stock can look messy quarter-to-quarter. But the bull case for 2026 is that as data-center redundancy, microgrids, and C&I solutions become a bigger slice of mix, GNRC can de-weather its earnings profile and earn a higher-quality multiple— if large-megawatt/data-center ramps translate into sustained bookings and margin resilience. How Else I Can Help You Beat Wall Street at Its Own Game Inside --- H.E.A.T. is our monthly webinar series, sign up for this month’s webinar below…. Why Covered Call ETFs Suck-And What To Do Instead Thursday January 15, 2-3PM EST Covered call ETFs are everywhere — and everyone thinks they’ve found a “safe” way to collect yield in a sideways market. The truth? Most of them suck. They cap your upside, mislead investors with “yield” that’s really your own money coming back , and often trail just owning the stock by a mile. Join me for a brutally honest breakdown of how these funds actually work — and what you should be doing instead. What You’ll Learn: 🔥 Why “high yield” covered call ETFs are often just returning your own capital 📉 How most call-writing strategies quietly destroy compounding 🚫 Why owning covered calls in bull markets is like running a marathon in a weighted vest 💡 The simple structure that can fix these problems — and where the real daily income opportunities are hiding Sign Up Here All rights reserved. Keep Reading No posts found View more caret-right 0