Deep Dive
Tesla-SpaceX Merger Math: Share Conversion and Valuation
James walks through the mechanics of a potential SpaceX acquisition of Tesla, starting with baseline 2032 price targets for Tesla standalone: bear case $2,700-3,100, expected $5,200, bull $8,000-8,200. With 320 shares, that translates to $864K-$992K in bear, $1.66M expected, $2.62M bull. But the magic happens when you layer in the merger premium. At a 50% deal premium (which James expects to be 40-60%), your 320 Tesla shares convert into 1,183 SpaceX shares at a 3.7:1 ratio, making you own 0.00000435% of the combined entity. This percentage-based thinking matters more than ticker symbols because both will likely merge under the X ticker. The combined sandbag valuation (very conservative) lands at 18 trillion, but expected case is 28-35 trillion by 2032 when you factor in Cybercab, Optimus, space-based data centers, Elon web services compute, turbine manufacturing for energy, and the Colossus GPU clusters. Your 320-share bag at expected case becomes $1.6-2.2M—far above the standalone Tesla bull case. Critically, James emphasizes that lower premiums under 25% get rejected by shareholders, all-cash deals won't happen because SpaceX needs cash for capex, and selling early kills compounding when each share's future value is $3,000-8,000.
Anthropic IPO Won't Trigger SpaceX Selloff
A Patreon member asks if Anthropic's likely $2 trillion IPO (versus SpaceX's current $1.75 trillion valuation) will cause investors to chase hype and dump SpaceX at 10-20% discount. James rejects the premise. Anthropic will need to raise $100B minimum to stay alive because it rents every GPU and watt from Google, AWS, and others—the entire business depends on renting infrastructure it doesn't own. Model weights fit on a USB thumb drive, making the core IP trivially replicable. SpaceX, by contrast, owns Starlink, direct-to-cell satellites, X payments infrastructure, $75B in government contracts, manufacturing capacity for turbines (because they're building their own power since there's a five-year wait), Colossus GPU clusters, and direct production of chips via co-design with Nvidia. James cites David Shapiro's critique that Dario (Anthropic's CEO) is a power-seeking narcissist using AI safety as a regulatory capture narrative, and that Anthropic's moral facade is crumbling as competition and open-source models ramp. James won't sell SpaceX for a chatbot that can be copied onto a flash drive—with $8 trillion in cash sitting sidelines and huge supply of dollars chasing returns, a 10-20% SpaceX dip off Anthropic hype is unlikely. Even if it happened, it would be a gift to stack more at discount.
Solana: Agentic Throughput and the Adoption Flywheel
Ken Ed asks where Solana's value comes from given it settles transactions at $0.0004, seemingly like free fiat trading. James reframes it as a bandwidth scarcity play. Solana is a decentralized state machine with massive throughput—26.1M daily active addresses, 10x Ethereum's 2.7M, and vastly ahead of Bitcoin (2.6M), Sui (2.5M), and Tron (7.4M). Stakers capture inflation plus priority fees under SAMD rules; non-stakers rely on fee burns. The app layer captures 93% of on-chain value while the base chain keeps 7%, so growth accrues mostly to applications. The real lever is adoption: AI agents exploded 14x in six months and are already picking Solana as the cheapest, fastest rails for transactions. Each additional agent 10xs the transaction throughput, driving more burn versus issuance and hardening the asset. The flywheel tightens as adoption spreads—more agents pick Solana, more burn, scarcer block space, higher fees, harder asset, price appreciation. James acknowledges risk (any kid in a basement could build a better blockchain) but notes Solana has network effect, developer adoption, and the winner-takes-most dynamics he's preached for years. His position: Solana is a high-beta call on agentic throughput, and the data supports it.
Covered Calls as a Trimming Tool
BB asks whether to use covered calls to trim overweight Nvidia (11%) and Tesla (25%) positions. James calls covered calls the optimal trimming tool from his 1990s options trading playbook. You get paid premium upfront instead of just hitting sell. If shares get called away, your effective sale price is strike plus premium collected—a win. If they don't get called, you keep the premium and the shares. The best approach: sell at-the-money calls about 40 days out to maximize premium while time value dissipates fastest over that window, collecting money on pure time decay with no intrinsic value risk. If the stock is at $350, sell the $350 strike, pocket the premium, and either cap your upside at $350-plus-premium or keep collecting on the next tranche. This beats a blunt market sale because you control the exact price and get paid for the privilege of selling. James doesn't recommend this as day-trading—it's about trimming positions in names you still believe in while maintaining the core bag.
Riot and Bitcoin Mining: Watch Out for Concentration and Obsolescence Risk
Frenchie brings up Riot's $9.1B Anthropic data center deal, pivoting from pure Bitcoin mining to AI infrastructure. James has been in and out of miners—it's a cutthroat business with heavy CapEx, construction risk, and tenant concentration. Riot's 191 megawatt deal doesn't start delivering full power until 2028, so it's far-out revenue with near-term CapEx pain. The stock popped 25% on the news, then immediately retraced all gains—no re-rate happened. James sees three hard problems: one, Elon has locked in 30-40% of the new Vera Rubin NVL72 GPU supply for SpaceX and co-design; two, those chips deliver 30-35x more agent throughput per megawatt than GB300, so anyone stuck with older Nvidia hardware faces massive cost-per-token disadvantage; three, Anthropic itself could be a house of cards (James thinks it's very risky). Revenue is flat at Riot despite $900M debt, though EBITDA and net income are getting less negative. His recommendation: don't hold more than 1% of portfolio in any Bitcoin miner. He's offloading his legacy positions (CleanSpark, etc.) because mining and AI compute are both races to the bottom unless you own the power, the chips, and the vertical stack—which Riot doesn't.