If you build it, they will come – popular idiom derived from “Field of Dreams”
Fifteen years ago, the world was touting the success of China’s real estate boom. Pundits lauded the “Chinese Way” that allowed for rapid progress where similar projects would take years in the West. Western fund managers moved capital to China because real estate collateral in China was deemed extremely safe.
Fast forward ten years and Chinese real estate is a never-ending meme that has destroyed trillions of dollars of value, China’s middle-class, and effectively bankrupted local governments that relied heavily on land sales. In typical Chinese Communist Party (CCP) fashion, a successful idea was taken to extremes.
But let’s be clear, the CCP didn’t invent asset bubbles, they copied them from the West – and from Japan just a generation ago. The bubbles in the US – housing, commercial real estate, stocks, private equity, etc. have been blown in a more controlled fashion, unlike the insanity of China over the past 30 years. But the results will be the same – eventually.
The AI boom is eerily similar to China’s real estate boom because it is driven by “top-down” motivations rather than bottoms-up demand. I write this with full understanding of how Chinese families were buying multiple apartments, levering themselves to the fullest in order to purchase newly built properties. The thing most people don’t realize is that the pricing boom was actually cultivated by regional Chinese government.
In the middle of the boom, Chinese regional governments effectively set the price of new apartments based on how much money they needed from the underlying land sales (leases) to match the needs for local government revenue. Yes, it sounds completely ridiculous but when you consider that Chinese savers had comparatively few options, it was possible for a time.
Ultimately, the underlying market was unable to generate sufficient cash flow to maintain these exorbitant prices. Chinese property investment declined 16% in May y-o-y as the decline in investment continues to accelerate. Regional governments in China are effectively bankrupt, cutting services and extracting revenue from citizens in any way they can.
Compare this to AI – nobody is really clamoring for AI apart from a relatively few people – it’s a demand-push story as opposed to a demand-pull story. Build it and they will come except that’s not what appears to be happening. Communities don’t want data centers, nobody likes Windows 11 or Co-Pilot, and neither ChatGPT nor Anthropic can price their models high enough to cover their costs. AI is not a function of the market clamoring for a product; it’s a function of large companies and government attempting to force it down our throats. The distinction is critical.
Deflation
In the same manner if any nation wasted part of its wealth, or lost part of its trade, it could not retain the same quantity of circulating medium which it before possessed – David Ricardo
This quote by David Ricardo is my favorite quote from the economic classics. People who have blinders on when it comes to inflation would be well-served if they gave the quote full consideration and extending it to understand how it works in modern financial systems.
No amount of “money printing” can save bad investments in the long run. The People’s Bank of China has been quietly going “Weimar” in its banking system for six years yet real estate – homes, factories, and retail – all continue to decline in value. In 1950, the city of Detroit was widely considered the wealthiest city in the US – 50 years later, that wealth was gone despite monetary expansion since 1970.
The investment case for data centers is absurdly ugly. If I want to build a data center today, I have to pay wildly inflated prices for electricity, GPU’s, RAM memory, HVAC equipment, on top of real estate. How can I make money in this business when my sunk costs are exorbitant? Nobody can which is why 50% of announced data center investments have already been cancelled or delayed and 1/3rd actually under construction. And even when constructed, there is still the question of electricity for the new data centers.
The cure for high prices is high prices. Companies that are critical for stock index expansion such as Nvidia, Micron, ASML, and TSMC are all reporting great earnings numbers even though their customers are languishing. Worse, the private credit that funded this recent round of data center investment is already “gating” withdrawals. The bagholders are the people who reached for higher returns by investing in garbage with poor fundamentals.
Given the number of AI data centers that are actually in use versus the number of AI data centers in some part of development is extremely small. This tells us that all those Nvidia GPU’s and Micron RAM chips are actually sitting in storage somewhere but are not actually generating revenue – simple math.
This is actually analogous to Chinese real estate developers such Country Garden and Evergrande who took deposits, used them to fund operations and in many cases, failed to deliver finished apartments. Private credit is heavily invested in data centers and yet, these projects are not getting completed and are not generating revenue. Instead, that cash flow has been used to prop up the earnings of Nvidia and Micron – for now.
There appears to be a significant number of GPU’s and RAM chips sitting in limbo even as the prices for new orders remain elevated. How long until those these chips hit the market due to forced liquidation? If I had to guess, it won’t happen until after the mid-term elections but the reality appears to be a situation similar to the moments before the Chinese real estate meltdown…
Enron
What Enron was doing, what caused investors to embrace it in a rapture of baffled awe, was hiding debt – P.J. O’Rourke
We don’t know who owns the majority of data centers that make up “the Cloud.” It appears to be a combination of SPV’s or special purpose vehicles and private equity. To the best I can tell, these SPV’s appear to be a partnership between large technology companies and private credit, supplied by asset managers such as Blue Owl.
