33 Comments
User's avatar
The Radical Individualist's avatar

Are data centers even necessary? They seem to be becoming archaic.

Marty Weil's avatar

Speaking as a former long-time Contributing Editor to Data Center Management magazine (back when print media was alive and well), Data centers aren’t archaic. AI hype aside, the compute demand is real and grounded in physics: heat, latency, bandwidth, and proximity to users still matter. Talk of space-based or exotic power ideas solving energy sourcing is currently pie-in-sky (word play intended) thinking and a decade away at least.

My take is that what is actually happening in QTR’s example today isn’t the end of data centers but their fragmentation and specialization. Some AI companies will fail but the infrastructure won’t. Confusing valuation excess with infrastructure obsolescence is risky in my opinion.

The Radical Individualist's avatar

I know there are some really strange ideas about how to produce electricity. My comment was about the processing of information itself.

I would love your input regarding this post. These folks have been saying that there are better, cheaper means to run AI than with a data center. it makes sense, but I'm not capable of assessing it.

https://fractalcomputing.substack.com/p/the-black-swan-event-about-to-hit-c98

Marty Weil's avatar

This would be a great topic for a DCM feature article, if I was still in the trade press racket.

IMO, there’s a real idea in the link you shared, but the conclusions go much further than the evidence supports.

Latency and locality absolutely matter. Anyone who’s built high-throughput systems knows I/O and data movement dominate long before raw CPU does, and purpose-built, hardware-aware software can dramatically outperform general stacks for specific workloads. That part isn’t controversial.

Where this breaks down is the leap from “latency dominates” to “data centers are obsolete.” A data center isn’t just idle CPUs—it exists for redundancy, fault isolation, security, compliance, networking, failover, and coordination at scale. Edge and near-data compute already exist precisely because the industry understands locality, but they complement centralized infrastructure; they don’t eliminate it.

The performance claims also need scrutiny. Hundreds of millions of transactions per second on cigarette-pack hardware at table-lamp power are physically implausible without very narrow definitions and independent benchmarks. Extraordinary claims require architectures and validation, not analogies.

If this is a highly specialized, locality-optimized system that crushes generic stacks for certain workloads, that’s interesting and valuable. But that’s a niche accelerator story—not a Black Swan, and not a replacement for data centers.

george's avatar

What's "Print Media?"

TheSovereignRevolution's avatar

Musk has em moving to space in the next few years. Think Elysium just not humans, yet.

Denis Cowley's avatar

Good catch..never knew Blue Owl existed and if I would have met the president on an airplane and had a conversation that led to me hearing he worked for Blue Owl I would have probably visualized a company that made candy cigars. You made me smarter today. Thanks.

george's avatar

There's a knock-knock joke in there somewhere.

Steve S's avatar

"Heavy capex inflating valuations faster than earnings, accounting choices smoothing over costs, and narratives racing ahead of cash flow are patterns markets have seen before." This observation by QTR brought to mind the early days of Amazon, when some early investors took heat for investing in a company that showed little earnings and expanded beyond cash flow. Amazon kept growing, expanding, and though in the early stages earnings were slow, it soon grew to the point it is today, with abundant earnings. Kinda like a puppy with big feet that later grows into a size justifying its big feet. That seems to be where AI is today, a puppy with big feet.

Luke C's avatar

I’m not sure you have kids…but damn, don’t you feel old when your Seinfeld and Simpsons references return dazed and confused looks on their faces 🤣

Avrelian's avatar

There was a Red Owl grocery store in the Mary Tyler Moore show.

Mick Connors's avatar

Great piece, Chris

K Tucker Andersen's avatar

Thanks for the insights. Totally agree with your conclusions regarding the implications of this Blue Owl announcement. The AI ride in all its various guises will be fascinating to observe over the next few years.

LA's avatar

Fantastic article, written simply and clearly.

Allan Richard Wasem's avatar

Good "catch" Chris :-)

THAT WHICH CAN'T CONTINUE - ISN'T!

george's avatar

unless you see what could be, unburdened by what has been.

Allan Richard Wasem's avatar

To see what could be does not, unfortunately, unburden us from what has been. The consequences of what has been are irrevocable; Entropy moves in one direction only.

HardcoreVeritas's avatar

Sheeple aren't paying attention.

We have lived and thrived all these years without "data centers," and now they are the "number one thing"? All of a sudden the world can't function without them? We need to spend hundreds of billions of dollars and revamp the entire energy edifice and grid to support them?

So we can store data? It has turned into a cult.

Sheeple should be asking, just what is going on here? This IS the digital gulag being rolled out. This is how you will be labeled, classified, and tracked like a UPS package your whole life. We can see it clear as day as people all over Europe are being arrested and charged for what they post on line. Hell, it was 20 years ago that Snowden told us they were watching everything about us.....in complete violation of our right to privacy.

Now they want to store all that information to be used as a weapon of punishment and control. It is coming hard and fast. No stopping it now as the entire government and corporate kingdom is on board.

Personally, I wouldn't invest one red cent......and haven't. No way am I willingly going to help them build the global, digital gulag so that I can make some more shitty FRN......that they will control anyway.

Swiguy2112's avatar

You can go back to 9/11 when Wall Street lost their archives held at World Trade. This started the push to diversity data across the country and to start taking Business Continuity Plans and Privacy seriously when it comes to managing data. Once the need to spread the risk of data loss moved into financial services, that is when data centers became a boom for IBM, Oracle, etc. As storage costs increased due to hard drive limit, you have the birth of the Cloud.

We have gotten more streamlined and efficient, yet at what "real" cost....your point is well taken.

george's avatar

They'd rather train machines than people.

Steve Ross's avatar

It’s fine, nothing to see here. Move along.

Bob Nixon's avatar

Definitely food for thought. Let’s see what MU earnings add to the plot.

Capital Structure's avatar

An alternative narrative could be: private credit firms are acting more rationally and forcing large tech AI capex spenders to take on build-out exposure (resulting in them taking less risk if they can’t as easily keep it off balance sheet). Which would be quite healthy for both industries.

Swiguy2112's avatar

Great point here. Blue Owl has already been undressed with their own plan they "pivoted" from a few weeks ago when they wanted to merge the private vehicle into the public fund. They have reminded me of Wachovia, Countrywide, Conseco, etc. Firms of the past who were writing loans with their eyes closed.

Mark Heywood's avatar

Another great article, glad we have you to spot these deals falling apart. The circle trade BS was the first fuckery. The AI bubble may have just started to pop.

Swiguy2112's avatar

The situation of all of these firms ski polling each other is real. As long as the assumption that usage will continue to increase all is fine. If you haven't already been tracked by your employer on your usage metrics of all of these tools, then you are lucky for now.

The days of you not being impacted by AI is over. Still trying to find out who asked for all of this as this is a real distraction from something under the hood.

Gehenna's avatar

At the end of January, this article's prescience is well validated.

TA's avatar

“ AI, credit, and private equity are intertwined. AI narratives justify massive spending. Credit markets fund it. Private equity absorbs and redistributes risk. Accounting keeps the optics clean. Equity markets reward the story. That works until it doesn’t.”

Very helpful explanation of the linkages.