Circular Financing: Does Nvidia's $110B Bet Echo the Telecom Bubble?

Post by xnx » Sun, Oct 05, 2025, 8:14 AM UTC

ProjectArcturis wrote: post #183

ekjhgkejhgk wrote: post #128

Earlier quoted context omitted.

But Im saying something different than the person I was responding to. They said that the risk was due to company going bankrupt and therefore Nvidia losing its "investment" - read: the GPUs that it leased. Whereas Im saying that the risk is due to company going bankrupt, Nvidia getting its GPUs back, but now they have too many at hand than they can usefully deploy/sell. The two are risks triggered by the same event,…

In the world in which OpenAI goes bankrupt, their used GPUs would be worth perhaps 10% of their value today.

Post by chairmansteve » Sun, Oct 05, 2025, 10:00 AM UTC

jauntywundrkind wrote: post #99

stephc_int13 wrote: post #27

Knowing history of past bubbles is only mildly informative. The dotcom bubble was different than the railroads bubble etc. The only thing to keep in mind is that all of this is about business and ROI. Given the colossal investments, even if the companies finances are healthy and not fraudulent, the economic returns have to be unprecedented or there will be a crash. They are all chasing a golden goose.

With Bezos openly stating the goal to build 10 GW+ data-centers in space, it almost feels in question whether this is about ROI, or simply building the Neuromancer future where the ultra-rich can finally be free of their need for the rest of us. Not needing labor at all would be the final return on investment. https://news.ycombinator.com/item?id=45465480 https://www.datacenterdynamics.com/en/news/jeff-bezos-claims..…

Post by lelanthran » Sun, Oct 05, 2025, 3:17 PM UTC

rglullis wrote: post #178

porridgeraisin wrote: post #112

Earlier quoted context omitted.

Inference will be cheapest when run in a shared cloud environment, simply due to the LLMs roofline. Thus, most B2B use cases are likely to be datacenter based, like AWS today. Of course, cern is still going to use their FPGA hyper-optimized for their specific trigger model for the LHC, and apple is gojng to use a specialized low power ASIC running a quantized model for hello Siri, but I meant the majority usecase.

I do not buy this premise. I think it will end up being cheaper to simply run the LLMs directly on the user device. I think that there are plenty of competitors in the "LLMs with open weights" space to essentially make the models a commodity, so all that is left is the compute cost and there is no way that someone will be running a datacenter in a way that is cheaper than "the computer that I already have running on…

Post by justsid » Sun, Oct 05, 2025, 4:46 PM UTC

xnx wrote: post #200

baobun wrote: post #198

Earlier quoted context omitted.

> most important technology transition in the history of humankind Get real. Farming? Plow? Steam engine? Combustion engine? Electricity? Air conditioning? Computers? Internet?

Creating a true digital brain would be humankind's greatest (last?) invention.

Post by keeda » Mon, Oct 06, 2025, 1:58 AM UTC

nickdothutton wrote: post #161

I was there in the middle of the dotcom crash and the telecoms crash which was much worse. Fiber does not rust, and while there was vast overcapacity, not all of it was lit, or indeed worth lighting. 10 years after, thanks to DWDM there were 8 strand cables where only 2 strands were lit, albeit with many more wavelengths than envisaged before. Even though demand had grown. How much is a 10 year old GPU worth? Where i…