Jevons Paradox And Why Demand For More AI Infrastructure Is Insatiable

The case against the memory stocks and AI infrastructure-related stocks across the board is that AI demand will peak and then fall, leading to a collapse in hardware prices. A related argument is that all the spending on data centre infrastructure will end in tears, with investment returns collapsing.

Bad stuff happens, but looking for the nigger in the woodpile when we haven’t even found the woodpile may be premature. I know the nigger word, shock, horror, but it’s just an expression with no racial implications whatsoever. Apparently, that is not true; it is deeply offensive and was first used in a cartoon to describe Abraham Lincoln’s initial indecisiveness about the slavery issue when he became president. We live and learn.

The counterargument is explained by Jevons’ Paradox. William Stanley Jevons was a Victorian economist who noticed that the more efficient steam engines became in their use of coal, the more coal demand rose exponentially. He predicted that Britain would run out of coal if these efficiency gains continued because the appetite for fast, cheap transport was insatiable. It didn’t happen because of oil, gas, automotive, and all sorts of other reasons.

The same applies to the AI revolution. Spotify is being taken over by AI. My favourite singer-songwriter is AI , operating under a host of different names. It is becoming a big problem for artists with three of the latest number one hits in the USA, AI generated. AI can write, and boy, does AI have a thrilling voice, and he/ she is raising his/ her game at an exponential rate. Soon, I expect AI to go touring. Watch out, Taylor Swift.

I write books, and AI can make synopses of my books that provide insights for me even though I wrote them. It is almost amusing how amazing AI has become, and this whole thing is just starting. As Jevons says, it doesn’t matter how fast and furious they make data centres, AI will want MORE! This is key because it suggests there may not be any cyclical peak.

As companies like SK Hynix, Nvidia, Micron Technologies, SanDisk and many others are suggesting, this boom is so ferocious and new applications for AI are going to be discovered/ invented at such an exponential rate that, however efficient data centres become and however much capacity is put in place, it will ALL be used, and more will be demanded.

I read a bearish piece on SanDisk over the weekend. It seemed like gobbledy-gook to me and boiled down to the old cyclical peak idea. I also listened to David Goeckeler speaking at the Mizuho Technology Conference on 9 June. The Seeking Alpha guy somehow drew bearish conclusions from listening to the same conference, which struck me as a triumph of the power of preconceived ideas.

What struck me from the conference is that the analyst interviewer predicted that SanDisk sales and eps for the year to 27 June 2027, a year that will soon begin, would be £45bn and 200p respectively. Both these projections went unchallenged by the CEO and CFO of SanDisk. My immediate thought is that they will do even better. Christ! Could we be looking at sales of nearer $50bn and eps around 225p? I have no idea, but it certainly gives the impression of a company that is (a) growing at an incredible rate and (b) if Jevons Paradox is right, faces an extended, exciting future, stretching out for the foreseeable future.

This is staggering because, as recently as last summer, SanDisk was valued at around $6bn. Its fiscal 2027 sales could be around eight times what it was valued at some 12 months ago, and the then level of sales. If SanDisk has become a structural AI growth play, which is perfectly possible, it could reasonably be valued at, say, 30 times earnings, given its growth rate and its critical importance to the AI infrastructure buildout.

Let’s do the sums. Thirty times earnings per share of $225 equals a share price of $6,750. No wonder the analyst at Mizuho was saying, not entirely joking, that even after its huge 30x rise, it could still do another 10x in the coming years.

I sometimes think there are two kinds of investors – contrarians who are always looking for trends to end, booms to fail and share prices to collapse and trend followers, like me, who believe, supported by observation, that trees do grow to the sky. The trick is to pick the right tree.

The biggest problem for SanDisk and so many of these AI hardware stocks with exponentially rising share prices is that many shareholders are sitting on such huge profits that at the first sniff of trouble, they run for cover, and the share price has a temporary swoon. Try to make this volatility your friend by using it to build your holdings.