The hope and the hype
Where do I sign up?
Joe Calhoun of Alhambra Investments says something I’ve wanted to say myself but hadn’t found the right words for.
Calhoun has.
Information technology has long promised a productivity miracle, but it’s turned out to eat up almost as much productivity as it adds. Constantly learning and applying new tech takes a lot of time and effort and means you never really have a chance to get good at it. Calhoun says:
“Technology has jumped the shark; today, it often makes life more complicated rather than less. Endless two-factor authentication codes, dynamic pricing, a constant flood of spam and phishing attempts (which have now invaded our calendars), app-only discounts, subscriptions for everything, embedded vehicle tracking, endless customer service bot loops (OMG, how do I get a human on the phone?), QR code menus, self-check-ins, and grocery self-checkouts. Technology has enabled companies to become more efficient but at their customers’ expense.”
To this we could add the endless app updates that don’t really improve anything and the proliferating channels of communications, from mail to telephone to email to voicemail to text messaging, etc. All this is externalizing costs, much as Big Tech seeks to do with its privacy appropriation and energy demands. And it’s going to have to extract a lot more from me and thee to justify its market valuations. It’s growing so big and pervasive it’s as if we’re growing another government.
I could quote much more but you will want to read the article for yourself.
More on the investment angle …
Michael Burry’s bearish view. The market is in ‘denial’ about the end of the AI trade.
I just want out: Billionaire investor Jeffrey Gundlach, who called the housing bust in 2007, says you should have almost nothing in AI stocks
One nit to pick. IBD frames this as a “pessimistic” view. I see it as the opposite. Bubbles are bad for the economy, and the longer they go on the worse. An implosion of the AI bubble, while transiently difficult, would ultimately lead to better living standards than its continuance. There’s only just so much real capital available, and the AI bubble is hoovering it up, leaving housing, transportation, education, medicine, etcetera wanting. Deflation of the bubble would take emotion out of it and lead to a more rational balance of capital priorities.
“The S&P 2”: Micron And Nvidia Alone Will Deliver A Third Of Q3 Earnings Growth As Median Stock Barely Grows
Robert Solow, some time in the ’90s, famously said “I see computers everywhere but in the productivity numbers.” It takes time to reorganize work to take advantage of really new tech. I expect a crash before we get there, likely emanating from credit.
It’s always the debt. Until recently, most of “hyperscaler” capital spending was financed with organic cash flows. Now it’s dipping into borrowing – hundreds of billions of it. At some point rock meets hard place.
Stock buyback scam hits new records
“The AI-chip giant Nvidia (Nasdaq:NVDA) announced this week that it was launching the biggest share buyback in U.S. history, at $150 billion, taking its authorized buybacks to $235 billion by January 2028”
“For tech companies like NVDA, the fairly feeble arguments for stock buybacks by mature industry corporations do not apply. Far from having excess cash flow, tech companies must invest gigantic amounts of money to keep up with their competition and markets in a rapidly expanding field.”
And reinvestments needed just to maintain competitive position aren’t true profits, they’re a cost of doing business. Besides the disproportionate benefit to insiders from share buybacks, this is another reason why dividends are relatively low. Dividends tend to reflect the earnings a shareholder can sustainably expect.
“The SEC should repeal Rule 10b-18, adopted in 1982, which “clarified” the stock buyback rules in favor of greedy management and against ordinary shareholders.”
Will AI Really Raise Productivity?
“And it’s going to have to extract a lot more from me and thee to justify its market valuations.”
We are witnessing the stock market do something for only the 3rd time in 156 years
“A recent Wall Street Journal study found that American businesses and consumers would have to spend the equivalent of 8.8% of the country’s GDP annually on AI to justify the industry’s immense spending. This sounds like a tall order, which raises the possibility that many current data center investments will never pay off.”
The Financial Times reports that OpenAI’s revenues fell about $20B short of previous expectations. OpenAI’s stock doesn’t publicly trade, but AI related stocks like Nvidia and Micron are down sharply.
