The real problem with data centers: Their buildout costs are driving up interest rates.
This is now as much a drag on the bond market as is the government.
I wonder how he Chinese are dealing with this?
Back to threaded view
Full conversation · 24 posts
-
Level 1
Socialized risk & cost with privatized gain. Pass.
(no message)Dead-eye Pete with 20 pounds of swingin' meat. -
Level 1
I think the proper way to think about data centers is like utilities or infrastructure.
(no message)-
Level 2
I think the proper way to think about data centers is how much useful AI we can make per $, per kwh
From these data centers.
-
Level 3
How much energy is spent on cat videos and other silliness?
AI at least has some utilitarian value. How much energy do we spend on entertainment? How much energy do we waste out of convenience, laziness or apathy? We should be advancing nuclear power in parallel and this criticism goes away.
Interesting stat I heard about the water usage, one of the viral criticisms lately, all of the data centers in the US use less than 3% of the water that all of the golf courses use. Yet I've never heard anyone argue we shouldn't be putting golf courses in the desert because of the water usage.
-
-
-
Level 1
Chinese don’t have this issue. China and U.S. have different routes in developing AI.
China doesn’t pursue computing power, therefore has very low cost in AI infrastructures such as data centers and energy facilities related to them.
AI business in U.S. is like Amazon, NFL, i.e. they are building an empire. AI business in China is like Alibaba, College football, i.e., open source, grassroots, integrated with real economy.
-
Level 1
And the burgeoning use of Credit Default Swaps by Hedge Funds isn't music to the ears of "Free Cash
Flow" adherents either...memories of the "Great Recession" are coming back.
Here's an AI summary (no pun intended)...
----------------------
AI OverviewYes, credit default swaps (CDS) are heavily being used in relation to AI capital expenditure (capex) spending, and they are signaling that financial risk is rising.
However, tech companies are not using CDS to buy chips or build data centers. Instead, bond investors and hedge funds are buying CDS as insurance policies to hedge against the massive debt tech companies are taking on to fund their AI infrastructure.Because major tech "hyperscalers" are aggressively leveraging themselves to fund AI capex, demand for these derivatives has surged, causing AI-tied CDS spreads to spike.
Why the Demand for AI-Linked CDS is Surging
Historically, mega-cap tech giants were viewed as impenetrable financial fortresses with massive cash piles and little to no debt. The AI arms race has radically flipped this reality:The AI Debt Wave: According to reports from Goldman Sachs, AI-related debt issuance has reached nearly $500 billion, with companies like Alphabet, Amazon, Meta, and Oracle issuing hundreds of billions in bonds to build infrastructure. Nvidia even tapped the bond markets for the first time.
Outpacing Cash Flow: Hyperscalers are on track to collectively spend more on capex than they generate in free cash flow. For companies like Oracle, projected capex is roughly double its operating cash flow.The Need for Hedges: Fixed-income investors who hold these bonds are using Credit Default Swaps to protect themselves against the possibility that these massive investments fail to pay off.
Is This Trend Increasing Financial Risk? Yes, the rising cost of CDS (widening spreads) is a direct indication that credit risk in the technology sector is increasing. While analysts emphasize that an immediate actual default by a major tech giant remains very low, several structural risks are building up:
1. Mismatched Asset Lifespans
A primary credit issue is asset-liability matching. Land and buildings for data centers can support long-term debt, but the highly expensive GPUs inside them have short technological lifespans. If the chips must be replaced and upgraded before the underlying debt matures, companies are forced into a relentless cycle of recurring capex.2. The Return on Investment (ROI) Timeline
Bond markets are pricing in heightened risk because they doubt whether AI commercial adoption can generate cash returns fast enough to service the ballooning debt. Even block-buster earnings are failing to calm fears about the long-term sustainability of AI revenue models.3. Rising Costs of Corporate Capital
Because investors are demanding higher yields to take on this AI debt, borrowing costs are surging across the industry. As noted via Apollo Global Management, bond yields for data-center debt have jumped significantly. If capital becomes too expensive, it may eventually force the AI capex cycle to self-throttle
------------------
Level 2
Just as with all the subprime mortgages at the time of the 2008 crash, there shall be serious losers
The funny thing about all this is that the folks pushing all this are the most intelligent and well educated amongst us... but they are compulsive slaves to "creative" financing while blind to the potentially catastrophic consequences.
It's as bad as our politicians legislating unlimited spending just to get elected.
-
Level 3
And most egregious are those who seek $1.5T Defense Expenditures, while giving Tax Breaks to those
who have no NEED for them. Looking at only one side of the ledger is simply unmitigated Greed...
Never forget that in recent memory, the only administration to produce a Surplus and Deficit Reduction was (D) Bill Clinton's...after much bipartisan work that respected Unions and Government workers' well-being. There is a Right Way to solve this problem...but 'Greed' needs to be left outside the meeting room. This is all documented fact...if you continue down your story's path, expect to be reminded.
