All-In's Best Ideas Pitch Competition: 4 Investors Present Their Top Trades Live

Why does it matter? Because the decade's best trades are hiding inside America's physical limits.

Four investors pitched their top ideas live — and together they mapped a thesis most markets are still ignoring: the biggest returns ahead won't come from AI software, but from everything AI physically runs into. Between them, they traced a power grid that went from 2x China's capacity to one-third in two decades, a Vegas company disguising the most coveted casino license in Asia, a cancer drug China cannot clone by design, and a crypto token bootstrapped by hobbyists that now outranks three incumbents with 30-year head starts.

• America's power grid was already in structural shortage before a single GPU came online — AI just makes the math unsolvable. • MGM has a $48 bid floor, a hidden Osaka license the market won't price for three more years, and Dubai optionality currently valued at zero. • Actis Oncology's cancer drug moat is physical: actinium comes from US nuclear weapons waste that China's program never generated. • GeoNet is returning $8.8M per year to token holders at a $150M fully-diluted cap — the world's largest RTK network, built in four years by paying hobbyists in crypto.

AI didn't break the US power grid — two decades of neglect did. AI just guarantees it can't be fixed quietly.

The power shortage doesn't need the AI boom to sustain itself. That's Daniel Drafus's most important claim — and the one that won him 50% of the audience vote.

From 2000 to 2020, America effectively demanded no net new power. LED lighting, smart HVAC, aluminum smelters shipped overseas — the grid went dormant during what should have been an investment cycle. "We do not need AI demand to keep the power markets incredibly tight for the next 20 years," Drafus told the room. "AI demand just turbocharges. That's all it does. And it creates shortages."

The trade is Talon Energy: 2 gigawatts of nuclear, 6 gigawatts of natural gas base load, enterprise value $25B against a $45B replacement cost. Reach replacement cost — management doing nothing — and you double the equity. Sign more data-center contracts, free cash flow moves from $50/share toward $70. Build 4 of the 106 gigawatts the PJM grid says it needs over the next decade and you get north of $100/share. The stock is in the high $300s.

The Sam Zell template, applied directly: buy hard assets below replacement cost for something the world must build more of, then sell at a premium when the market wakes up.

The US went from 2x China's power capacity to one-third. That's not an infrastructure gap — it's a binary outcome.

America started this technology cycle with twice China's power generation. Today China has three times ours.

Drafus doesn't frame this as policy failure. "If you believe that artificial intelligence is going to be helpful for national security and military affairs, then you either have it or you're dead." No middling outcome, no graceful managed decline.

The refinery analogy sharpens the investment case. A data center is just a modern refinery: crude oil in, jet fuel out; electricity in, tokens and intelligence out. Jensen Huang recently said the world needs a thousand times more compute than it currently has. If that's even remotely accurate, every supply chain feeding power generation — nickel super alloys, silver for photovoltaics — is already constrained. "We know that we don't have the raw materials to meet this level of demand that's coming our way." Existing generation becomes structurally valuable regardless of which AI company eventually wins.

Barry Diller bid $48 for MGM. Aaron Cowan thinks that's underpriced by at least two countries.

Most people see MGM as a Vegas play. Cowan's pitch: it's a disguised Asian casino company with a Japan option the market hasn't priced and a Dubai free call embedded in 300,000 square feet of empty floor space.

Japan's gambling market is $40 billion — bigger than Macau at $30 billion, four times Las Vegas at $10 billion. MGM holds the only casino license in Osaka, opening in 2030. The timing insight comes from Wynn's Macau experience: the stock started moving about three years before the doors opened. "The reality is it tends to be about 3 years before it opens. Well, we're almost in that time frame." The Osaka license was approved in 2023 and barely appears in MGM's investor presentations.

The geography makes the opportunity concrete: from Shanghai, Osaka is shorter than Macau or Singapore; from Beijing, roughly Macau-equivalent distance, far closer than Singapore. First-world destination, premier Asian gaming market.

Diller's $48 bid creates the floor — he owns 26% and has 80% of his NAV committed, which Cowan reads as a financial buyer trying to capture Japan, not a strategic acquirer. Cowan wouldn't sell. The asset stack: Vegas plus existing Asia gets you to the low $60s. Japan adds ~$50/share. Dubai legalization adds $40-50 more. Path to $100-150. Rarely does a company buy back half its float in six years while an insider of this caliber concentrates this hard.

