23andMe BANKRUPT, Saylor Still Buying Bitcoin, + Founder Interviews | E2102

Episode Summary

Executive Summary: The episode opens with two major business stories: 23andMe’s bankruptcy, framed as a cautionary tale about weak recurring revenue and SPAC-era public listings, and MicroStrategy’s continued Bitcoin accumulation, discussed as a circular bet on scarcity and market premiums. The bulk of the show features founder interviews on three very different but complementary infrastructure plays: Etched’s transformer-specific chips, Skyflow’s data-security platform evolving into AI data protection, and MedServe’s secure medication cabinets for outpatient care. Across all segments, the theme is specialization and defensibility in a rapidly changing market.

Main Topics: 23andMe bankruptcy as a cautionary startup cautionary tale (Priority: 5/5): The hosts discuss 23andMe’s Chapter 11 filing, declining revenue, ongoing losses, cash burn, and the difficulty of building a repeatable business around a one-time consumer DNA test. They frame it as a SPAC-era outcome and a lesson in recurring revenue fundamentals. MicroStrategy’s Bitcoin accumulation and market circularity (Priority: 4/5): They debate whether Michael Saylor’s strategy is fundamentally different from a leveraged/recursive scheme: raising capital to buy more Bitcoin, thereby potentially inflating the value of the asset and the company’s stock premium. M&A and exit activity returning in startups (Priority: 3/5): The hosts note a sharp rise in billion-dollar startup sales early in the year, suggesting a thaw in liquidity and exit markets, with Wiz cited as the largest driver of the headline number. Etched’s transformer-only ASIC bet (Priority: 5/5): Etched explains its thesis: if transformers remain the dominant AI architecture, a non-programmable chip optimized solely for transformer inference can outperform general-purpose GPUs/TPUs on throughput, latency, and cost per token. Skyflow’s shift from privacy infrastructure to AI data security (Priority: 5/5): Skyflow describes how it began as a vault for PII and evolved into a core layer for enterprise AI, helping companies protect structured and unstructured sensitive data while still training and deploying AI models. MedServe’s secure medication cabinets for non-hospital care settings (Priority: 4/5): MedServe addresses drug diversion and inventory control in outpatient clinics and other smaller healthcare facilities with a lower-cost, cloud-connected cabinet system that replaces insecure locks and paper logs.

Key Arguments: 23andMe’s collapse illustrates how a company can have a compelling consumer novelty but fail without recurring revenue and a second act. SPAC listings can create fragile public companies that face sharper downside when growth slows and capital markets tighten. MicroStrategy’s Bitcoin strategy is presented as circular: issue equity or debt to buy more BTC, which may support both BTC and MSTR’s premium valuation. ETched argues that programmable chips sacrifice efficiency; purpose-built transformer ASICs can deliver far higher utilization and lower cost per token. The AI boom does not eliminate the need for privacy infrastructure; it makes data governance more important because enterprises want to use sensitive data safely in models. Skyflow argues that trust, compliance, and data residency become sticky infrastructure concerns, producing durable revenue and near-100% retention. MedServe argues that healthcare medication security is still under-digitized outside hospitals, creating a real market for affordable, secure, auditable cabinets. The show repeatedly emphasizes that specialization—whether in chips, privacy, or healthcare workflows—creates defensibility and pricing power.

Data Points: 23andMe revenue: $44 million - Latest financials discussed for fiscal Q2 2025 23andMe revenue change YoY: -12% - Revenue down from the year-ago period 23andMe operating costs: $84 million - Reported operating costs in the same period 23andMe operating cost change YoY: -17% - Costs reduced versus prior year 23andMe GAAP loss: $59 million - Quarterly loss reported Bitcoin holdings added by MicroStrategy: 7,000 BTC - Purchase that pushed holdings over the 500,000 BTC mark MicroStrategy share sale proceeds: About $600 million - Raised from selling roughly 2 million shares in the cited period MicroStrategy preferred share sale proceeds: About $1 million - Additional STRK share sales mentioned MicroStrategy total Bitcoin bought since inception: $33.7 billion - Total cumulative BTC purchases cited by the hosts Average Bitcoin purchase price: $66.6K per coin - Average cost basis noted for Saylor’s holdings Current Bitcoin price discussed: About $88K - Host referenced approximate market price at the time Startup acquisitions over $1B in Q1: 11 deals - CB Insights data discussed from Bloomberg reporting Total value of those deals: $54.5 billion - Aggregate value of billion-dollar startup sales in Q1 Value of Wiz deal: $32 billion - Largest deal contributing to the Q1 total Etched seed round: $120 million - Funding raised after the transformer-ASIC thesis gained traction Browser Use seed round: $17 million - Raised to make websites more agent-friendly for AI systems Etched claimed inference advantage: 10x - Company said its chip can outperform NVIDIA Blackwell on transformer inference Etched claimed utilization: >90% - Expected flops utilization for its ASIC versus much lower on GPUs Typical GPU utilization discussed: ~33% or less - Hosts and founder discussed effective utilization versus headline specs Meta Llama 3 utilization: 38%-42% - Referenced as an example of strong but still partial GPU flop utilization Blackwell die improvement: 25% more flops - Compared with Hopper/H100, according to discussion NVIDIA 8x H100 system cost: About $300,000 - Used to illustrate AI server CapEx Power draw for 8x H100 system: About 10 kW - Referenced as yearly operating cost context Power cost rule of thumb: $1 per watt per year - Used for rough operating expense estimation Skyflow time to first $1M ARR: ~9 months - Founder claimed early traction with startups Skyflow early customer ACVs: A few thousand to $100,000/year - Initial deal sizes discussed MedServe hospital cabinet cost: About $150,000 each - Legacy enterprise hospital systems MedServe cabinet cost: $8,000 - Shown as the lower-cost alternative for outpatient settings MedServe software price: $150-$600/month - Recurring software fee layered on hardware Healthcare provider self-reported substance abuse: ~7% - Used to underscore diversion risk MedServe cabinet height: 36 inches - Described during product demo MedServe strategic exit timeline: 3-5 years - Founder suggested acquisition could happen in that window

Pivotal Quotes: "The startup angle here is twofold. Therefore, one, don't go public via SPAC. And two, SAS is popular because. It involves inherently recurring revenue." — Alex: After discussing 23andMe’s bankruptcy and why one-time consumer DNA testing struggled as a business "It's just, it's just going into condition more stock because the stock went up, and then you can buy more." — Lon Harris: During the discussion of MicroStrategy’s circular Bitcoin-buying strategy "Our chip is not programmable. It does one thing and one thing only, and that is run transformer inference." — Gavin Uberti: Explaining Etched’s core hardware thesis for building a transformer-specific ASIC

Implications: The episode highlights a market shift toward specialized, infrastructure-heavy startups with defensible economics. AI will reward companies that improve cost, trust, and workflow integration, while consumer novelty businesses and undifferentiated growth stories remain vulnerable.