In August I wrote an investment piece titled "AI Bubble Talk is Cheap" for Zacks Confidential to highlight the facts and figures of the 5th industrial revolution being driven by companies like NVIDIA ((NVDA Quick QuoteNVDA - Free Report) ), Taiwan Semiconductor ((TSM Quick QuoteTSM - Free Report) ), and OpenAI.

So I've spent no small amount of time since August continuing to collect the "AI Bubble" arguments and pit them against sound research from AI investment wizards like Coatue Management. More on team Laffont coming up.

And it's always fun. There's no better way to fight the AI bears and doomers who doubt the revolution than simply by buying more NVIDIA, TSMC, Google, Micron, and other smaller innovators.

Last week saw two big attention-getting pieces...

From the Financial Times on Thursday there was 'Phantom' data centres muddy forecasts for US power needs

This article asked good questions about US datacenter developers "flooding utilities with inflated growth plans, muddying efforts to plan for future power needs."

And from the Wall Street Journal on Friday we were given When AI Hype Meets AI Reality: A Reckoning in 6 Charts 

Writer Christopher Mims cites research from JPMorgan about what kind of revenues will be required to support the buildout.

JPMorgan analysts built a financial model that assumes global AI infrastructure investment reaches about $5 trillion by 2030. Then they asked how much extra yearly revenue that pool of hardware must generate to give investors a reasonable return.

Their answer is that the AI stack would need to produce around $650 billion of additional revenue each year for decades to hit a 10% annual return. That's more than 150% of Apple’s current yearly sales and far above OpenAI’s present revenue of about $20 billion.

Two Big Drivers the Bubble Blowers Miss

These articles followed corrective action in several AI-related stocks like Meta Platforms ((META Quick QuoteMETA - Free Report) ), Oracle, and CoreWeave ((CRWV Quick QuoteCRWV - Free Report) ), convincing the doomers and bears that they are right about a big bubble given lots of little ones.

But even Wall Street analysts have consistently underestimated the persistent demand for NVIDIA GPU-driven accelerated computing systems. 

I was the first analyst to raise my NVIDIA price target to $200 in June of 2024 because I did some rough math that told me the company would have sales of $500 billion and be the first $5 trillion company by 2029.

Turns out even I was too conservative on the valuation. But the analysts are still way behind on projecting NVDA sales of only $275 billion for next year (FY'27 begins in February).

So I don't blame journalists for trying to sort through the big numbers and confusing guesstimates about this company and this technology revolution, and coming up skeptical.

To help investors figure it out, I offer two big ways to think about what's happening and why the AI infrastructure capex spending will cross $1 trillion in 2028 (Goldman Sachs and Bank of America research) and total $5 trillion cumulative in 2030 (Citi and JPMorgan research).

1. Traditional economic analysis is treating the AI revolution like a one-and-done additive technology, such as mobile, the cloud, or high-speed connectivity. 

But AI systems multiply economic activity because they are not built on static software but on generative and agentic systems that are constantly producing new tokens of information and value. 

This scale of real-time intelligence will vault economic productivity in every industry. And it requires new and faster computing power.

Jensen Huang recently spoke at The Future of AI Conference in London, hosted by the Financial Times, and said this on a panel...

"Software in the past was pre compiled, and the amount of computation necessary for the software is not very high, but in order for AI to be effective, it has to be contextually aware. It can only produce the intelligence at the moment, you can't produce it in advance and retrieve it. Intelligence has to be produced and generated in real time. And so, as a result, we now have an industry where the computation necessary to produce something that's really valuable in high demand is quite substantial. We have created an industry that requires factories. That's why I remind ourselves that AI needs factories to produce these tokens, to produce the intelligence.

"...it's never happened before, where the computer is actually part of a factory and and so we need hundreds of billions of dollars of these factories in order to serve the trillions of dollars of industries that sits on top of intelligence. AI is intelligence that augments people. And so it addresses labor, it addresses work. It does work. I think we're well in the beginning of intelligence. And the fact of the matter is, most people still don't use AI today, and someday in the near future, almost everything we do, you know, every moment of the day you're going to be engaging AI somehow. And so between what we are today, where the usage is quite low, to where we will be someday, where the usage is basically continuous."

2. After Generative-AI and Agentic-AI add hundreds of basis points to GDP over the coming 5 years, then the emergence and impact of Physical-AI will begin to be felt in all kinds of autonomous machines, from self-driving cars and humanoid robots to automated factories and smart-city systems.

The infrastructure required for training and inference in Physical-AI is possibly beyond the scope of most analysts to project. Human safety is paramount, and so the compute for billions of machines must be robust and redundant, especially off-cloud, aka “the edge.”

I should also add that while investors are focused on the spending and balance sheets of AI companies, most don’t look at the possibilities for AI to transform the world as a turbo-boost for science, medicine, energy, materials, transportation and the ability to solve poverty.

AI Bubble Talk: Staying Sharp on the Behavioral Metrics

Who moves the market over the long run? Large institutional investors with a steady, growth-oriented process. Like Baillie Gifford of Edinburgh, Scotland. They call their long-term philosophy "actual investing."

The firm were early investors in Tesla ((TSLA Quick QuoteTSLA - Free Report) ) and now NVIDIA is their largest position.

BG says "actual investing" explains how making the best returns for their clients comes from discovering the companies that contribute most to progress over the long term. This means they prefer to think in decades, not quarters.

In early October, I wrote a blog piece sharing the latest views of another thematic, long-term investor, Coatue Management, from their quarterly slide deck presentation. At the end of October, I wrote the following and shared a dozen of their key slides...

I went through the deck again, as I suggest you do, and found more insight worth highlighting. Because there will continue to be a flood of journalist articles and broken clocks that want to be right or famous when a bubble pops.

And since this is not an exact science (few things in markets rarely are), it's primarily a behavioral phenomena where beliefs, biases, and blind spots dominate crowd (and our own) actions. Right now, I still think the AI revolution is underhyped and we are nowhere near a euphoria stage of reckless capex and out-of-control speculation in markets.

But we still need to stay vigilant and know how to sort through the garbage, so that we pay attention when some research or viewpoint seems valuable. A lot of Coatue research can help us do that. I start with these 3 slides and share more below...

(end of excerpt from AI Bubble Talk: Staying Sharp on the Behavioral Metrics)

Bottom line: Expect another beat and raise quarter from NVIDIA as Blackwell rack-scale systems are rolling out to high demand. Which means Wall Street analysts will have yet again to raise their growth estimates and price targets for the stock.