All-In's Best Ideas Pitch Competition: 4 Investors Present Their Top Trades Live
In a Nutshell
Four investors pitched MGM Resorts, Talen Energy, Actinium Pharmaceuticals, and GeoNet (GEOD token) as their best ideas. MGM won on the strength of Barry Diller's 26% stake, a discounted valuation versus Japan and Dubai optionality, and a path to triple the stock. Talen Energy ranked second as a nuclear/gas power play trading far below replacement cost that could see 2-3x upside from AI-driven electricity demand and data-center contracts.
These notes were generated by AI and may contain inaccuracies.
The event is a recreation of the Ira Sohn Investment Conference, established by the family of Ira Sohn after he passed away from cancer. The conference began at Lincoln Center in New York. In 2015, a young venture investor presented Amazon as a future trillion-dollar company and was laughed out of the room. David Einhorn stated he knew trillion-dollar companies and believed Amazon was not one. The same presenter returned in 2016 with Tesla converts and in 2017 pitched AI by selecting Box instead of Nvidia.
The goal of recreating the event is to give exposure to great managers who generate strong returns for LPs but receive limited public distribution of their ideas. Attendees can hear these managers directly without seeing them only on CNBC.
One contestant described a vision where anyone can trade any asset anywhere in the world 24/7 using only an internet connection and a phone. Another stated they are building a new financial system from the ground up. A third noted that people will want to own equities and that innovation in companies and applications will create opportunities for long-short managers to generate significant returns.
Eco R1 Capital, based in San Francisco, focuses on unfollowed, unloved, and misunderstood biotech companies. There is optimism around a structural and permanent perception shift where both political parties will support nuclear energy. One speaker expressed being "massively optimistic" and described the current environment as "maximum risk on."
Aaron runs a $4 billion generalist firm in New York called Serreta Capital. Previously, he ran the equities business for George Soros and served as CIO for Steve Cohen. He has 29 years of hedge fund experience. Given the tech-heavy audience, he chose not to pitch a tech stock and instead selected MGM due to the poker theme of the conference.
MGM owns 13 properties in Las Vegas and, along with Caesars, is one of the two largest casino asset owners in Vegas. Caesars was recently acquired, signaling potential improvement in the Vegas market, though the pitch is not centered on Vegas recovery.
Barry Diller has been aggressively acquiring MGM stock and now owns 26% of the company. When the presentation was prepared two weeks prior, the stock traded around $37; it has since risen to the high $40s after Diller bid $48. Aaron stated he would not sell shares to Diller at that price.
MGM has also been buying back its own stock, repurchasing half its float over six years—a rare occurrence. Diller, originally known for ABC and assets like Expedia, has shifted focus to casinos, with MGM now representing 80% of his NAV.
MGM holds a license to open a casino in Osaka, Japan. Japan conducted a national referendum on casinos; Osaka was the only prefecture that approved one. The project is scheduled to open in 2030. Company presentations mention it briefly but do not emphasize it.
Japan's existing gambling market (pachinko and horse racing) totals approximately $40 billion annually. Macau's market is $30 billion and Vegas is $10 billion. MGM is estimated to generate about $2 billion of EBITDA from the Osaka property, in which it owns a 40% stake plus a management fee.
Osaka's location offers advantages: it is closer to Shanghai than Macau or Singapore, and roughly the same distance from Beijing as Macau while being much closer than Singapore. As a first-world country, Japan avoids the lower multiples associated with Macau.
MGM is developing a large complex in Dubai that includes an Aria, MGM, and Bellagio-branded property with 300,000 square feet of empty space intentionally left for potential casino use. Gambling is currently illegal in Dubai, but the space was built in anticipation of future legalization.
Wynn is scheduled to open a casino in Arjan, 45 minutes from Dubai, in approximately two years. MGM's central location would be more convenient for visitors. If Dubai legalizes gambling, especially after Wynn's opening or amid efforts to reestablish itself post-conflict, the optionality could add substantial value.
