Micron FQ3 24 – AI Canary?

Micron spooks the fickle market

• Another correction from an AI-related company has the market asking whether the question as whether the AI bubble is beginning to burst which is a question that only Nvidia will be able to answer.
• Micron reported good FQ3 results but failed to guide as strongly as many had hoped raising fears that demand for AI has peaked and that there are declines ahead.
• FQ3 revenues / EPS were $6.81bn / $0.62 well ahead of the EPS forecast of $6.67 / $0.48 but guidance was less robust.
• Here, Micron expects that FQ4 24 revenues / EPS to be $7.4bn – $7.8bn / $1.00 – $1.16 broadly in line with consensus of $7.61 / $1.06 and it was this failure to beat and raise that caused the shares to decline 8% in after-hours trading.
• Micron stated that its AI-related HBM3E memory is sold out for calendar 2025 which is a sign that Micron has more demand for this product than it can make.
• The very strong profitability performance is another sign that demand for its AI-related products has not topped out yet.
• This mirrors precisely what is going on at Nvidia which is also running as hard as it can to meet demand but where revenues are likely to be determined by how much capacity it has reserved at TSMC as the phones are still ringing off the hook.
• Hence, I suspect that the party is not over yet and that when Nvidia reports in August, there will be a huge sigh of relief and the valuations will hold for a little longer.
• This is going to remain the case for as long as capital continues to pour into the sector with almost no questions being asked.
• No one seems to wondering how all of these start-ups are going to make any money or worry that as the supply of models increases, the prices that can be charged to use them will fall.
• Furthermore, with the performance from one to the other becoming increasingly similar, the stage is set for brutal competition that could bring the price paid for these services down by 90% or more.
• There is still good money to be made at these prices, but it requires scale and for the services to run on user devices rather than in the cloud.
• Open-source and free models that perform almost as well as the paid versions are only going to exacerbate the pricing pressure.
• Hence, at some point, someone is going to start wondering where the returns on invested capital are which could trigger a substantial reset of expectations.
• It is clear from the empirical evidence that statistical-based systems such as deep learning and large language models are not going to deliver superintelligent machines, but this is what the market is pricing in and what everyone seems to expect.
• This is on full display in Micron’s commentary where it states “We are in the early innings of a multiyear race to enable artificial general intelligence, or AGI, which will revolutionize all aspects of life”, for which there is no evidence whatsoever.
• It is this kind of hyperbole that is setting the market up for a large correction.
• Instead, what has been created is a technology that allows humans to converse with machines in a natural way allowing voice to finally become a great man-machine interface as well as next-level database building and retrieval.
• There are large revenue and profit opportunities from exploiting these abilities, but this does not include machines taking over from humans in almost all intellectual and economic activities.
• Consequently, while I think that there is a reset coming, this report from Micron is not the canary in the coal mine and I am pretty sure that the party will continue when Nvidia reassures the market in August.
• Despite its rally, Nvidia is one of the cheapest AI opportunities available in the market today (Meta, Microsoft and Google are not AI investments) and so if I had to have a position in generative AI, Nvidia would be it.
• However, I prefer to look more laterally at inference at the edge or nuclear power to run all of these new data centres that are springing up.
• I have positions in both of these themes.