-
-
-
Level 1
Good grief - stop already
Minimal - if that - impact on interest rates. That one month old article is a failed attempt to tie company-specific debt to interest rates. The old, boring reasons - federal deficits, Fed policy, inflation expectations, etc - still drive rates.
Data centers are the new boogeyman. No doubt we'll soon learn they are also racist and probably transphobic, too.
'I define fear as standing across from Joe Louis and knowing he wants to go home early.' - Max Baer-
Level 2
I tend to agree. I also look at who is against data centers - that always help clarify things when
BaronVonZemo 64373 postsit's otherwise murky.
-
Level 2
I'm still trying to figure out the fear of datacenters.
I suppose people freaked when Ford was first building factories??? Or, when the Jacquard Loom was replacing people in textile factories?
AI in general is more fearful than datacenters. If the fear is of AI, I might understand it. (And, I'm not talking about AI merely taking some jobs away.)
Evil preaches tolerance until it is dominant, then it tries to silence good.-
Level 3
So, data centers = AI.
Compute is the limiting factor. The race between the US and China is a race for compute because the more compute you throw at AI the smarter it gets. There's more involved, but this is a simplified way to look at it.
-
Level 4
So, stopping AI is as simple as stopping data centers. Good to know.
(no message)Evil preaches tolerance until it is dominant, then it tries to silence good.
-
-
Level 3
The objections are rooted in the distain that datacenter builders have for local communities...odd
you never sought to use the product of such centers to help you understand...;-)
Here's one AI Summary of the situation...
----------------AI Overview
Roughly 70% of Americans oppose building new data centers in their communities due to fears over soaring electricity bills, massive water consumption for cooling, persistent industrial noise, and the perception that multi-billion-dollar tech corporations offer few permanent local benefits while burdening municipal infrastructure.USA Today
Core Environmental and Resource Concerns
Grid Strain: Enormous power demands from servers and AI training risk destabilizing local power grids and driving up monthly electricity rates for everyday residents.Gallup News
Water Consumption: Facilities require millions of gallons daily for evaporative cooling, raising alarms in drought-prone areas or small municipal water districts.Facebook
Noise Pollution: Round-the-clock humming from cooling fans, chillers, and backup diesel generators disrupts nearby neighborhoods and sleep quality.The Conversation
Economic and Quality-of-Life Discontent
Sparsely Created Jobs: While construction phases bring temporary jobs, completed facilities require very few permanent, full-time local workers.Harvard Gazette
Opaque Tax Deals: Tech companies often secure heavy local tax breaks, leaving residents feeling that they shoulder the environmental costs without receiving proportional tax windfalls.YouTube
·PBS NewsHour
Land Use Changes: Sprawling warehouse-style footprints permanently alter rural farmlands, quiet suburbs, and local community aesthetics
---------------- -
Level 3
There are Luddites in every generation. The bigger problem is that so many do not fear debt.
Maybe one reason our younger generation leans socialist is that we boomers are saddling their future with ginormous debt for them to pay.
-
Level 4
Yes. And anyone who looks at the issue knows we can only reduce debt by inflating the dollar.
So, if they've thought about it, they know it will get worse for them unless they are pulling their savings out of cash and putting it in inflation proof forms (e.g., stocks, commodities, possibly real estate).
And if they don't have savings at all, then they need to steal from those who do.
So, there will be pressure to select socialism as a short term fix (redistribute wealth) to a long term problem (diminishing wealth and increasing national debt). Socialism just replaces who wins: capitalism rewards merit; socialism rewards those in power.
Evil preaches tolerance until it is dominant, then it tries to silence good.-
Level 5
Without AI I think we'd be in a recession or perhaps a depression.
And it may be the thing that saved us, but I have some doubts.
-
Level 6
At least as the stock market measures those things.
(no message)Evil preaches tolerance until it is dominant, then it tries to silence good.
-
-
-
-
-
Level 2
A mere $1.4 Trillion... can't possible impact credit markets. Right?
And I am pro-datacenter. I just feel we need to anticipate the larger impact.
Some DSA members (El-Sayed) are already arguing that we need government control of the whole data center/AI industry.
That's a first step toward rule by progressive elites.
Link: https://tech-insider.org/us-utility-1-4-trillion-ai-data-center-energy-2026/
-
Level 3
That's a bit misleading
The $1.4T is a projection. Regardless of what the figure actually is, a significant portion will be funded by the hyperscalers' cash flow, with more still funded by Nvidia. The amount subject to be financed - while still a huge number - is far less than the scary headline number. I still argue that the buildout will have minimal impact on interest rates.
'I define fear as standing across from Joe Louis and knowing he wants to go home early.' - Max Baer-
Level 4
Long term I totally agree.
But in the near term... our borrowing for the AI buildout is an important factor in the rise of interest rates,
-
-
-
End of conversation