All-In's Best Ideas Pitch Competition: 4 Investors Present Their Top Trades Live
Summary
- MGM Resorts won the Besties’ vote because Barry Diller’s $48 bid provides a near-term floor while Osaka and a possible Dubai casino supply asymmetric upside. Aaron Cowen valued the existing Vegas and China assets in the low $60s per share, Osaka at roughly another $50, and Dubai at $40-$50 if gambling is legalized—supporting his claim that MGM “could be easily worth over 100, could be worth 150.” With Diller owning 26% and MGM having repurchased half its float in six years, Cowen’s instruction was blunt: “I would not sell my shares to him.”
- Talen Energy won 50% of the audience vote as the liquid way to own an increasingly unavoidable U.S. power shortage. Dan Dreyfus framed its 2 GW of nuclear and 6 GW of gas generation as $45 billion of replacement cost available for a $25 billion enterprise value; even without new projects, he projected $50 per share of annual free cash flow against a stock in the high $300s. “We do not need AI demand to keep the power markets incredibly tight for the next 20 years. AI demand just turbocharges.”
- Dreyfus’s Talen upside ladder runs from a passive double to more than $100 of annual free cash flow per share if it builds 4 GW of new capacity. Additional data-center contracts or higher PJM prices could lift free cash flow from $50 to $70 per share, while a blended 15x infrastructure multiple supplies the rerating logic. The discussion identified political and regulatory intervention as a risk; Friedberg also said higher interest rates could compress the valuation multiple before the longer-dated buildout matures.
- Aktis Oncology is a well-funded but binary wager that precision radiopharmaceuticals become a reusable cancer platform. Oleg Nodelman described mini-proteins that find tumors and deliver a radioactive blast across roughly 100 microns, with imaging helping verify target engagement before efficacy is known. AKTS had a roughly $1 billion market cap, $500 million enterprise value, over three years of cash, and 2027 data catalysts; one successful program could, in his estimate, support $10 billion or $200 per share.
- GEODNET offers the most explicit token cash-flow mechanism—and the hardest institutional sizing problem—of the four pitches. Kyle Samani said its approximately 22,000 RTK nodes deliver two-centimeter positioning, versus roughly two meters for GPS, while 80% of a later-cited $11 million revenue figure funds open-market GEOD purchases. He also said the business had just crossed about a $1 million annualized run rate. At roughly $150 million fully diluted value and more than 3x growth, it looked cheap to him, but Chamath argued that even a $1 million purchase could move the market.
- The decisive portfolio distinction was not excitement but positionability, downside, and time horizon. Chamath viewed MGM and Talen as positions that could absorb tens of millions, versus smaller “lottery ticket” allocations to AKTS and GEOD; Friedberg worried that satellites could displace terrestrial RTK and that a failed Aktis readout could erase 50%-75%. As Jason Calacanis summarized the defensive pair: “People will always gamble and leave the lights on.”
Deep dive
1. The competition demanded investable conviction rather than television fame
Jason Calacanis traced the format to the Ira Sohn Foundation’s investor pitches. He recalled predicting a trillion-dollar Amazon in 2015 and being laughed at, pitching Tesla converts in 2016, then identifying AI in 2017 but choosing Box instead of Nvidia: “I could have been an Irish legend.”
The aim was to surface managers producing returns for LPs without relying on CNBC distribution. In the post-pitch discussion, Chamath invoked the Stanley Druckenmiller school: “If you don’t have any skin in the game, you don’t care.”
The four ideas deliberately spanned different payoff structures: MGM’s bid-backed equity, Talen’s power scarcity, Aktis’s clinical binary, and GEODNET’s revenue-linked token. The Besties therefore separated presentation quality from risk-reward, liquidity, duration, and feasible position size.
2. MGM’s existing assets and Diller’s bid establish the floor
Cowen, a generalist with 29 years in hedge funds, chose MGM because pitching another technology stock to this audience would be “absolutely stupid.” The poker-themed conference instead got a casino stock whose thesis had already begun moving before he presented it.
MGM traded near $37 when Cowen prepared the deck, then moved into the high $40s after Barry Diller offered $48. Diller owned 26%, his exposure represented roughly 80% of his NAV, and MGM itself had repurchased half its float in six years—an unusually concentrated alignment of buyers.
