When considering such headlines, it's important to remember the history of major innovations in order to avoid conflating market excesses with the productivity promise of AI diffusion. This history—especially for innovations with general-purpose attributes—is full of initial speculative excesses.. There are many reasons for this, including the human behavioral tendency to initially over-produce and over-consume something promising whose barriers to entry have suddenly dropped. (Just think of past examples such as the steam engine and, more recently, fiber optics.) #economy #innovation #AI #markets
Yes, there’s an AI bubble. Good. Bubbles finance the infrastructure we keep: data centers, fiber, tooling, and talent. The dot-com bust left a “surplus” of fiber/CDNs exactly what later powered Google, Amazon, Netflix, YouTube, Zoom. Same pattern now: capex overshoots first; productivity shows up later on the S-curve. How to separate hype from payoff Distribution moats: embedded in OS/cloud/enterprise suites. Proprietary data + workflow integration (real usage, not demos). Unit economics trending right: rising GPU utilization, revenue per seat/query, payback < 12 months. Power plan: secure, scalable energy footprints. Most startups won’t make it. That’s normal. What matters is who captures a lower cost of cognition across millions of users. Investor playbook: picks-and-shovels (chips, cooling, power, DC builders/REITs), model-agnostic platforms, and verticals with measurable cost takeout (support, coding, compliance, marketing). Headline angst ≠ economic reality. Overbuild is a feature, not a bug.
We live in a dynamic market where uncertainty, panic, and fear are natural byproducts of disruptive shifts in both consumer and business markets. Yes, AI today feels like an unstoppable wave, almost viral in its spread. But rather than react , I lean on the principle of being fearful when others are greedy, and greedy when others are fearful. Proactive risk averse, taking an analytical Real option approach. I was always a believer that Technology assets, when grounded in real productivity, will outperform traditional financial services! And it did! The challenge lies in valuations running far ahead of actual earnings. A company’s market cap should reflect tangible asset performance, with economic and social value creation, not just speculative hype. But within the noise, the true long-term winners are those who convert potential into measurable, sustainable value. The key is to separate reactive market players, those chasing headlines from proactive builders who harness AI’s real capabilities as a tool to solve problems and create durable wealth. That’s where I choose to focus when others panic.
Thanks for sharing, and for the realism here. The AI bubble, in my view, is only a bubble because the cannon is pointed toward the wrong field of battle. Deep Learning, LLMs, SLMs, etc... these systems are not intelligent in the way humans are intelligent. Most of the poor capital allocations of today are because an investor is attempting to anthropomorphize the technology, and believes that AI can do something that it cannot do. Neural Nets are outstanding at pattern-matching, but not systems-level thinking. At least not yet. Once we see capital pointed (and spent) in the direction of promising AI innovations that celebrate their pattern-matching abilities, instead of AI's poor generative abilities, I think the world will be onto something.
The telling number is enterprise AI implementation penetration over the next 1~2 years, Mohamed El-Erian. Most studies put it around 15%. That doesn’t account for shadow AI, but that would pale in comparison in terms of revenue.
We see a lot of hype cycles come and go (e.g. ICOs), but tech that survives those circles (e.g. the internet) shape the entire economy.. it’s likely that AI is in that latter category
But the good thing about this bubble is it will leave infrastructure for future use or innovation at lesser cost as happened with rail mania of 1850s and telecom bubble of late 90s....for deep insights on financial history one should read book titled: Innovation, Euphoria and Financial Crisis
Perhaps measured amounts of spending on proof of concept projects until there’s a thorough understanding of the technology would be wise. Grow investment from there. Nothing speaks truth like results. I have no doubt the technology is here to stay and will only improve with time.
Mohamed El-Erian we recently wrote a paper for the Absolute Strategy Research Ltd clients about the link between the cost of capital typically seen in General Purpose Technology ‘bubbles’ - the excess capex build out that each of these cycles brings and the ‘schumpterian waste’ that is essential to the eventual ubiquity of these technologies. The insights of William Janeway are relevant here. The main point is that while society gains from these bubbles in the long run - it is equity holders that pay - often with declines of 70%-80% in the value of their equity holdings. Happy to share the paper if you message me. Ian H
Owner - Family Owned Businesses
3wLet me share my experience with AI and its power. I have a Wall Street trading desk, which is basically a couple of computers 6 monitors at this point and all the related electronic gadgets. I am using AI to conduct in depth market research. So modern financial markets have about 425 years of day-to-day recorded history as well as the history of the surrounding environment. So an amazingly robust and spread out collection of databases. I often search for trading patterns that requires the acquisition, study and interpretation of large numbers of subsets of this global data. The trick is asking AI the right question and giving it the right data sources. So in a matter of two or three seconds it will give me a 100% accurate answer to extremely complicated data inquiries. To put it in contrast it would take a group of 10 Cambridge PhD‘s at least 90 days to comb through this data at these levels and yet they still would not get the right answers. In addition same Desk is used to manage presidential campaign. AI helps create million dollar videos in just a few minutes. Through Google Ads and other platforms I am able to reach 350 million Americans within 24 hours. So 1 guy does the work of 30 people. Cost is gear & connection.