The Connected News Podcast · Business Post

The Connected AI Podcast: AI and accountability

·1 hr 9 min·5 clips

CoreWeave made $1.9 billion in 2024 revenue, but its net losses since 2022 total $1.5 billion.

1. The Connected AI Podcast centers this episode on CoreWeave’s AI IPO, AI accountability, and the legal fights around training data. 2. Charlie Taylor and Elaine Burke host the episode, and Abebe Birhane joins later as a guest who leads an AI accountability lab at Trinity College. 3. The episode asks what happens when AI companies, debt financing, and data extraction collide in the same market. 4. Charlie and Elaine describe CoreWeave as a New Jersey startup founded in 2017 as a crypto-mining business before pivoting to AI compute. 5. They say CoreWeave has raised more than $14.5 billion, with a large share coming from debt financing rather than equity. 6. They note that CoreWeave’s IPO target is about $2.7 billion for 49 million shares, implying a valuation near $32 billion. 7. The hosts stress that CoreWeave’s customers include Meta, IBM, Microsoft, and OpenAI, while its competitors are cloud providers like Microsoft and Amazon. 8. They explain that CoreWeave rents about 30 data centers instead of owning them, and that many are leased from Core Scientific. 9. They say Core Scientific’s shares have plunged 40% over the past year, which adds another layer of risk to CoreWeave’s infrastructure. 10. The discussion turns to CoreWeave’s revenue growth from $16 million in 2022 to $228.9 million in 2023 and $1.9 billion last year. 11. Elaine and Charlie highlight that the company is still not profitable, with net losses since 2022 totaling $1.5 billion. 12. They call the debt structure “Ouroboros” and “chicken and egg stuff,” because contracts are used to raise capital and capital is used to build the contracts. 13. The hosts say roughly two thirds of CoreWeave’s revenue comes from Microsoft, even though Microsoft is also a competitor. 14. They point out that Microsoft pulled back on a $12 billion option to buy more capacity from CoreWeave, and OpenAI then took that contract. 15. They add that OpenAI will receive $350 million worth of shares through a private placement at the IPO. 16. Charlie compares the CoreWeave story to the dot-com bubble and says the company may become a warning sign for the wider IPO market. 17. Abebe Birhane describes her Trinity lab as an effort to do empirical accountability research on datasets and AI models. 18. She says her team audits data sets, evaluates AI models, and brings findings to vendors, developers, and the public. 19. The episode’s style moves from news analysis to a long interview, with a conversational tone and frequent concrete examples. 20. Listeners interested in AI governance, data audits, and tech market risk will get the most from this episode; listeners wanting light AI news may skip the detailed debt discussion.

As heard by us

A clear-eyed look at who can audit AI, and why so much of the data remains out of reach.

This episode treats AI accountability as a matter of access and power, not branding. Charlie Taylor and Elaine Burke speak with Abebe Behan, described here as setting up an AI accountability lab at Trinity College, and the conversation moves from headlines to the basic question…

Read the full review in PlayNext →

Why you'd press play

For when you want the AI accountability story without the conference-brochure polish.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from The Connected News Podcast