Cowen valued MGM’s Vegas and China assets in the low $60s per share, already above both the original price and Diller’s offer. Diller, he stressed, “is not a strategic buyer. He is a financial buyer, and he’s doing it to get rich.”
The bid made MGM immediately “in play” while supplying shareholders with a firm offer near the prevailing price. Cowen said the market should begin caring about Osaka as it approached the recognition window that had applied to Wynn’s Macau expansion.
3. Osaka supplies MGM’s hidden NAV and Dubai remains a free option
MGM’s Osaka resort is scheduled to open in 2030. Cowen contrasted Japan’s roughly $40 billion existing gambling market with Macau’s $30 billion and Vegas’s $10 billion, projected about $2 billion of EBITDA, and noted that MGM owns 40% plus earns a management fee.
Osaka is closer to Shanghai than Macau or Singapore and roughly as accessible from Beijing as Macau. Cowen argued that Chinese demand plus Japan’s first-world jurisdiction could justify a better valuation; he assigned the opportunity approximately $50 per MGM share.
MGM’s Dubai complex includes Aria, MGM, and Bellagio brands plus 300,000 square feet of casino space, although gambling remains illegal there. With Wynn opening two years from now in Al Marjan, roughly 45 minutes from Dubai, Cowen treated legalization as another $40-$50 per share—but conceded it was merely a “free option.”
Asked why the market ignored an Osaka approval dating to 2023, Cowen cited Wynn’s Macau precedent: investors began valuing the project roughly three years before opening. MGM is approaching that window, and shareholders could still more than double on his math if Dubai never happens.
4. Loyalty and entertainment determine whether MGM converts traffic into cash
On international customer credit, Cowen pointed to MGM’s large customer database and loyalty program. His failed investment in the Rio supplied the cautionary example: buying at $200 per square foot looked cheap, but separation from Caesars also removed the loyalty network that drove repeat play.
Cowen admitted, “I don’t know the answer,” when asked whether Diller could materially improve entertainment monetization, reiterating that his thesis was Asian rather than Vegas-led. The candor matters because entertainment became a major part of the later upside case.
Friedberg cited a Las Vegas property generating roughly $1 million of incremental EBITDA each day that the Sphere hosted a show at the Venetian hotel, because entertainment traffic drove gambling revenue. He had separately heard that Diller was working on reinventing MGM’s entertainment offering, potentially adding upside before Osaka or Dubai contributes.
5. Talen is a replacement-cost bet on unavoidable power scarcity
Dreyfus described power demand as cyclical: appliances and air conditioning drove one boom, while LEDs, smart HVAC, and offshoring power-hungry industry produced two decades of stagnation. A new technology cycle is now beginning, and AI adds shortages rather than creating the thesis.
Applying Sam Zell’s rule—buy a needed hard asset below replacement cost and sell when the market pays a premium—Dreyfus valued Talen’s 2 GW of nuclear and 6 GW of gas generation at $45 billion of replacement cost versus a $25 billion enterprise value. Because Talen carries debt, merely reaching replacement value would more than double its equity.
His strategic comparison was stark: America began the last 20 years with twice China’s generation capacity, while China now has three times America’s. If AI drives science, robotics, and national security, power availability determines whether the United States receives those benefits: “You either have it or you’re dead.”
Dreyfus likened a data center to a refinery costing roughly $50 billion per gigawatt: electricity goes in and “photons or tokens or intelligence” come out. Jensen Huang’s claim that 1,000 times more power may be needed, if remotely correct, means nuclear, gas, solar, orbital systems, and every bridge technology must participate.
6. PJM’s deficit gives Talen three distinct routes to a rerating
PJM forecasts the need for 106 GW of new power over ten years, roughly Japan’s present consumption for one U.S. region. Dreyfus called ten infrastructure years “tomorrow morning” in geological time and doubted the system could retire coal while simultaneously sourcing enough minerals, turbines, and other constrained equipment.
Microsoft’s agreement to restart Three Mile Island illustrated hyperscaler urgency: with market power near $50 per MWh, Dreyfus said Microsoft offered $100 per MWh for 20 years at a minimum price. Regulators are resisting arrangements that remove grid supply from households, creating the “crunch time” behind Talen’s scarcity value.