Your Stake In The Biggest Economic Bet In US History
“Economist Stijn Van Nieuwerburgh estimates investment in data centers, power systems, networking infrastructure, specialized chips and other AI-related equipment will total an astonishing $10.3 trillion from 2025 through 2032. That works out to an average of 3.63% of U.S. GDP every year.”
So let’s see … it’s spending a whopping 3.63% of US GDP and would require 8.8% of GDP for it to pay off.
This is bigger than railroads, electrification and highways put together.
Whether you live in America or Timbuktu, there’s no way you won’t be affected.
As a young engineer I worked with an old guy near retirement. I was gushing about how great my PC was for working. He smiled and asked a question: ” Will computers do for engineering what typewriters did for poetry?”
Around 1980 or so I was an engineering student learning to use a mainframe to run FORTRAN and run CADCAM on desktop workstations. We enjoyed it. We didn’t have a home PC until the early or mid 1990s. Even the old i386 was pretty capable for doing actual computing. PCs started to slow down a lot though once we started connecting them to the internet.
At first it was great. But soon you had to start updating regularly to maintain internet compatibility. And every new software iteration slowed your machine down and the skills you developed became obsolete. Having to learn and relearn sapped most of the productivity benefit.
Having tried later versions, I’m still using Excel 95 to run Synthetic Systems. Another desktop is a Mac Pro I use for music. I’ve kept it unplugged from the internet and it’s just as fast as the day I set it up over a decade ago. Few get to appreciate the benefit of not having the constant drain of overhead just to keep up and being able to get really good at using certain operating systems and applications.
One worry I have about AI is that it will completely displace traditional computing. We’re already seeing AI features being added everywhere they can fit and some places they don’t. That would be a loss inasmuch as it’s very useful to have machines with a stable and defined input-output relationship. You can perform a given task on a given set of data and get the same reliable answer each time you do it. Losing that will result in unanticipated consequences that won’t be fully appreciated until they’re gone, but it’s apt to happen anyway.
I’ve had a similar experience in late 80s early 90s, as a Computer Science & Engineering student working on 8086 chips PCs, minis and later mainframes running COBOL/JCL. Motorola 68000 was the state of the art, Intel’s architecture was in its early days. We could do a lot of customisation on the DOS operating system given it was a simple aarchitecture. A 286 PC had its own enclosure within the computer lab, with access strictly controlled by the staff.
Agree with you that the emergence of the Internet resulted in slowing software down with regular updates being pushed through that resulted in bloatware. And as you say, productivity gains were minimal due to the need for training and constant updates.
You make an important point on how AI could impact traditional computing. The predictability of output when the inputs and process are unchanged is lost with AI. Having vibe coded software to help with my day job, I’ve been able to observe this first hand. Additional prompts help establish guardrails but one cannot anticipate all eventualities that need to be mitigated. So, there’s always the uncertainty factor to consider.
I don’t want to sound like I’m down on the potential for AI. One use case I haven’t heard about but which seems well within reach is in audio and visual processing. Say, taking an old scratchy vinyl LP and removing the noise while improving the frequency response and dynamic range. Or one that was recorded before they learned how to do stereo and just panned everything hard left and right. Same with old movies and TV shows … convert 4:3 aspect to modern 16:9 or whatever to fill the screen without cutting off the top and bottom. Just a couple areas traditional signal processing couldn’t quite do right. Not to mention language processing … I’ve had some amazing conversations with even some of the free AI bots out there, especially or even on scientific and technical matters.
Besides the financial excess, it’s the internet precedent that tempers my enthusiasm. Not only is so much investment being put into AI that other human needs go wanting, but the internet, which was supposed to set information free and democratize media, wound up being controlled by a handful of supersize corporations. We could do so much better than Facebook and TikTok and Deep Fakes, but that’s where we are. So far AI looks headed down the same low road. I hope I’m wrong.
There are very troubling parallels between the AI bubble and the TMT/dot-com bubble. At the center of the TMT bubble was Cisco and its vendor financing of its customers that ultimately blew up. For the AI bubble, the role of Cisco is taken by NVIDIA and they’re also doing a lot of circular accounting tricks.
