Where Are The AI Startups? — With Rick Heitzmann
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
In this episode of Big Technology Podcast, host Alex Kantrowitz sits down with Rick Heitzman, managing partner and founder of FirstMark Capital, to explore a fascinating paradox in the tech world. Despite the transformative potential of generative AI, there's been a notable absence of the startup wave many expected to see. (00:49) The conversation reveals that OpenAI's comprehensive product offering and the lack of differentiated training data have created barriers for consumer AI startups, while enterprise applications with proprietary datasets like Harvey (legal) and EvolutionIQ (insurance) are thriving.
• Main themes include the consolidation of AI capabilities within established platforms like ChatGPT, the investment implications of massive AI infrastructure spending, and the potential automation of white-collar work
Speakers
Rick Heitzman
Rick Heitzman is the managing partner and founder of FirstMark Capital, a venture capital firm known for investments in companies like Discord, DraftKings, Shopify, and Airbnb. (00:56) He's a frequent guest on CNBC's Closing Bell and has been investing in technology companies for decades, witnessing multiple waves of innovation from the dot-com era through today's AI revolution.
Alex Kantrowitz
Alex Kantrowitz is the host of Big Technology Podcast and a veteran technology journalist. He previously worked at BuzzFeed covering consumer tech and has built a career spanning freelance journalism, marketing, and now multimedia content creation including video, audio, and television appearances.
Key Takeaways
AI Startups Face a "Breadth and Depth" Problem
The most significant barrier for consumer AI startups isn't technical capability—it's that OpenAI has created a product with both exceptional breadth and depth that's difficult to compete against. (02:29) As Heitzman explains, successful AI applications typically need either highly specific datasets (like Harvey's legal documents) or discrete regulatory requirements that prevent general-purpose models from serving the use case effectively. For general consumer applications like fitness coaching or travel planning, ChatGPT's broad knowledge base often provides sufficiently good results, making it hard for startups to justify their existence to investors or users.
Data Ownership Creates Sustainable Competitive Advantages
The most successful AI companies are those that control proprietary, industry-specific datasets that can't be easily replicated by general models. (03:33) Companies like EvolutionIQ in insurance and Henry in commercial real estate succeed because they have access to "discrete and sometimes private datasets" that enable better, more specific applications. This suggests that rather than competing on model capabilities, successful AI startups should focus on securing exclusive access to valuable training data within specific verticals.
Enterprise Applications Require Security-First Thinking
As AI adoption accelerates, data privacy and security concerns are becoming major limiting factors for how organizations deploy these technologies. (12:15) Heitzman notes that companies are increasingly wary about sharing sensitive information with general-purpose models, creating opportunities for AI solutions that operate within "walled gardens" of proprietary data. This trend toward private, secure AI deployments represents a significant investment opportunity and could reshape how enterprises adopt AI technologies.
White-Collar Automation Is Accelerating Job Market Changes
Unlike previous technological revolutions that primarily automated blue-collar work, AI is simultaneously targeting both white-collar and blue-collar jobs, fundamentally changing hiring patterns. (37:01) Companies are becoming more cautious about hiring, particularly for roles that involve routine knowledge work like creating presentations or writing emails that generate little value. This shift is contributing to the difficult job market for recent graduates and longer job search times, even in a seemingly strong economy.
Historical Patterns Suggest Creative Destruction, Not Permanent Job Loss
Despite concerns about AI-driven unemployment, historical precedent suggests that technological advancement typically creates more jobs than it destroys, albeit in different categories. (37:57) Heitzman points to the transition from an agrarian economy (93% of Americans were farmers in 1900 versus 3% in 2000) during what became "the greatest century of an economy of any civilization's economy in the history of civilization." The key insight is that while specific job categories may disappear, human creativity and entrepreneurship typically generate new forms of valuable work—though the transition period can be challenging for individuals.
Statistics & Facts
- At the beginning of the 20th century, about 93% of Americans worked in the agrarian economy as farmers, while by the end of the century, only 3% of the American workforce were farmers. (38:01) Heitzman uses this statistic to illustrate how massive job displacement can coincide with economic prosperity through creative destruction.
- NVIDIA recently committed $100 billion to OpenAI, with Jensen Huang putting up $10 billion initially with plans to contribute another $90 billion in increments. (27:31) This represents one of the largest technology investments in history, though structured as a "partnership first, investment second."
- Rick Heitzman mentions that last year was "the worst year since the financial crisis for recent college graduates" in terms of job market conditions. (40:15) This statistic highlights the current challenges facing young professionals entering the workforce during the AI transition.
