AI frenzy continues as Macquarie commits up to $5B for Applied Digital datacenters Fears of an AI bubble have yet to scare off venture capitalists and private equity firms from pumping billions of dollars into the GPU-packed datacenters at the heart of the machine-learning craze. Case in point, on Tuesday Macquarie, one of Australia's top financial services firms, revealed plans to inject up to US$5 billion …
This feels like a larger-scale repeat of the early Bitcoin boom years, where manufacturers were falling over themselves to produce datacenters and warehouses full of ASICs that are now completely unusable electronic waste.
At least the GPU shovel salesmen for this gold rush fad are making out like bandits. It's still mind blowing to me that any sincere C-leve genuinely expects a ROI on actually _using_ AI, rather than just selling the faff associated with it.
Boom and bust may be a thing in this space as you've pointed out.
There's money to be made for sure, but not at the coal face of AI, but all the subsidiary supporting services and the hardware.
I've not seen tangible output in any organisation that would act as a whitepaper/blueprint for how to spend millions on AI to make XXXX profit or YYYY savings.
Let's see where we're at in 5 or 10 years.
On the plus side if an AI bubble bursts there is going to be lots of compute available for cheap. Guessing a timeline on that is hard, but lets say 3 to 4 years*. So right now any contracts for colo or cloud on a 3 to 5 year horizon are unlikely to have dramatic price changes. But if I was renewing a contract in 3 years time, I would be evaluating future compute costs pretty hard before committing to any contract.
as for investing, 3 fold return in 4 years is a pretty big upside, enough that the investors are almost guaranteed to break even and given they measure success by beating the index by a few bps (100th of a %) , plenty of portfolio managers are only constrained from plunging in head first by their Investment Management Agreements with their clients and agreed risk profiles.
I was holding off on any more NVIDIA stock but with announcements like that probably** safe to buy and hold for another 2 years.
* even if the bubble bursts next year there are a lot of contracts running that won't expire immediately, there will be plenty of companies who don't believe it and continue for a while.
** not financial advice, my track record in picking stocks sucks.
"as for investing, 3 fold return in 4 years is a pretty big upside, enough that the investors are almost guaranteed to break even and given they measure success by beating the index by a few bps"
Other than Nvidia, where did you get that figure from?
Are you suggesting that there is going to be OVER $600 billion ROI from these investments, given that Anthropic and OpenAI are making obscene losses, only kept going by venture capital funding.
They have no plans on how to make money, other than "you can do stuff"?
I took the 3 fold return number from the article "According to our sibling site The Next Platform investing $1.5 billion in infrastructure — enough to build and network a datacenter with roughly 16,000 H100-class GPUs to rent out — will net you about $5.27 billion in revenues within four years, assuming a decent mix of spot instances and long-term commitments from customers."
Wow, well that didn't take long for me to be proven wrong :) https://www.nextplatform.com/2025/01/27/how-did-deepseek-train-its-ai-model-on-a-lot-less-and-crippled-hardware/?td=rt-3a
12 days from prediction to proven wrong, damm near a new record for me. Maybe I should make more predictions on stocks and then others can make money by buying exactly the opposite.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
