23andMe BANKRUPTCY?, Saylor and MORE Bitcoin, + 3 Founder Interviews

Key Concepts

• 23andMe Bankruptcy: Voluntary Chapter 11 proceedings, potential sale of assets including genetic data.
• MicroStrategy & Bitcoin: Continued accumulation of Bitcoin, potential Ponzi scheme dynamics.
• M&A Activity: Increase in startup acquisitions, potential drivers and implications.
• Browser Use: AI agent-friendly website rendering, $17 million seed round.
• Etched: Transformer-specific chips, high throughput and low latency, Sohoo chip.
• Skyflow: Data privacy and security, AI data stack, Skyflow GPT Privacy Vault.
• Medserve: Secure medication cabinets, drug diversion prevention, point-of-care solutions.
• AI Scaling: Pre-training, post-training, test-time computer reasoning.
• Asic: Application Specific Integrated Circuit
• HBM: High Bandwidth Memory
• PII: Personally Identifiable Information

23andMe Bankruptcy

• Main Topic: 23andMe's filing for voluntary Chapter 11 bankruptcy proceedings.
• Key Points:
  • Revenue declined 12% year-over-year to $44 million (as of November 2024).
  • Operating costs reduced by 17% to $84 million.
  • Gap loss of approximately $59 million.
  • Special committee formed to explore a sale.
  • CEO offered to buy the company but the deal didn't materialize.
  • Potential sale of genetic data as a key asset.
  • Recommendation: Users sensitive about their DNA data should consider deleting it from 23andMe's platform before a sale.
• Arguments:
  • Another example of a SPAC-led debut gone wrong.
  • Lack of recurring revenue model similar to Instant Pot's challenge.

MicroStrategy & Bitcoin

• Main Topic: MicroStrategy's continued investment in Bitcoin.
• Key Points:
  • Increased Bitcoin holdings to over 500,000.
  • Sold 2 million shares to generate $600 million in revenue for Bitcoin purchases.
  • Purchased an additional 7,000 Bitcoin.
  • Michael Saylor has purchased $33.7 billion worth of Bitcoin at an average price of $66.6k per coin.
• Arguments:
  • Potentially resembles a Ponzi scheme, increasing asset value by buying more of it.
  • Circular process of borrowing money or selling stock to purchase more Bitcoin.
  • Hodling strategy drives scarcity and price appreciation.
• Counterpoint: Bitcoin has historically appreciated, making hodling profitable.

M&A Activity

• Main Topic: Increase in startup acquisitions in Q1 2024.
• Key Points:
  • 11 startup sales worth over $1 billion each, totaling $54.5 billion (CB Insights data).
  • In Q1 of the previous year, there were only two such acquisitions, totaling $3.2 billion.
  • Whiz acquisition accounted for $32 billion of the $54.5 billion total.
• Arguments:
  • Potential drivers include changes in regulatory environment and market momentum.
  • "Vibes" or market sentiment plays a role in M&A activity.
• AI's Role: AI could potentially analyze market data and predict optimal times for IPOs.

Browser Use

• Main Topic: Browser Use's $17 million seed round.
• Key Points:
  • Company renders the visual web into a text format for AI agents.
  • Used by Manis, an AI agent.
  • Felicis led the round, with participation from Paul Graham, A Capital, and Nexus Venture Partners.
• Relevance: Addresses the need for AI agents to interact with websites more effectively.

Etched: Transformer-Specific Chips

• Main Topic: Etched's development of transformer-specific chips.
• Key Points:
  • Etched is building chips specifically designed to run transformers.
  • The company's chip is not programmable; it only runs transformer inference.
  • The company claims their chip will outperform existing solutions by 10x.
  • The company raised a $120 million round in June of last year.
  • The company's system is called Sohoo.
  • The company is working with TSMC's emerging businesses group.
• Arguments:
  • Transformers are the bedrock of most AI models.
  • Nvidia's chips are flexible but less efficient for transformer inference.
  • Moore's Law is slowing down, creating an opportunity for specialized chips.
• Technical Details:
  • Multiply accumulate (MAC) operations are key to AI model performance.
  • Nvidia H100 can perform approximately 1 quadrillion MAC operations per second (FP16).
  • Etched aims for over 90% utilization of headline flops.
  • Etched is aiming for 10x better tokens per watt.
• Reasoning Models:
  • Reasoning models can increase the size of the market by 100x a,000x overnight.
• Hiring:
  • The company is hiring for firmware and kernel engineers.

Skyflow: Data Privacy and Security

• Main Topic: Skyflow's approach to data privacy and security in the context of AI.
• Key Points:
  • Skyflow provides a vault for personal information, offering bank-grade security.
  • The company's early customers were fast-growing startups.
  • The company's product is no longer two distinct complete solutions.
  • The company has partnerships with Snowflake, Amazon, and Service Now.
  • The company's gross retention is very close to 100%.
• Arguments:
  • Companies are drowning in data and need secure ways to manage it.
  • Personalized service and AI require the exchange of personal information.
  • Trust is essential in the security and cyber security market.
• Skyflow GPT Privacy Vault:
  • The company's model training and drag infrastructure product line.
• Value Creation:
  • Value is created for customers through revenue and ARR.
  • Value is also created through the multiple, which is created when you are truly building something differentiated.

Medserve: Secure Medication Cabinets

• Main Topic: Medserve's development of secure medication cabinets for outpatient clinics.
• Key Points:
  • Medserve provides digital medication security and management at the point of care.
  • The company's cabinets are priced at $8,000.
  • The company's software ranges between $150 to $600 a month.
  • The company's system is scalable and modular.
  • The company's cabinets are not safes.
• Arguments:
  • Drug diversion is a growing problem in healthcare.
  • Legacy solutions are expensive and not suitable for smaller facilities.
  • The company's cabinets are more secure than kitchen cabinets with locks.
• Use Case:
  • A nurse scans their badge to sign in.
  • The nurse selects a patient.
  • The nurse counts how many vials are in the cabinet.
  • The nurse takes the medication.
• Hiring:
  • The company is hiring for account executives that have experience in healthcare technology sales or capital equipment medical equipment sales.

Synthesis/Conclusion

The podcast episode covers a range of topics from the financial struggles of 23andMe to innovative solutions in AI and healthcare. It highlights the importance of recurring revenue models, the potential risks of SPAC-led debuts, and the increasing M&A activity in the startup world. The interviews with the founders of Etched, Skyflow, and Medserve provide insights into their respective industries and the challenges and opportunities they face. The episode emphasizes the need for deep knowledge, trust, and strategic partnerships in building successful companies.