Michael Dell prédit un surinvestissement massif dans l’IA malgré l’explosion des ventes actuelles
L'effervescence artificielle
Michael Dell sonne l'alarme : la ruée vers l'intelligence artificielle suit la trajectoire classique des bulles technologiques. Les chiffres de vente explosent, mais le patron de Dell Technologies anticipe déjà le moment où l'infrastructure dépassera les besoins réels.
Le paradoxe de l'hypercroissance
Les data centers se multiplient plus vite que les cas d'usage concrets. Les investisseurs jettent des milliards dans des GPU qui pourraient finir sous-utilisés - une situation qui rappelle étrangement la folie des dot-com, mais avec des puces plus chères.
La loi des rendements décroissants
Même les optimistes reconnaissent que tous ces investissements ne généreront pas de ROI immédiat. Certains projets d'IA consomment plus d'électricité que des pays entiers pour des résultats parfois... discutables.
Typique : les VC parient sur l'IA comme s'il n'y avait pas de lendemain, alors que la plupart des 'révolutions' technologiques finissent en désillusions coûteuses pour les investisseurs naïfs.
Dell expands sales with Nvidia chips and big customers
Michael explained that the company’s AI servers are powered by Nvidia’s Blackwell Ultra chips.Those machines are then sold to heavy hitters like CoreWeave, a cloud service provider, and xAI, the startup run by Elon Musk.The stock market reacted quickly to this boom.
Dell shares rose more than 3% on Tuesday, right after the company raised its long-term growth outlook in front of analysts.
Michael lifted the company’s expected annual revenue growth to a range of 7% to 9%, up from the earlier 3% to 4% forecast.Diluted earnings per share are now expected to grow 15%, almost double the 8% previously targeted.
This followed strong second-quarter earnings in August. Looking further ahead, Michael said the company plans to ship $20 billion worth of AI servers in fiscal 2026, which is double last year’s figure.
But while sales are exploding, there’s a problem no one can ignore: power.
“It’s the clear constraint that we hear about from our customers, including OpenAI,” Michael said. “Many customers, in fact, will tell us, ‘Well, don’t deliver it until this day because we won’t have power in the building to support it.’”
Power supply questions loom over AI buildout
The scale of these projects is staggering. In September, OpenAI announced a partnership with Nvidia to build at least 10 gigawatts of data centers.
That equals the annual energy use of about 8 million U.S. households, based on analysis from CNBC using data from the Energy Information Administration. The largest tech names (Microsoft, Google, and Amazon) have also committed billions of dollars to new AI facilities.
Michael pointed out that while Dell can design servers to be as energy-efficient as possible, the reality is that these new projects will consume staggering amounts of electricity.
The concern is whether that energy even exists yet. Government data shows the U.S. grid is expected to add 63 gigawatts of capacity in 2025. The 10 gigawatts linked to OpenAI and Nvidia would eat up almost 16% of that total increase.
“At the end of the day, if you’re going to generate tens of trillions of tokens, and you’re going to create intelligence and drive the economy forward, you’re going to need computing power and energy,” Michael said.
Get seen where it counts. Advertise in Cryptopolitan Research and reach crypto’s sharpest investors and builders.