# data center revenue — X 热门讨论 (2026-10-09 08:03 UTC)
## @rogueiq (CK) · 10-08 14:59 · ♥50 ↻5 💬4 Two legendary investors just said the AI bubble is close to popping
Ray Dalio: AI is a "classic bubble." "I think we're close to that."
Michael Burry: "The bubble in AI may burst sooner than later."
Meanwhile, the market disagrees and keeps going for ATHs
Who's right?
Nobody knows. Not Dalio. Not Burry. Not the guy on X with his charts. Certainly not CNBC's "experts."
The good news: you don't need to predict. You need to be positioned.
3 shifts that helped me a ton, after plenty of mistakes:
1. Stop predicting.
Be positioned so a burst doesn't break you, and a continued rise keeps you in without FOMO.
2. Spread your AI bets.
Know what gets hit hardest vs least, then pick what you want to own:
Hit hardest - Story stocks with no earnings underneath (pre-revenue AI, quantum, small-cap "AI pivots") - Companies borrowing against GPUs that age out in a few years, with one or two big customers - Commodity parts like memory chips, sitting at record margins right now - Gear sold project by project (optical, cooling, power equipment). Orders vanish when builds pause
Hit hard, but survive - AI chip designers. Hugely profitable, but most revenue rides on AI spending. Nvidia fell about two-thirds in 2022 - Networking gear and data center landlords
Hold up best - Chip equipment monopolies and grid equipment, where AI is part of the demand, not all of it - The one foundry every chip has to go through - Big platforms funding AI from cash (cloud, ads, software). For these, AI is a cost they can cut, not their revenue - Regulated utilities. Regulated returns, AI is the bonus
3. Think long term.
Staying in the game is what gets you to your money goals.
What lasts owns its own customers, has a cash engine outside AI, sits at a chokepoint, or earns regulated returns.
What gets wiped out carries debt against a fast-aging asset, sells a commodity, or depends on a couple of customers who aren't profitable yet.
You don't need to know when it pops.
You only need to know what you own if it does. https://x.com/rogueiq/status/2108210803941707886
## @IREN_Bull (IREN Bull) · 10-08 05:38 · ♥49 ↻5 💬3 Kent Draper- $IREN COO - gives a deep dive into IREN, customers, financing, growth and much more:
So, we're a vertically integrated AI cloud company…. we provide compute to end users, and this includes leading names in the industry like Perplexity, Cohere, Prometheus, Microsoft, Nvidia and we do that from our own data center facility.
So, that is the the difference with most of our competitors is we actually own the underlying physical infrastructure that sits behind the compute layer that we provide.
We've got a a large roster of the leading names in the industry. One of the nice things in this industry is demand is not a problem. There is multiples of demand for every bit of capacity that we can bring online, but strategically, we're spending a lot of time thinking about where we want to be 5 to 10 years from now, diversifying our customer base and making sure that we have a really strong ecosystem of end users within within the AI sphere.
We're in a whole new paradigm here where AI is going to be a totally different type of compute to traditional cloud computing. And that actually means that there is room for new hyperscalers within this industry. Andso our thesis, you know, going back many years now was that the digital world is going to grow exponentially, the physical world wouldn't be able to keep up with it. We've built a very large pipeline of land and secured power and are now building out the data centers to to service the compute needs going forward.
And that gives us the ability to become a new hyperscaler in the AI space. So that ultimately is is the target.
And one of the key things we keep saying is execution in this area is really hard. Land, power, supply chains, labor.
And it is getting harder and harder as we move forward. You know, when the industry went from 10 gigawatts to 20 gigawatts, that's one thing. When you're going from 40 gigawatts to 80 gigawatts in a single year, as forecast to happen, that is going to be extremely challenging.
So execution is really key and that's something that we pride ourselves on.
Financing markets are developing extremely quickly. I think historically, if you had investment grade customers, there's been a well-developed market for that and we've seen very strong financing products in that area.
Now, with our new customers we're signing, they are sub-investment grade. But we've seen in just a matter of months that market really develop.
And we were able to obtain for our most recent set of customer contracts, 90% of the GPU CAPEX in debt financing. And in addition to that, our customers are actually giving us prepayments for 50% of the GPU CAPEX. So, we're actually funding between debt financing and customer prepayments, more than 100% of the GPU CAPEX.
So, you know, that's a great position to be in and we expect to see financing markets to continue to develop over time.
How durable is AI demand?
I think incredibly durable. I mean, one one of the things we've seen particularly over the last 18 months is a real emergence of real world use cases. So, if you rewind in this industry, the biggest drivers have been training runs for large model development.
But increasingly, we're seeing manufacturing, R&D, drug discovery. Obviously, all enterprises are incorporating it into their daily use cases, whether it's internal or customer facing. And that is driving a significant level of demand that is really tangible and we think going to be very durable as we move forward.
What's next for Iren?
Executing as I mentioned on our existing pipeline. We have a very large portfolio of secured land and power. And as soon as you bring this capacity online, we will have a revenue line. So our focus in the near term, very much on execution, but we continue to add additional sites behind our existing ones to give us that runway for continued growth over time. https://x.com/IREN_Bull/status/2108069425341444320
## @Abbycadabby87 (Abby) · 10-08 17:43 · ♥33 ↻3 💬0 Among these neoclouds, $NBIS is clearly further along on going global.
In the US it has Kansas City. Independence, Missouri is where the first GW-scale AI factory is moving.
Europe is more built out. Finland 75MW, Lappeenranta 310MW, Lille in France 240MW, and four UK sites at about 65MW combined.
I noticed the playbook changing. Partners put up the money for the data center and the GPUs. $NBIS brings the architecture, the software stack, the supply chain, and the customers.
That means someone else takes part of the hit. Expansion and capital efficiency should both get better.
$IREN, the one I like, is still mostly a North America story.
Power and land aren’t the shortage. A real global ramp probably waits until after 2028.
So the next job for IREN is turning the power edge into data-center and GPU revenue faster, including outside the US. https://x.com/Abbycadabby87/status/2108251879004086464