# The AI boom is a lie: Fake data centres and unused GPUs | Ed Zitron Source: https://www.youtube.com/watch?v=nxUEOdC4VzU Recap page: https://rapidrecap.app/video/nxUEOdC4VzU Generated: 2026-03-29T12:57:22.641+00:00 --- ## Quick Overview The AI boom is built on hype and speculation, not reality, as evidenced by the fact that major AI companies like Anthropic and OpenAI are not generating substantial revenue from their core products, leading to questions about the sustainability of their massive data center build-outs and the eventual price crash for AI-related hardware like GPUs. **Key Points:** - The speaker estimates that only about 3 gigawatts of the planned 190-240 gigawatts of AI data center capacity in the US by 2025 is actually under construction. - Anthropic, despite claiming $1 billion a month in revenue, is suspected of overstating figures, having reportedly lost over $200 million doing so. - The speaker suggests that the reality of AI economics is that large language models (LLMs) are not yet profitable enough to sustain current investment levels without continuous hyper-scaler funding. - The financial community, including private equity firms, is reportedly starting to question the irrational exuberance surrounding AI investment. - Nvidia's annual upgrade cycle means that even if data centers are built now, the GPUs might be obsolete or underutilized by the time they come online, as seen with the 11 gigawatts planned for a datacenter in Texas that was halted. - The speaker believes that when the real economics hit, there will be a reckoning, possibly leading to a price crash for GPUs and a negative sentiment toward the entire AI sector. - The speaker points to the historical precedent of the Dot-com bubble, where hype outpaced actual utility and revenue. ![Screenshot at 00:04: Ed Zitron asserts that the majority of large AI projects, like those from Meta, Amazon, Microsoft, and Google, are essentially 'vaporware' or 'cash machines' that lack real underlying economic viability.](https://ss.rapidrecap.app/screens/nxUEOdC4VzU/00-00-04.jpg) **Context:** The video features a discussion between Ed Zitron, author and host of the Better Offline podcast, and the host of The Tech Report, focusing on the current massive investment and hype surrounding Artificial Intelligence (AI) infrastructure, particularly data centers and GPUs. The conversation critically examines the gap between the proclaimed success of AI companies and the underlying economic reality, suggesting the boom might be inflated by speculation rather than genuine cash flow. ## Detailed Analysis Ed Zitron argues that the current AI boom is largely hype, pointing out that most major AI projects are 'vaporware' or 'cash machines' that are not generating sufficient revenue to justify the massive infrastructure spending. He estimates that only a small fraction (3 gigawatts out of a planned 190-240 GW) of the projected US data center capacity for AI will actually be built by 2025. Zitron cites Anthropic's reported $1 billion monthly revenue as suspicious, given that the company allegedly lost over $200 million in the previous year, suggesting the figures are inflated to maintain investor confidence. He contrasts this with established software companies like Microsoft's 365 or Google Cloud, which have real, demonstrable revenue streams. The core problem, according to Zitron, is that the cost of building these massive AI data centers is not being met by real customer demand or actual output, leading to a situation where companies are building infrastructure they cannot fully utilize or monetize. He notes that even when data centers are announced, like the one in Louton, Texas, they can be halted due to issues like power availability, or they end up being smaller than advertised. Furthermore, the rapid annual upgrade cycle from Nvidia means that hardware installed today could be obsolete by the time these massive facilities become operational, creating a risk of asset devaluation. Zitron concludes that the current irrational exuberance mirrors the Dot-com bubble, predicting that when the true economics catch up, there will be a reckoning, potentially involving a price crash for GPUs and a general downturn for the AI sector, as the underlying demand simply isn't supporting the current investment frenzy. ### AI Hype vs. Reality - Most major AI projects are 'vaporware' or 'cash machines' - They are not generating sufficient revenue to justify massive capital expenditure. ### Data Center Buildout Concerns - Only 3 GW of the planned 190-240 GW of US AI data center capacity might be built by 2025 - Projects like the one in Louton, TX, have stalled or under-delivered. ### Questionable AI Revenue - Anthropic claims $1B/month revenue but allegedly lost over $200M last year, suggesting inflated figures to maintain investor confidence. ### Hardware Obsolescence Risk - Nvidia's annual upgrade cycle means GPUs installed now may be obsolete by the time massive data centers come online, potentially leading to asset write-downs. ### The Financial Reckoning - The current AI bubble is compared to the Dot-com era, where hype outpaced real utility; eventually, someone will need to check the actual numbers, and the underlying demand may not support current investment. ### LLM Economics - LLMs are not yet profitable enough to sustain current investment levels without continuous hyper-scaler funding; the software market is being destroyed by bad debt and bad business. ![Screenshot at 00:04: Ed Zitron argues that the majority of large AI projects are essentially 'vaporware' or 'cash machines' lacking real economic viability.](https://ss.rapidrecap.app/screens/nxUEOdC4VzU/00-00-04.jpg) ![Screenshot at 01:47: A graphic transition shows disassembled portraits of major tech CEOs \(likely including Musk, Cook, Nadella, Pichai\) before transitioning to the show's title screen.](https://ss.rapidrecap.app/screens/nxUEOdC4VzU/00-01-47.jpg) ![Screenshot at 02:55: Ed Zitron expresses skepticism, claiming that the majority of AI projects are 'vaporware' and that Nvidia GPUs might become obsolete before new data centers are even fully utilized.](https://ss.rapidrecap.app/screens/nxUEOdC4VzU/00-02-55.jpg) ![Screenshot at 12:22: Ed Zitron expresses disbelief and frustration over the lack of real progress or revenue matching the hype, suggesting the situation is 'sickening'.](https://ss.rapidrecap.app/screens/nxUEOdC4VzU/00-12-22.jpg)