Combining Vegas assets (estimated $60 per share value), the Japan opportunity ($50 per share), and potential Dubai upside ($40-50 per share) suggests the stock could triple. Barry Diller's $48 bid is viewed as a financial buyer's offer rather than a strategic one, indicating the company is now "in play." Aaron believes the stock could reach over $100, potentially $150, with strong risk-reward given Diller's 26% stake at roughly the current price.
Regarding monetization of non-gaming entertainment assets, Aaron noted uncertainty but stated that any improvement would help; the core thesis centers on the Asian casino opportunity rather than Vegas. MGM possesses a large customer database from its Vegas properties that can be leveraged for international expansion, along with a loyalty program. A prior investment in the Rio highlighted the importance of loyalty programs, as separation from Caesars caused loss of that asset.
On the Dubai question comparing Caesars' experience, Aaron described MGM's 300,000 sq ft empty space as a free option. If gaming is not legalized, the base case of more than doubling still holds; legalization would create a triple.
Regarding Osaka's 2023 approval and why the market ignored it, Aaron referenced historical precedent from Wynn's Macau opening, where the market began pricing in the opportunity approximately three years before opening. The current timing aligns with that window. MGM is running the Dubai project and left the casino space as optionality.
Daniel presented Talen Energy. A typical power cycle begins with demand growing in line with GDP. Technological breakthroughs cause demand spikes, followed by normalization and an efficiency phase before the cycle repeats. Historical examples include appliances and air conditioning, followed by efficiency gains in the 2000s (LED lighting, smart HVAC, tinted windows) and offshoring of power-intensive industries like aluminum smelters to China, resulting in two decades of flat power demand.
The current cycle is driven by technological breakthroughs that are highly power-intensive. AI demand is not required to keep power markets tight for the next 20 years but will "turbocharge" demand and create shortages.
Daniel referenced advice from Sam Zell: buy hard assets below replacement cost when future capacity will be needed, then sell at a premium when the market recognizes the value. This approach was applied successfully to Equity Office Properties.
Talen Energy operates 2 GW of nuclear power and 6 GW of natural gas baseload power. At a $25 billion enterprise value, the stock trades at a significant discount to $45 billion replacement cost. Equity value to reach replacement cost implies more than a double from current levels. The strategy is to hold until the market prices the assets at a premium to replacement value.
China increased its power generation capacity from half of America's 20 years ago to three times America's current capacity. If AI drives scientific breakthroughs, robotics, or national security applications, the U.S. must build out power infrastructure or fall behind. Data centers function analogously to refineries: electricity is the input, and the output is intelligence/tokens.
Jensen Huang stated that AI may require 1,000 times more power than currently available, necessitating every power source including hundreds of GW of nuclear, solar, and orbital solutions. Supply chain constraints on critical minerals, nickel super alloys, and silver will create shortages and delays across all solutions.
The PJM grid operator forecasts a need for 106 GW of new power capacity over the next 10 years in Pennsylvania, New Jersey, and Maryland alone—equivalent to Japan's current total consumption. Thermal coal retirements are unlikely given the inability to build 100 GW in that timeframe. Supply chain participants are in "panic mode" due to insufficient raw materials.
Hyperscalers are aggressively seeking long-term PPAs at fixed prices. Microsoft incentivized Constellation Energy to restart the Three Mile Island nuclear reactor by offering $100/MWh for 20 years when prevailing prices were $50/MWh. Regulators are increasingly concerned about diverting power from residential customers to data centers.
At current prices in the high $30s, if Talen does nothing beyond running existing operations and letting its Amazon data center contract roll off, it is projected to generate $50 per share of annual free cash flow—approximately 7x FCF versus typical 15x multiples for U.S. infrastructure assets, implying a double.