If we look at Meta’s Hyperion data center campus in Louisiana, it is 80% owned by Blue Owl Capital and 20% owned by Meta but with Meta fully controlling operating decisions. Meta’s operating control should result in prudent financial analysts consolidating this project on Meta’s balance sheet and treating Blue Owl’s stake as secured lending but that’s not what’s happening.
The 20% equity threshold keeps the project off Meta’s books but the full operating control tells us that it represents an operating unit of Meta’s. This is where accounting policies fail us.
Applying the cockroach theory, how many SPV’s are actually off-balance sheet financing vehicles for major technology companies? What would their financial statements actually look like if consolidated under the prudent man rule? In effect, Meta’s shareholders socialized the credit risk to a third-party fund but still retain the operating risk associated with dislocation connected to financial disruption at data centers.
Which brings me to Enron. In 1998, I covered the energy sector and the partners at my firm tasked me with the job of giving the firm a reason to purchase Enron for clients. After days of analysis, I couldn’t understand how Enron made money given their aggressive use of SPV’s. My conclusion was that I could not recommend Enron in good conscience.
Keep in mind that it took three years from when I wrote that research report on Enron in 1998 until things fell apart. Not coincidently, I had moved to another firm by that time…
George Hudson
When running a Ponzi scheme, how does one avoid enormous, unexpected withdrawals – runs on the bank, so to speak – that would pull back the curtain and reveal a little man blowing smoke? One way would be to attract a core of investors who could be counted on to never withdraw more than a small percentage of principal each year – Mitchell Zukoff
George Hudson was the “Railway King” in Great Britain during the 19th century. He controlled around 1/3rd of British railways at his peak and yet, like Enron, his empire collapsed in 1849 due to financial fraud. The fraud, which was paying dividends out of capital, not earnings, appears tame by today’s standards.
Fairly or unfairly, I personally view people such as Zuckerberg, Musk, Jensen, and Bezos as scions of George Hudson in the figurative sense. If I’m right about the impact of SPV’s on their operations, we need only watch the problems in the private credit market for indications of when this storybook romance between the market and hyper-scalers ends. As some have already noted, the current strategy seems to be to roll unwary retail investors into this crap so we may be waiting for some time before it ends.
I don’t believe these men are destined to remain at the top for much longer but it is critical to understand that maintaining the value of stock indexes is vital for our national economic well-being. Our nation’s savings are invested in these companies, effectively making them “too big to fail” for now.
I view this situation as representing a “stay of execution” and not a permanent pass to destroy capital as they invest in artificial intelligence. Furthermore, I view buying short positions on their companies as traps – for now. The market is structured to perpetually squeeze short positions on these vital national interests.
Conclusion
A man who has bought a theory will fight a vigorous rearguard action against the facts – Joseph Alsop
It will take something enormous to shake confidence in the stock market enough to spark a major decline before the election. But that doesn’t mean that it’s smooth sailing for the Mag 7 stocks and AI in general. The negative year-to-date returns in Microsoft, Meta, Tesla, and Oracle are sending a loud message along with the modest failures of the Cerebras and SpaceX IPO’s. People are waking up slowly.
The story about Nvidia GPU’s and Micron RAM chips sitting in storage and not lit-up has been known for at least nine months. Increasingly, more and more people are realizing that using AI does not generate a positive ROI in the collective. The problems with private credit’s investment into data centers is not going away and will need to be managed through the mid-term elections.
In general, while unpopular, the belief that AI is our future is still relatively intact and that’s all that matters for now. The US government is still fully behind attaining a Pyrrhic victory in the AI race with China, even as the US runs the full race-course and China takes shortcuts at every turn. I believe the “victory” will ultimately destroy the capital structures of the Mag 7 stocks and the savers in the private credit markets.
I don’t attend crazy parties where people lose control and I certainly don’t advocate such behavior in the capital markets. The over-reliance of a few stocks is similar to the year 2000. The good news is that there are plenty of places in the stock market that have gotten cheaper as everyone has jumped on the AI boat. These stocks represent the life rafts when the AI boat ultimately tips over whether it’s this week or next year.
The financialization cycle is over after 125 years. The future is about re-building a destroyed industrial base, not about trading crypto and Mag 7 stocks. If you’re interested in learning more, visit us at https://geovestadvisors.com/ and give us a call.
Philip M. Byrne, CFA
