Post

Free newsletter: In July, Apollo Economist Torsten Slok said that "AI better work out, or we have a problem." Facing ever-growing high-interest debt, negative-EBITDA AI labs and $200bn+ in warehoused GPUs, that's exactly what the AI industry has. www.wheresyoured.at/ai-had-bette...

wheresyoured.athttps://www.wheresyoured.at/ai-had-better-work-out/If you liked this piece, you should subscribe to my premium newsletter, and you can subscribe on the following links: $70 a year, $18 a quarter, or $7 a month. In return, you get a weekly premium…

As I've written extensively in the past, the AI bubble is nothing like the dot com bubble. The scale and concentration is different, and a simple 20% correction across the main AI stocks would be worse than the worst day of the Dot Com crash. www.wheresyoured.at/ai-had-bette...

The AI bubble is sustained by vibes and hype, and both have started to sour, with Gartner saying that we have more compute capacity than demand, and that 80% of neoclouds will go out of business by 2030, adding there's not enough revenue to justify them. www.wheresyoured.at/ai-had-bette...

AI is also concentrated around a handful of companies, most notably NVIDIA, which is forecasting over $670bn in revenue next year, at a time when the debt its customers need to buy its chips is increasingly unaffordable as lenders sour on AI debt. www.wheresyoured.at/ai-had-bette...

It's now confirmed that Amazon has warehouses of GPUs that can’t be installed, with in some cases "no clear path to doing so" despite taking billions in deliveries monthly. The RAM crisis was created by NVIDIA to boost its revenues through early delivery. www.wheresyoured.at/ai-had-bette...

11 reposts80 likes

The GPU trade is the irrational core of the bubble, and the argument that hyperscalers are hoarding scarce commodities falls over when you see that *everyone* has billions in dark GPUs. NVIDIA seems to be channel stuffing, and we all suffer as a result. www.wheresyoured.at/ai-had-bette...

In fairness, the whole "abandoning reason" has seemingly already happened. This late in the bubble, there is no consensus on how much a gigawatt of capacity should cost, and many of the projections thrown around exceed the limits of profitability. www.wheresyoured.at/ai-had-bette...

Profitability, of course, assumes that the customer for the compute exists, and they have customers. Unfortunately, OpenAI and Anthropic are overwhelmingly dependent on VC-backed startups, at a time when VC's underperforming and running out of cash. www.wheresyoured.at/ai-had-bette...

They will try to shake down governments, installed 'ai' ministries and internal offices already diverting budgets, while using public broadcasters (e.g. CBC) &prime ministers (Carney, Albanese, etc) to promote the replacement of services with 'ai' ($). Not saying it will work,but it's long in motion