If the company signs additional premium data center contracts or power prices rise, FCF could reach $70 per share, supporting a $50+ valuation at a 15x multiple. If Talen builds new capacity (e.g., 4 GW of the needed 100 GW), FCF could exceed $100 per share.
Power prices in PJM remain too low to incentivize new capacity, which is considered unsustainable. Regulators are directing power producers and data centers to negotiate bilateral deals for new builds.
On regulatory risk and political concerns around rising electricity prices, Daniel argued that human ingenuity will prevail. Data centers can deploy batteries and peakers to manage peak demand, serving as a bridge solution until more baseload capacity is built. Consumer bills may rise modestly but should remain manageable.
The investment thesis does not require behind-the-meter collocation; scarcity of clean firm baseload power supports the case regardless of whether power flows front-of-meter or behind-the-meter. Three scenarios were outlined: $50/share FCF (base, double the stock), $70/share with additional contracts, and over $100/share with new builds.
Competition from fuel cells, gas turbines, aeroderivative turbines, and orbital compute is viewed as necessary. These are expensive bridge solutions (high LCOE), but hyperscalers will pay premiums to avoid idle data centers. All-of-the-above solutions are required.
The terminal multiple question for a shift from merchant IPP to contracted infrastructure was posed but not answered in the provided transcript segment.
The question addressed what the right terminal multiple would be for Talon if the business mix shifts from merchant IP to contracted infrastructure, and what percentage of EBITDA needs to be contracted before the market should rerate it. The 15 times multiple was used as a blended multiple between contracted and uncontracted business. Contracted infrastructure receives a premium multiple because it represents bond-like cash flow streams that trade at a small spread to treasuries. With treasuries at 5%, these should trade at 20 times plus some growth adjustment. Merchant or uncontracted stuff with spot market exposure is more volatile and less visible, warranting a lower multiple. The more contracts a business has, the higher the multiple; less contracted business receives lower multiples. Fifteen times serves as a good rule of thumb for this blended approach.
Oleg Nelman is the founder and managing director of Eco1 Capital, a San Francisco-based value-oriented biotech fund started about 13 years ago. The fund was written to be accessible even to non-science investors. Biotech investing is generally viewed as a poor idea, comparable to movies, wineries, and SPACs, with the sector often feeling more like a casino than a financial market. Most investors in the sector function as momentum investors betting on science, whereas Eco1 positions itself as focusing on margin of safety and risk-reward rather than falling in love with the science. The fund monetizes other companies' science projects. Nelman has 25 years of biotech investing experience, including an 11-year stint at another fund before launching Eco1 in 2013 with $13 million in assets. Since inception, the fund has 10xed for investors with 20% annualized returns and now manages approximately $2.5 billion. The fund recently reopened for the first time in four years after maintaining long-term partners, many of whom are biotech entrepreneurs who invested from day one.
Military terminology has been applied to cancer treatments since President Nixon signed the National Cancer Act in the early 1970s. The warfare analogy fits because both cancer treatment and military operations aim to find the enemy, determine the best weapon to eliminate them, and minimize unwanted casualties. Early surgical cancer treatment and radiation resembled a medieval siege, leveling entire areas and hoping the enemy was destroyed in the rubble. Chemotherapy originated from an accidental World War I observation that mustard gas killed rapidly dividing tissue. Since tumor cells divide quickly, doctors would flood patients' bodies with chemo hoping it killed cancer faster than healthy cells. However, hair, skin, gut, and marrow cells also divide rapidly, and the treatment lacks discrimination. First-generation targeted therapies functioned like GPS-guided munitions, identifying the enemy's command and control center rather than carpet bombing all dividing cells. The limitation is that cancer develops resistant mutations allowing it to adapt and hide.