His base scenario requires nothing new to happen: allow the Amazon contract to roll up and Talen could generate $50 per share annually, about seven times free cash flow at a high-$300s stock. More premium contracts or higher prices could produce $70; building just 4 GW could lift the figure beyond $100.
Dreyfus used 15x as a blended terminal multiple: contracted, bond-like cash flows deserve more—potentially around 20x with 5% Treasuries—while volatile merchant generation deserves less. Batteries and peakers could protect peak hours, but later discussion highlighted government price caps, political intervention, and interest-rate sensitivity as risks to the valuation.
7. Aktis converts radiopharmaceutical precision into a platform wager
Nodelman called biotech “a horrible idea sandwiched somewhere between movies, wineries, and SPACs.” EcoR1 approaches the sector as poker rather than slots: it avoids falling in love with science, focuses on margin of safety, and tries to “monetize other kids’ science projects.”
His military history moved from surgery’s medieval siege, through chemotherapy’s indiscriminate poison and targeted therapy’s GPS-guided strike, to immunotherapy’s recruited T cells. Modern radiopharmaceuticals are “a swarm of microdrones” that recognize a tumor and detonate across roughly 100 microns—the diameter of a single cell.
Aktis had a roughly $1 billion market cap, $500 million enterprise value, and enough cash for more than three years. Its $300 million IPO was 18 times oversubscribed and included a $100 million Eli Lilly order, providing financing beyond the approaching clinical milestones.
The mini-protein platform can carry different radioactive payloads and clear the body quickly. Imaging shows whether a candidate achieved target engagement in the tumor, reducing one major development uncertainty before efficacy data arrive.
8. Aktis’s 2027 readouts offer platform upside but preserve binary risk
Aktis selected validated targets for its first programs: Nectin-4, which is critical in bladder cancer, and B7-H3 across major solid tumors, including prostate, colorectal, and lung. Initial data for both are guided for 2027, with Nectin-4 potentially arriving in Q1.
A signal in either program could revalue both the drug and the broader platform—“the holy grail in biotech.” Nodelman estimated $10 billion, or $200 per share, if one program reaches market, against a sector that has seen roughly $15 billion of recent radiopharmaceutical M&A and dealmaking.
In the Q&A, a questioner explained that Aktis’s actinium payload is manufactured from radium 233 associated with U.S. nuclear programs and said actinium is not available in China. Nodelman said China could reproduce it for its own market but that he had seen no meaningful competitive transfer into the U.S.
The stock had stayed flat since its IPO because biotech investors considered even eight or nine months too long to wait. Nodelman expected accumulation in the second half before Q1 data, while acknowledging no marketed products or recurring revenue, possible dilution, and the possibility that safety or efficacy failures damage the thesis.
9. GEODNET’s economics impressed, but liquidity and satellites decided its rank
Samani said GEODNET’s real-time kinematics locate devices within roughly two centimeters, versus two meters for GPS. Founded in 2021, it had approximately 22,000 nodes—nearly twice the combined 12,000 of Trimble, Hexagon, and Topcon—covering 150 countries, 11,000 cities, and 80% of the global population outside sanctioned markets.
Instead of centrally deploying infrastructure, GEODNET lets owners buy a station for a few hundred dollars, mount it on a roof, and earn GEOD. Customers include TomTom, DJI, robotic-mower makers, and agricultural systems; spending reportedly rises from about $60,000 in year one to $170,000 in year two.
Samani first said the business had just crossed about a $1 million annualized run rate. He later referred to $11 million in revenue, with 80%—about $8.8 million—contractually used for open-market token purchases. GEOD’s fully diluted value was about $150 million, versus the $60-$70 million circulating figures shown on some crypto sites; he disclosed owning tokens and having led a prior venture round.
The sharpest disagreement concerned obsolescence. Samani argued terrestrial stations cost one-third to one-quarter as much as proprietary systems and beat satellites on cost and energy, especially for drones; Friedberg thought dense LEO constellations would replace both RTK and GPS within a decade. The audience ranked GEODNET fourth with 5%, while Aktis took 21%, MGM 24%, and Talen 50%; the Besties reversed the top two, awarding MGM first and Talen second.