Join institutional investors, analysts and strategists from the world's largest banks: Subscribe today
Already have an account? log in
Just Another Tech Bubble Article
The good news is, the rally in tech stocks is "justified."
That's also the bad news to the extent t
Echoes of the nifty fifty. It’s happened before in my lifetime.
+1. They were very similar times, too.
I can’t deny that I’ve found AI useful for relatively simple tasks in my life and business. Image generation is pretty good (Heis seems to have been able to fire his human graphic artist). I’ve used it for code generation for simple algos, but found it absolute crap for anything more than 50 lines of coding. It seems to do a pretty good job of cleaning up my lackluster grammar skills or writing simple things (one time ChatGPT did a bang up job writing a letter to get me out of jury duty). But all of these feel like parlor tricks.
A great testing ground should be retail customer service. There is infinite marking fluff telling us how great AI is for this. But here’s a question for the gang out there: From a retail customer’s standpoint has anyone had a good experience with AI handling relatively simple Q&A from phone or chat?
Seems I always end up waiting furiously to speak with a human representative who often ends up doing a crap, but better than AI job at answering my questions. If wasting my time is the goal then AI does a great job, otherwise I see it failing miserably at tasks that a 6th grade educated human can handle. The Emperor’s cloths have fallen off in my opinion.
Representative please!!!
I had posted an intention to remain quiet while our societal confusion plays out. However your post on effects of AI have my engineer wheel a turning.
I have used AI recently and in most cases the tools and my ‘less than beautiful brain’ is able to come up with a superior product over that of not using these tools at all. At least my brain tells me the result is superior. I even think the tools and I are able to come up with a product that is more quintessentially me than without the tools.
One use case was in the job of a financial advisor. I have been a self directed small investor who has done quite well. However in this success I need to take a different tack. One that better takes into account our current financial imbalances. Working with AI, I was able to come up with a good plan which addressed all the risks I have thought I was managing. I am certain there are other risks that I cannot be aware of therefore cannot be managed a priori. I expect I can use AI to help me manage changes as they occur.
I also used the tool to help write a review publication recently. The publication ended up with 163 references. About 60 of which were found by AI. I also used the tool to change writing style from that of scientific style, my native method, to a review style writing. The paper got published and the underlying thoughts are now in the public domain.
What I see in both of these endeavors is the ability to improve productivity for a person later in their years. I have hopefully some wisdom but not with the tactical knowledge I need to complete the tasks professionally. I know the questions to ask, understand solid philosophical methodologies. However implementation at times is less than I would hope.
I have used it for a few other endeavors: house deal hunting, Tech support and Tax lien business plan formation.
If this is how people will use AI as a skill and cognitive enhancer, then it has the potential to improve productivity. Improved productivity has been stated to be the only way an advanced economy can advance standard of living. In all cases the results are faster than I could do without AI. However it has become obvious that the speed of cognitive evolution in these endeavors was greatly enhanced by the increased speed AI brought to the task. In short I was able to make more high quality revisions than otherwise and therefore improve the products.
Not to say the concerns surrounding AI business model are unfounded. Light processing and Quantum computers have the potential to enhance current data center capacity orders of magnitude beyond current capacity. Will we need all these data centers? Will all be universally upgraded? Will energy demand to flop ratio stay the same? Afterall I did all the above with a ‘free’ tool. My queries are part of the training protocol therefore I have been trading inputs for results. Would I pay for something if the sufficient is available for free? As a retiree, I think not. However society is benefitting nonetheless without resorting to the role of replacing an 8th grade graduate trained in customer service. This post was not written with any AI help.
‘Representative’ is my favorite word when communicating with chat bots.
BTW, I saw an article on Horizontal Morality versus Vertical Morality the other day as a potential pathway out of our current fog. Vertical Morality is the equivalent to following Rome’s wishes, Horizontal morality is the equivalent of love thy neighbor. I hope our problems are resolvable through philosophical framing. Signing out until the fog clears, engineer.
Interesting. I have recently chatted with my daughter about this topic. She is a product developer and project manager for a Silicon Valley software provider specializing in the pharma industry. Her products are AI based and data intensive. So the other day I asked her where’s the money coming from, for your company and even more, for its customers? Everyone who uses some AI personally has stories like yours. But my question remains where’s the money? Chip makers are making money making the machines used to produce AI outputs. But where’s the economic value, the cash flow from those who use AI actually coming from? Are we really effectively monetizing the use of AI? I see Hopium Dealer has similar questions to mine.
Firstly, I also had too much time on my hands this Thursday…
It is highly probable, forward looking “earnings” for the companies in the AI Bubble are not correct, just like they were not correct in 2007 and 2000. 2007/2000 models simply showed/used recent results as the justification for the continued future growth/gains. Spurious/weak analysis, but it is what keeps the analysts employed.
Total MAG-7 Revenue for 2026: $2.6T (estimated, source ChatGPT, but looks at least reasonability inline with real world)
2025 AI Cap Ex Spend: $1 Trillion+ (source: HR article “How AI Spending Could Rob Stocks Of Their Biggest Buyer”)
– Moore’s Law is currently estimated to be 3-years, meaning today’s CapEx will be obsolete in 3 years, meaning a reasonable Deprec/Amort period of 3 years is fair.
Doing the math, adding $330B in costs represents 12.7% (1270 bps) in margin on the $2.6 Revenue base, then add that same cost again for next year’s spend, you’re at about 25% NP margin impact. Total NP Margin for MAG7 for 2026 is 25% (estimated). So essentially, only Apple and Nvidia will be profitable by 2027, everyone else would be unprofitable. Pause and think about that. Think about the 67 P/E ratio the MAG-7 currently has (or 37 if you exclude Tesla, because it’s a meme stock). Then keep in mind, that there is no Revenue/Profit ROI model for the App layer of AI yet. It maps to the Internet bubble almost exactly.
Yes, this analysis is very assumption/proxy driven, but the numbers don’t need to be precise to show the point, forward earnings estimates are wrong. The current models are either: under estimating Depre/Amort or over estimating Revenue gains, probably both.
Hopeful One, don’t be so pessimistic! A good paper chart and a stiff steel ruler is all you need to forecast future revenue and profit growth. As long as you are forecasting growth, that is. Others here and in the tech rather than Wall Street world have laid out more exact forecasts which all support your calculations.
“So essentially, only Apple and Nvidia will be profitable by 2027” . Those who are unfortunate enough to know me outside of this metaverse know that I am hardly an Apple fan boy, but I have huge respect for the company for refusing to squander shareholder money creating LLMs. Which, you may recall triggered murmurings that “Apple has missed AI” among the wise men on Wall Street and in the financial press.
Do you want promises or money?
As Ray will put it, one company’s capex is another company’s EPS.
The last four paragraphs are awesome. Thank-you for including them.