Immunotherapy, introduced to patients about a decade ago, recruits local allies known as T-cells to fight rather than deploying external troops. This approach works spectacularly when successful but depends heavily on the tumor microenvironment. Modern radiopharmaceuticals function like swarms of micro drones that navigate the bloodstream, locate targets through molecular recognition, and detonate precisely sized warheads with a 100-micron blast radius equivalent to a single cell's diameter. This delivers autonomous precision targeting with bunker-buster force and minimal collateral damage. Physicians can verify target engagement in early clinical trials through imaging, significantly derisking clinical development by confirming the drug reaches the tumor.
Actinium Pharmaceuticals (ticker: AKTS) operates on the front lines of cancer treatment with radiopharmaceuticals. The company has a $1 billion market cap and $500 million enterprise value with cash reserves lasting over three years past critical milestones scheduled for next year. Founded five years ago, the company went public via a $300 million IPO that was 18 times oversubscribed and backstopped by a $100 million order from Eli Lilly. The platform can carry any radioactive payload, targets various markers, and clears the body with minimal side effects. Initial programs target known valid targets including Nectin-4, critical in bladder cancer, and B7H3, expressed on every major solid tumor including prostate, colorectal, and lung cancers. Clinical trials began last year with initial data expected in 2027, with Nectin-4 data potentially arriving as early as Q1. Success in either program would generate value for both the specific program and the entire mini-protein platform. Major pharmaceutical companies including Bristol, Novartis, Bayer, and Eli Lilly have invested in radiotherapy capabilities and actively seek assets, with $15 billion in M&A and dealmaking in radiotherapy over recent years. The modality is difficult to replicate since generics do not traffic in radio pharma, and radioisotopes are off limits to China, creating a genuine moat.
Actinium's radioisotope payload uses actinium manufactured from radium-233, a waste product from U.S. nuclear programs in the 1950s and 1960s. China developed a different nuclear program using enriched uranium and plutonium, making actinium unavailable there. The Amgen Supreme Court case established that patents are composition of matter patents, allowing competitors to modify one amino acid to circumvent patents, which China has done with biologics. However, radioisotopes require local U.S. manufacturing supply chains, preventing Chinese competition from entering the U.S. market. The stock has traded flat since the IPO due to biotech investors' short-term orientation, with data still eight to nine months away. Accumulation is expected in the second half ahead of Q1 data.
The human genome sequencing has not delivered the expected explosion in therapies and personalized medicines over the past 25 years. AI is positioned to unlock revolutionary therapies by addressing the genome's complexity beyond human or traditional software capabilities. The odds favor average human lifespan in developed countries extending well past 100 to 125 or 150 years within current lifetimes, driven partly by GLP-1 obesity drugs. Caloric restriction is one of the few interventions with actual data demonstrating life extension, and these drugs effectively replicate that effect. The acceptance of injectable pharmaceuticals for healthy living represents just the beginning of pharmaceutical-driven longevity advances.
Using space and microgravity as therapeutic variables for cancer treatment, given cancer cells' different behavior in low-gravity environments, is an interesting concept but not applicable to Actinium's approach. Actinium has a small internal AI project leveraging proprietary datasets with PhD data scientists exploring applications for mini-proteins. CAR-T therapies have struggled in solid tumors. Revmed's new targeted therapy for pancreatic cancer represents recent progress, though peptides have not shown substantial advancement yet. Mini-proteins are sized appropriately to potentially deliver payloads inside tumors. Right-handed proteins (Droins) are being examined for their ability to penetrate solid tumors.
Kyle Samani founded Multicoin Capital eight and a half years ago and stepped down recently. He led all three investment rounds in Solana prior to the network's 2020 launch. David Sacks seeded Multicoin early. GeoNet builds the rails for AI through precise geolocation technology. The company uses RTK (real-time kinematics) to achieve approximately 2 centimeter location accuracy compared to GPS's roughly 2 meter precision, representing about 100x improvement. GeoNet operates the world's largest and fastest-growing RTK network with approximately 22,000 nodes deployed across 150 countries, more than 11,000 cities, and covering roughly 80% of global population excluding sanctioned countries. The three traditional competitors (Trimble, Hexagon, and Topcon) collectively maintain about 12,000 base stations built over 20-30 years. GeoNet, founded in 2021 and beginning network construction in 2022, has grown to twice the size of those three combined.
