Blitzhiring and the Complicated Case of Windsurf, OpenAI and Google | E2151

Key Concepts

AI, Machine Learning, Wildfire Detection, Startups, Venture Capital, Acquisitions, Regulation, Autonomous Vehicles, Open Source AI, Server Depreciation, Talent Acquisition, Backdoor Acquisitions, QSBS (Qualified Small Business Stock), AI Safety, Job Displacement, Technology Trends.

Wildfire Detection and Prevention

• Problem: Arson and wildfires are a significant problem, and current detection methods are insufficient.
• Example: The video discusses an incident where influencers in Runyon Canyon caught an arsonist.
• Startup Angle: Problems like wildfire detection present opportunities for startups to develop solutions.
• Existing Companies: Pano AI and Dryad Networks are mentioned as companies working on wildfire detection using cameras. Pano AI raised $44 million last month.
• Technological Solutions:
  • Cameras: The use of 360-degree cameras (Insta360, GoPro), iPhone cameras, and security cameras (Nest, Dropcam) for early fire detection is discussed.
  • 5G Solar Cameras: These cameras can be deployed in remote areas without Wi-Fi, using 5G data plans. They cost around $129.
  • Drones: The potential for drones to deliver fire retardant material is explored as a faster and cheaper alternative to helicopters (which cost $10,000/hour).
• Call to Action: The hosts suggest that every trail should have cameras every five miles to detect fires quickly.
• Cost Analysis: The cost of deploying 5G solar cameras is considered relatively low, making it a feasible solution.
• Zipline Analogy: Zipline, a drone delivery company, is suggested to pivot to fire retardant delivery.

Windsurf Acquisition and "Backdoor Acquisitions"

• Windsurf Deal: Windsurf, an AI company, was initially in talks to be acquired by OpenAI for $3 billion (at a 100 million ARR), but the deal fell apart. Google then acquired the company for $2.4 billion in a "backdoor acquisition." Cognition AI (Devon) is buying what's left of Windsurf.
• 49% Agreements/Backdoor Acquisitions: These deals involve acquiring a significant portion of a company (e.g., 49%) without a full acquisition.
• Illy's Thesis (Category VC): These deals are driven by the need to acquire talent quickly and avoid regulatory delays associated with larger acquisitions (HSR Act, Sherman Antitrust Act). Deals over $150 million require government notification and can take up to a year to process.
• Tax Implications:
  • Asset Sale/Licensing Deal: Revenue into the company, subject to corporate tax and potential double taxation when dividends are distributed.
  • Capital Gains: Taxed at the federal capital gains rate (potentially lower in states with no income tax).
  • QSBS (Qualified Small Business Stock): Offers a waiver for the first $10 million in gain if the company's asset value is under $50 million.
• Structuring the Deal: It's important to leave enough of a "carcass" (remaining company) to avoid regulatory scrutiny.
• Examples of Backdoor Acquisitions:
  1. Microsoft/Inflection AI ($650M, IP deal)
  2. Amazon/Adapt AI (Nine figures, licensing deal)
  3. Google/Character AI ($2.5B, non-exclusive IP)
  4. Meta/Scale AI ($14.8B, stake in the company)
  5. Google/Windsurf ($2.4B, license, co-founders, talent)
• Motivation: Speed and avoiding regulatory hurdles are the primary drivers.
• Jurassic Park Quote: "Life finds a way" is used to describe how VCs and corporate M&A find ways to circumvent regulations.

AI Trends and Job Displacement

• AI Prize: The AI market is estimated to be worth $10 trillion in market cap.
• Skepticism: Some believe the AI hype is a bubble and that the technology won't reach full completion.
• Job Displacement: AI is expected to displace many jobs, particularly those involving routine tasks.
• Static Team Size: Companies are focused on hiring top-tier AI experts to replace larger teams of less specialized workers.
• AI as a Job Replacement Tool: AI is being used to monitor workers and train systems to automate their jobs.
• Example: Safety drivers in autonomous vehicles are training the AI to replace them. Developers using tools like Cursor are inadvertently training the AI to take their jobs.
• Cognition AI (Devon): Devon is an AI engineer that can be hired to work with companies.
• AI Coding Assistants: Google (Gemini Code Assist, CLI), Anthropic (Cloud Code), OpenAI, and Amazon (Hero) are all developing AI coding assistants.
• Vellum: A company focused on bridging the gap between AI prototypes and production, claiming to have SAS gross margins.
• Token Pricing: The price of tokens for AI models is decreasing due to competition, potentially improving margins for AI companies.

Autonomous Vehicles and Regulation

• Tesla's Expansion in Austin: Tesla is expanding its robotaxi service in Austin, potentially surpassing Whimo in geographic footprint.
• Safety Concerns: The hosts discuss incidents involving Tesla's autonomous vehicles, such as tire touching, wrong lane driving, and phantom stops.
• Importance of Safety Drivers: The hosts emphasize the importance of safety drivers (or monitors) to mitigate risks.
• Regulatory Scrutiny: Regulators are likely to intervene if autonomous vehicles have too many incidents, even minor ones.
• Local Regulation: The hosts argue that local regulators should have the authority to set rules for autonomous vehicle deployment.
• Slow and Steady Approach: The hosts advocate for a cautious approach to autonomous vehicle deployment, prioritizing safety over speed.
• Invite-Only Access: Tesla's robotaxi service in Austin is currently invite-only.
• Pricing: A ride from the northern tip of the service area to the base costs $4.20, suggesting subsidized rides in the early days.

Open Source AI

• Moonshot AI (China): This company has released high-quality, open-source AI models (Kimmy K2).
• Significance: This challenges the trend of closed-source AI dominance and provides a valuable resource for the global community.
• Comparison to OpenAI: Moonshot AI is fulfilling the original vision of OpenAI (before it became a for-profit company) by making AI freely available.
• Potential Impact: Open-source AI could become as prevalent as open-source software like WordPress and MySQL.
• Mistral (France): A French AI company that is ranked sixth in the world for token usage on Open Router.
• AI Model Leaderboard: The hosts discuss the constantly shifting landscape of AI model leadership, with different models (Grock, Gemini, Claude, OpenAI) taking the lead at different times.
• Poly Market: A prediction market where people can bet on which company will have the best AI model.

Server Depreciation

• Hyperscaler Depreciation Timelines:
  • Alphabet: 6 years
  • Meta: 5.5 years
  • Microsoft: 6 years
  • Amazon: 5 years
• Financial Impact: Extending server lifespans can save hyperscalers billions of dollars.
• Coreweave: Depreciation and amortization costs at Coreweave have risen significantly due to increased server investments.
• Revenue vs. Cost: The key question is whether the revenue generated by these servers justifies the cost.
• Afterlife of Servers: The hosts question what happens to servers after their depreciation period (resale, repurposing, recycling).
• Nvidia Recycling Program: Nvidia has a recycling program for its GPUs.

Conclusion

The episode covers a wide range of topics related to technology and startups, including wildfire detection, AI acquisitions, job displacement, autonomous vehicles, open-source AI, and server depreciation. The hosts emphasize the importance of innovation, but also caution against moving too quickly without considering safety and regulatory implications. The discussion highlights the dynamic and competitive nature of the AI landscape, with new models and companies constantly emerging. The "backdoor acquisition" trend is identified as a way for companies to acquire talent and technology quickly while avoiding regulatory hurdles. The episode concludes with a discussion of server depreciation and the need to understand the long-term economics of AI infrastructure.