Where Are The AI Startups? — With Rick Heitzmann

Episode Summary

Executive Summary: The conversation examines why consumer AI startups haven’t exploded the way many expected, arguing that broad models like ChatGPT now cover many “wrapper” use cases, while defensibility increasingly comes from proprietary data, regulated workflows, and security. Rick Heitzman also discusses AI-driven job disruption, the likely persistence of creative destruction, and why hyperscaler spending and frothy AI market expectations may still keep expanding.

Main Topics: Why consumer AI startups haven't boomed (Priority: 5/5): Heitzman argues that ChatGPT’s breadth, depth, and rapid product improvement have absorbed many obvious consumer point solutions, making it hard for startups to offer step-function improvements. Data as the main moat in AI (Priority: 5/5): The strongest AI companies, in his view, are those with proprietary, private, or highly structured data sets—especially in enterprise verticals like legal, insurance, and commercial real estate. Security and privacy as investable layers (Priority: 4/5): He sees data security, model security, and walled-garden architectures as major opportunities because companies and users may not want sensitive data flowing into large public models. AI’s effect on work and jobs (Priority: 5/5): The discussion turns to automation of white-collar work, with Heitzman expecting some job loss in specific fields but also significant creative destruction and reallocation of labor, similar to past technological shifts. Market froth and hyperscaler spending (Priority: 4/5): Heitzman says the AI boom is being fueled by massive hyperscaler investments, much of it funded through operating cash flow rather than external capital, which keeps the cycle going and raises expectations. Consumer and commerce interfaces may change, not disappear (Priority: 3/5): He argues commerce and consumer products may increasingly be initiated through chatbots, but the underlying infrastructure—payments, logistics, fraud, fulfillment—will remain essential. Portfolio examples: Discord, DraftKings, Shopify, Airbnb (Priority: 3/5): He gives quick perspectives on privacy-oriented communities, sports betting regulation, AI-enabled commerce, and the regulatory damage of Airbnb’s New York restrictions.

Key Arguments: ChatGPT has become good enough across many general-purpose tasks that it suppresses demand for many standalone consumer AI startups. Startups are more defensible when they have distinct, private, or regulated data sets that improve model performance and product outcomes. Many AI products that succeed today are effectively wrappers, and those without a meaningful step-function advantage are hard to fund. Security and data governance are emerging as core AI investment areas because enterprises need control over what data is shared with models. AI will automate many white-collar tasks, but historically technology has also created new jobs and categories over time. The biggest near-term risk is not only job displacement but also overcapacity and frothy market expectations tied to AI infrastructure and model partnerships. Consumer interfaces may shift toward chatbots, but e-commerce and software back-end infrastructure still matter because the physical and financial world does not disappear. Discord-like semi-private communities reflect a broader move away from public social broadcasting toward curated, smaller-group interaction.

Data Points: FirstMark largest check: $200 million - Heitzman says FirstMark’s biggest check to date is around this amount. NVIDIA commitment to OpenAI: $100 billion - Discussed as a historic, partnership-like investment/commitment rather than a conventional check. OpenAI initial contribution: $10 billion - Heitzman notes the deal was framed as $10B initially, with another $90B planned in increments. OpenAI follow-on contribution: $90 billion - Planned additional funding in the partnership structure discussed on the show. Public market reaction scale: $100 billion previously exceeded the market cap of all but a few companies - Heitzman notes how mind-boggling the number is in historical context. U.S. farm workforce share at start of 20th century: 93% - Used to illustrate creative destruction and labor reallocation. U.S. farm workforce share at end of 20th century: 3% - Used in the same historical comparison to show massive employment shift. Big Technology Discord community: 11,000 to close to 12,000 likes on a Substack post - Referenced when discussing how widely a post about corporate-job meaning resonated; the post itself was liked this many times. Shopify example: Revenue up while employee base reduced - Cited as a company using AI/productivity gains to maintain or increase output with fewer people. AI and recent college grads: Worst year since the financial crisis - Heitzman says recent graduates are experiencing a difficult job market.

Pivotal Quotes: "ChatGPT has done a great job of making a very good product that has both breadth and depth." — Rick Heitzman: Used to explain why many consumer AI startup opportunities are being absorbed by one dominant general-purpose product. "The best people are going to be the people who understand it a little ahead of time." — Rick Heitzman: Said in the context of Gen Z and workers needing to adapt to AI-driven labor change. "I think there is a creative destruction that I think capitalism is really good at." — Rick Heitzman: Used to frame AI disruption as a long-run reallocation of labor rather than a simple end to work.

Implications: AI startups must prove unique data, workflow, or security advantages or risk being subsumed by foundation-model platforms. Expect more job churn, especially in entry-level white-collar roles, but also new AI-native business models and communities to emerge.