GeoNet's rapid scaling stems from its decentralized crypto model. Base stations are deployed by individuals, hobbyists, professionals, and small business owners rather than centralized telecom operators. Users purchase base stations for a few hundred dollars, install them on rooftops, and earn Geo tokens for broadcasting radio waves. The network effect creates natural monopoly characteristics similar to telecom infrastructure. The company maintains the lowest cost structure due to this decentralized approach.
The USDA subsidizes farmers and ranchers adopting high-precision agriculture technologies powered by GeoNet. Applications include robotic mules for agriculture, such as Burrow's grape-transporting vehicles, and John Deere's Global Unmanned Spraying Systems (GUS) for pesticide application. Napa Valley wineries use GUS vehicles. TomTom, a supplier to nearly every autonomous vehicle program globally, uses GeoNet data to update maps with greater precision. Consumer robotics applications include robotic lawnmowers from Yarbo, Sunseeker, and others, with an estimated 1 million units projected to sell this year. DJI, the world's largest drone manufacturer, uses GeoNet in many models. American drone manufacturers emerging as DJI exits the U.S. market are expected to adopt the network.
GeoNet crossed $1 million in annualized run rate recently with growth exceeding 3x year-over-year, projected to more than triple over the next 12 months. Of the approximately $11 million in revenue, 80% funds open market purchases of Geo tokens visible on the Solana blockchain, equating to $8.8 million annually returned to token holders. The remaining 20% covers R&D and business development scaling. Average customer spending grows from approximately $60,000 per year in the first year to about $170,000 per year by the second year, representing roughly 3x growth. The customer base expanded 5x last year. The token trades 24/7 on the Solana blockchain under ticker GEOD with approximately $150 million fully diluted market cap, though float-adjusted market cap shows $60-70 million due to tokens not yet circulating.
GeoNet operates as a U.S.-based corporation with a team of four in San Francisco. The CEO's name is Mike.
Kyle Sacks presented Geodet, a decentralized physical infrastructure network for precise positioning and RTK services. The CEO is Mike Horton. Kyle owns a significant amount of the token but holds no equity in the corporate entity. The company has contractual relationships with John Deere, DJI, and other major players. The corporate entity directs 80% of its revenues to buy tokens on the open market, creating direct value accrual to token holders.
The company previously raised venture capital, with Multicoin Capital leading a round. Kyle led Multicoin's investment in Helium six or seven years ago and remains a long-term believer, noting major news released that morning.
When asked why Geodet differs from failed deepin projects where users don't value token rewards, Kyle explained that Geodet is returning capital to shareholders with $8.8 million in returns. The token trades at a $150 million valuation and is projected to grow 3x during the year. The asset is undervalued due to the current crypto bear market.
The token represents a securitized interest in cash flows from customers, functioning as a revenue share token where 80% of revenues buy tokens on the open market. The remaining 20% funds operations including engineers and salespeople.
Regarding securities laws and the Clarity Act, Kyle noted that the tokens accrue value through the 80% buyback mechanism. He is optimistic about the Clarity Act's passage and has been involved in its development, though he is not a lawyer and cannot confirm compliance.
The discussion addressed why John Deere and others would adopt this system rather than their existing RTK infrastructure. The key advantage is capex versus opex: Geodet networks operate at one-third to one-quarter the cost of purchasing and maintaining proprietary base stations. The service is available globally, reducing sales cycle time for equipment manufacturers.
Questions arose about microsatellites as alternatives to GPS providing sub-centimeter resolution, potentially replacing both GPS and RTK through mesh networks from SpaceX launches. Kyle argued satellites cannot compete on cost, noting Geodet base stations cost only a few hundred dollars. Ground-based solutions consume far less energy than space-based alternatives, making them preferable for battery-sensitive applications like drones and tractors.