Actinium doesn't exist in China's supply chain — and that makes Actis Oncology's moat impossible to clone.

The most durable moat in the pitch competition wasn't a patent or a network effect. It was nuclear history.

Actis Oncology's radioisotope payload is actinium, manufactured from radium-233 — a waste product of US nuclear programs in the 1950s and '60s. China's program ran on enriched uranium and plutonium. Actinium was never produced. They've cloned virtually every biologic in US biotech by changing a single amino acid; they cannot manufacture this isotope. Oleg Nelman framed it simply: "unlike most of biotech, there's a real moat."

$15 billion in radiotherapy M&A in the last few years. Nelman's read: early innings. Actis (AKTS): $1B market cap, $500M enterprise value, cash beyond 3 years, first readouts from two programs in 2027. Eli Lilly backstopped the $300M IPO with $100M on an 18x oversubscribed deal. The delivery mechanism: "a swarm of micro drones small enough to navigate the bloodstream and find their target by molecular recognition, then detonate a precisely sized warhead with a blast radius of 100 microns — the diameter of a single cell." Imaging confirms tumor engagement in early trials, removing a core uncertainty from clinical development before late-stage spend begins.

GeoNet beat three incumbents with 30-year head starts in four years by paying hobbyists crypto to build its network.

Trimble, Hexagon, and Topcon have spent 20 to 30 years building RTK precision-location networks. Combined, they have roughly 12,000 base stations worldwide. GeoNet was founded in 2021 and today has roughly 24,000 nodes — twice the competition — live in 150 countries, covering 80% of global population.

Kyle Samani's engine: buy a base station for a few hundred dollars, put it on your roof, earn GEOD tokens. No AT&T-style build crews. Cost structure a third to a quarter of traditional capex deployment. Revenue sits at ~$11M annualized and is growing 3x year-over-year. Customers spend ~$60K in year one; by year two they're at ~$170K. The network routes 80% of revenue to open-market GEOD token purchases — $8.8M per year at a $150M fully-diluted cap. The last 20% funds all operations.

Customers include DJI, TomTom, and John Deere. The USDA is subsidizing farmers to adopt high-precision ag tech, most of it running on GeoNet. Friedberg raised the long-term structural risk: LEO satellite constellations could eventually replace ground-based RTK entirely. Samani's counter: a base station costs a few hundred dollars. Competing with that from orbit on cost is not realistic.

The edge in biotech is refusing to hire anyone who understands it.

"Investing in biotech companies is a horrible idea sandwiched somewhere between movies, wineries, and SPACs." Nelman opened with that — then backed it up: $13M to $2.5B AUM in 13 years, 10x for LPs at 20% annualized.

No PhDs or MDs at Eco R1, by design. Scientists fall in love with the science and lose their grip on risk-reward. The fund treats biotech like poker: hunt for margin of safety, stay cold on scientific elegance. "We consider ourselves poker players in a sector where virtually everyone else is a momentum investor betting on science." The specific structural edge: biotech investors are "so insanely short-term oriented that even though we're now eight or nine months from data, that's still way too long." Actis has traded flat since its IPO. Nelman expects accumulation to begin in the second half, ahead of Q1 2027 readouts.

GLP-1s are already the best longevity drug ever discovered — most people taking them don't know it.

Caloric restriction is the only intervention with hard data on life extension. GLP-1s mechanically replicate caloric restriction. When Gavin asked whether lifespans past 100 were achievable for people in the room, Nelman took the over without hesitation — "we already have one of the best longevity drugs out there and folks don't even realize it."

The secondary effect may matter as much as the primary biology: GLP-1s have trained a generation that healthy living through pharmaceuticals is both possible and normal. That behavioral shift is infrastructure for everything that follows in the longevity space.

Reframe the addressable market: not people with obesity, but everyone who wants to outlive their projected expiration date.

The real AI trade isn't in the models — it's in everything the models run into.

All four pitches were bets on physical constraints that software spending cannot route around. Power grids don't scale like code. Casino licenses are geographically finite. Isotope supply chains are tied to nuclear history. Location networks need hardware placed everywhere. When the next wave of capital chases AI, it will increasingly find that the bottleneck is matter, not math — and the investors who built positions in matter first will collect the toll.

---

Topics: energy infrastructure, power markets, casino gaming, MGM, radiopharmaceuticals, biotech investing, crypto tokens, RTK precision location, AI infrastructure, longevity, GLP-1, investment pitches, hedge funds, China competition, nuclear energy