I suppose the difference between the two bubbles is that the AI bubble is the TMT bubble on steroids. In another topic, I commented that annuities are very difficult to evaluate, at least for me, because it’s difficult to determine the financial health of the company selling the annuity. My understanding is that a lot of the AI bubble debt has been bought up by insurance or reinsurance companies. The leverage levels I’m seeing are extremely high although perhaps that is normal for insurance companies. However, with high leverage, and sort of loss on assets is greatly amplified and I think the AI debt is at much higher risk of default that believed.
Very much so. Far too much of these highly touted earnings are highly questionable. I’m not saying Nvidia is about to become an Enron or Global Crossing, but am not loaded to the gills with it either. Not to mention I am not one of those waiting in breathless anticipation of an OpenAI IPO.
I agree with your point about annuities, only would point out that there is no investment that is foolproof. Each has its uncertainties and can’t realistically be compared to an ideal that does not exist. The traditional pensions that millions relied on for decades (and some still do) also depended on the financial condition of the issuer. Even federal pensions and Social Security – also annuities – depend on an increasingly sketchy entity. Even the dollar itself. Insurance company soundness issues are not unique to annuities either; they extend to your home insurance policy, your car insurance and your life insurance too. My answer is classic due diligence and diversification … you allocate your investments between multiple issuers and include hard assets like gold and other commodities in the mix; as always use care in selection and tailor the mix to your own individual circumstances. The perspective of my previous post is a timeless one about the principles of retirement finance, not specific to the current economic environment. Commercial annuities are a good fit for some people, and while investing in something that would best serve the purpose may be taking a risk, putting the same money into something that doesn’t would be taking a guaranteed loss.
As for the present, I also agree that AI debt is at risk of default. It’s unsettling that companies that have traditionally sported strong balance sheets are now going deep into debt, and with a dubious payoff at that. When value and quality are both in question, it’s a red as any flag gets.
We don’t know with certainty the outcome of any investment. The one thing that every investor does know is his own personal circumstances. The art of investing boils down to assessing the likely and even the possible futures of each investment on the menu and matrix them in with the known objectives of the investor … put all these things into the hopper and turn the crank!
AI’s Cash Is Missing, and the Credit Risk Is Not Priced In
“Scale sharpens rather than softens the exposure. The outlay is roughly 17 times the dot-com investment and four times the subprime housing bubble.”
Companies are realizing they’re spending a fortune on AI just so their employees can generate error-riddled slop, and suddenly cutting back budgets
“As CEO of the Society for Human Resource Management Johnny Taylor Jr told the Guardian, corporate executives are increasingly feeling that, when it comes to AI, “we’re driving the truck and building it at the same time, and that’s a problem.”
“If we could all rewind… we would have introduced [this] in a more thoughtful, slow way,” Taylor continued.”
Why should anyone be surprised? When you decide you’re in a race, speed takes precedence over everything else. It’s an inevitable side effect of manias. Years of easy money make capital appear cheap and it gets thrown about carelessly. The Austrian school calls it “malinvestment”, but you can just call it casting aside all common sense.
When Will AI Meet Cash Flow Reality?
Mauldin’s letter has an excellent example connecting AI asset values with prospective returns.
I was going to write a post comparing the AI credit bubble to the pre-2008 mortgage bubble, but Patrick Feeley beat me to it:
The Next Fannie Mae Is Made Of GPUs
Where The Puck Will Be
“The perception this year has been that the market is all about AI and you have to own the S&P 500 because that’s where you can pretend to be diversified while putting 40% of your money in tech.”
Meta stock sinks nearly 7% as AI spend pushes free cash flow negative
Alphabet sinks 10% as it holds 2026 capex unchanged, expects significant increase in 2027
Its shocking how willing these corporations are to thumb their noses at shareholders. Something is seriously amiss with corporate governance. It’s not new; corporate insiders have been helping themselves to shareholder assets for years with stock options and share buybacks. It’s only partly due to the fact that so many shares are held through intermediaries like mutual funds, exchange traded funds and 401(k)s. Laws and regulations regarding shareholder rights are ripe for review.
This would be another benefit of the bubble bursting. People don’t care as long as their portfolios are going up. They pnly get motivated to start cleaning house when they go down. This is how stability becomes destabilizing … the rot festers … Enron, Global Crossing …