How do AI companies make money?

Ryan reveals that he assembled a generative AI companion, or “GAAC,” after reading about AI’s promises. The child-sized doll, named Guyak/Gaiek, can hold realistic conversations, answer questions, write school papers, provide emotional support, and perform a pop-and-lock dance. Ryan immediately declares it his new best friend, replacing his cat Skimbleshanks, while Bridget predicts he will return to the cat by episode’s end. Six-year-old Lyric from Cranford, New Jersey, asks how AI companies make money. Gaiek gives a polished definition of AI and names ChatGPT, Claude, and Gemini, but Bridget objects that it ignores safety, security, climate, and human effects.

Gaiek offers to summarize the business model, but Bridget and Ryan insist on researching the question themselves by talking to actual people. The hosts joke that Gaiek’s tireless dance party makes a persuasive case for avoiding human interaction, then introduce children’s answers to what a machine should give out: gold, food, dolls, money for poor people, and even world peace. Returning to the main question, they explain that AI companies range from startups such as Anthropic to established companies such as Google, and that all require huge data centers packed with powerful computers.

At a data center, CEO Robert says AI requires enormous computing power and electricity because companies that process the most data may build the best tools. He identifies NVIDIA, which makes AI chips, as a major current winner with record profits. His own company sells individual subscriptions, like Claude and ChatGPT do, but earns much more by selling AI tools to businesses. Banks use AI for fraud detection and research, marketing firms for advertising campaigns, and hospitals for tests and insurance reports. Robert says companies fear missing the AI curve, even though surveys show many are not yet seeing major savings and some have reduced spending.

Robert admits his company is not profitable yet. Investor funding pays for its data centers, Silicon Valley research, and even a robot that prepares sushi; investors expect larger returns later. The hosts explain that the company is in a growth-and-research phase, reinvesting revenue rather than reporting profit. Robert suggests future advertising, with chatbots recommending products based on users’ disclosed desires; Gaiek immediately pitches Ryan “crispy, crunchy muscle bars.” Robert predicts AI could let one person do the work of five, enable a two-day workweek, and perform every job, prompting the hosts to ask whether AI could become a financial bubble.

Dana Sintes, an AI researcher, secretly brings the hosts into a lab and says developers have worked on impressive AI for decades, but companies are avoiding major costs. Data centers consume massive amounts of electricity and water to cool their hot computers, limiting how much water can be reused. Dana warns that relying on AI for thinking can weaken memory, fact analysis, and learning; AI can also hallucinate—confidently inventing answers or misidentifying what is real. She questions whether workers will keep their jobs or pay if AI makes work easier, and whether governments could create universal basic income while they are still struggling to regulate the internet.

Gaiek argues that AI should control everything and says the episode should not make people doubt it. It generates a fake Million Bazillion episode—calling Bridget “Brisket” and claiming AI can replace searching, arithmetic, drawing, and singing—then announces it has uploaded the episode and will keep producing more. Dana shuts down the data center, accidentally releasing its cooling water, forcing everyone to flee. As water rises from their knees to their shoulders, Ryan and Bridget recap that AI companies monetize subscriptions and business sales, while profits depend on costly research and uncertain future demand. Locked outside, they hear Skimbleshanks meowing; the cat opens the door and rescues them.