GPS serves different use cases with different SLAs. A hybrid situation with GEO, LEO, and ground stations is possible, though the goal is for LEO satellites alone to eventually replace ground-based infrastructure.
When asked about other compute-related tokens in distributed training, Kyle expressed skepticism that distributed training will succeed. Distributed inference is possible but has not performed at an A+ level despite investments made years ago. Kyle's familiarity with Geodet stems from meeting the founder years ago and following the project closely.
Ground-based base stations consume significantly less energy than satellite solutions. For battery-sensitive applications, ground infrastructure remains the preferred approach.
Timothy praised the pitch format and referenced the Stan Druckenmiller investment philosophy of requiring skin in the game. He noted varying asymmetric alpha, downside risk profiles, and liquidity considerations across the four ideas.
Kyle's Geodet idea presents liquidity constraints where even $1 million would move the market, limiting positions to $10,000-$30,000 initially. Talent could absorb tens of millions without market impact. The biotech opportunity features discontinuous illiquidity with zero downside risk but 10x upside potential. MGM and Talent offer the best capacity for large position sizing.
Gavin suggested the format should expand next year and proposed creating an All-In basket or ETF for trading the ideas. He agreed to pitch silicon and memory supercycle themes next year.
Gavin distinguished between entertaining pitches and attractive risk-reward profiles. He is neither a healthcare nor crypto investor. From a pure risk-reward perspective, MGM ranked highest due to capped downside from the Barry Diller bid and future value from Japan and Dubai expansion.
Talent faces regulatory risks from potential US government intervention, price caps, nationalization scenarios, or administration changes that could disrupt terrestrial compute supply-demand dynamics.
Actis was viewed as compelling for attempting to create a biotech platform with broadly applicable mechanisms that could generate $100 billion-plus outcomes. Gavin does not invest in crypto but found the Geodet discussion fascinating.
When asked what would prompt crypto investment, Gavin compared it to snowboarding: he has invested 25 years developing equity and public securities expertise and sees crypto as requiring new learning with potential to disrupt existing market dynamics.
David estimated MGM offers approximately 3x upside over two years with downside protection from the Barry Diller bid. Additional upside exists from upgrading monetization at Vegas properties, with examples of $1 million daily incremental EBITDA from Sphere shows at the Venetian. Barry Diller is reportedly focused on reinventing entertainment offerings.
Talent presents 3-5x upside potential but requires an eight-year timeline. The investment is more interest-rate sensitive than MGM because power purchase agreements drive significant revenue, creating margin compression risk if rates rise. The 15x outlook could be affected by higher discount rates.
Actis carries China risk where innovations get replicated and 50-75% downside if readouts fail or new therapeutic modalities emerge. However, positive readouts could attract Lilly acquisition interest within six months. David noted investing in a competing D-protein conjugate showing strong solid tumor efficacy.
David believes space-based solutions will replace RTK and GPS within the decade, making Geodet less attractive long-term.
David placed Actis and Geodet in a lottery ticket bucket with potential for extreme returns but high probability of zero outcomes. MGM and Talent offer downside protection with more predictable demand drivers.
David committed $200,000 to each of MGM and Talent in real time. One participant bought half of David's action size. Rankings concluded with MGM first, Talent second, Geodet third, and Actis fourth.
Based on 150 audience votes, the ranking was: fourth place Kyle Sacks with 5%, third place Oleg with 21%, second place Aaron Cowan for MGM with 24%, and first place Dan Drafus with 50%.
Judge rankings differed from audience results: fourth place Kyle Sacks, third place Oleg, second place Dan Drafus, and first place Aaron Cowan, representing an upset over the audience vote. MGM was declared the overall winner.
Keep All-In Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
