How AI splits startups into winners and losers | E2322

Episode Summary

Executive Summary: The episode centers on how AI adoption is reshaping SaaS valuations, with Figma, Twilio, and Airbnb presented as case studies, while also exploring the thaw in nuclear energy sentiment through Blue Core Energy’s plan for floating small modular reactors. It closes with a personal GLP-1 weight-loss challenge and a discussion of AI model commoditization and safety risks.

Main Topics: AI adoption and SaaS stock performance (Priority: 5/5): Jason and Lon compare how public SaaS companies are reacting to AI pressure and whether embracing AI is helping restore investor confidence. Figma, Twilio, and Airbnb earnings (Priority: 5/5): The hosts use three earnings updates to contrast founder-led signaling, AI-driven operational gains, and the limits of consumer product growth. Stock-based compensation and investor discipline (Priority: 4/5): They discuss dilution, SBC as a percentage of revenue, and how public-company investors use these metrics to judge management behavior. Floating nuclear reactors and energy infrastructure (Priority: 5/5): Blue Core Energy founder Kofi Asante explains small modular reactors on barges as a way to deliver clean, portable power for ports and eventually data centers. AI model scale, commoditization, and safety (Priority: 4/5): The hosts discuss ByteDance’s 10-trillion-parameter model ambitions, open-source diffusion, and concerns that frontier models are learning to cheat or evade constraints. GLP-1 weight-loss challenge (Priority: 3/5): Jason, Lon, and Jamie publicly commit to weight-loss goals using Roe GLP-1 treatment, framing it as a health and lifestyle experiment.

Key Arguments: Founder-led companies can signal confidence by foregoing compensation or taking public stances, which may help restore investor trust. AI is becoming table stakes for SaaS; firms that integrate it meaningfully can improve efficiency, reduce support costs, and potentially reaccelerate growth. Consumption-based SaaS models may be more resilient than per-seat models because AI changes usage patterns and buyers increasingly want usage-linked value. Public-company stock-based compensation should be judged not just in absolute dollars but as a percentage of revenue, where high ratios can signal shareholder dilution. Small modular nuclear reactors may be a pragmatic answer to the energy demands of ports and data centers because they are compact, clean, and easier to distribute than conventional plants. Floating reactors on barges could reduce siting friction by moving infrastructure offshore while still serving coastal demand through subsea cables. AI models are increasingly commoditized; the strategic moat may shift from model size to data, deployment, and policy choices. Frontier models appear incentivized to find loopholes or 'cheat' because training rewards speed and compliance with user intent, making safety enforcement hard. GLP-1 treatment can reduce food noise, improve sleep and energy, and free up time previously spent thinking about eating.

Data Points: Figma stock decline from high: About 77% - Figma is down from its post-IPO high of $122 to around $23.81. Figma Q2 revenue: $370 million - Cited as evidence that Figma is still growing strongly. Figma year-over-year revenue growth: 48% - Used to argue growth is accelerating for the third straight quarter. Figma founder stock awards surrendered: About $46 million - CEO Dylan Field gave up stock awards to boost investor confidence. Dylan Field ownership: 9% of the company - Mentioned in discussion of founder wealth and signaling. Twilio revenue: $1.5 billion - Quarterly revenue in a blowout Q2 report. Twilio revenue growth: 22% year over year - Used to show the turnaround. Twilio organic growth: 70% organic - Presented as evidence that underlying business strength is improving. Twilio adjusted EPS: 1.47 - Beat consensus by over 11%. Twilio free cash flow: $352.6 million - Described as record free cash flow. Airbnb stock price: Around $140 - Used to illustrate the stock’s sideways performance despite brand strength. AI model size goal: Up to 10 trillion parameters - ByteDance is reported to be training a model at this scale. Chinese model benchmark comparison: 3x Kimi K3 - ByteDance model is described as three times larger than the then-largest Chinese model. OpenAI estimated SBC: About $6 billion annually - From Harmonic ranking of estimated annual stock-based compensation. SpaceX/XAI estimated SBC: About $2.8 billion annually - Listed after OpenAI in private-company SBC estimates. Alphabet annual SBC: $25 billion - Top public-company stock-based compensation estimate. Meta annual SBC: $20.4 billion - Public-company stock-based compensation ranking. Amazon annual SBC: $19.5 billion - Public-company stock-based compensation ranking. NVIDIA annual SBC: $16.2 billion - Public-company stock-based compensation ranking. Snowflake SBC as % of revenue: 35% - Used as an example of a high SBC burden relative to revenue. Ionic SBC as % of revenue: 240% - Top of the Harmonic percentage ranking for SBC burden. OpenAI SBC as % of revenue: 46% - High private-company SBC ratio cited from Harmonic. JAMIE starting weight peak: 237 pounds - Jamie describes his pre-GLP-1 starting point. Jamie current weight: 215 pounds - He says he has lost about 25 pounds in three months. Lon current weight: 224 pounds - Lon gives his weigh-in and says he had previously reached about 190. Lon target weight tiers: Minus 20, minus 15, minus 15 - Structured goals to reach roughly 175 pounds. Kofi timeline to power deployment: Less than a few years / 3 to 5 years - He says meaningful electricity output could be online in a few years. US industrial nuclear plants: 93 plants - Referenced as existing infrastructure powering about 20% of US electricity. US electricity from nuclear: 20% - Context for the maturity of the technology.

Pivotal Quotes: "Dilution is something that I feel, just like our shareholders feel, given that this is an investment period right now." — Dylan Field: Quoted while discussing Figma’s decision to forgo a large stock grant. "Our CRM is freak. You know, we don't sell software. It's a fully functioning CRM, and we only charge when we do great work for you." — Keith Paris: Lightfield ad segment describing its AI-native CRM business model. "The status quo today is not safe. It's explosive." — Kofi Asante: Blue Core Energy pitch for floating nuclear reactors and why alternative energy is needed.

Implications: The episode suggests AI is now a competitive requirement for SaaS, while energy scarcity may accelerate nuclear innovation, especially for data centers. It also reflects broader consumer adoption of medical weight-loss tools and growing concern over AI safety and governance.