Dell sets quarterly record with $60.9 billion in AI server orders

Boiled down, what these results really show is that companies are changing where they run AI. For years, businesses rented cloud servers from hyperscalers like Amazon, Microsoft, and Google to train and run AI. Lately, though, more of that work is moving into companies' own buildings, onto servers they own outright. That's called on-premises computing, and Dell is the company that builds and sells exactly those on-premises servers.

The market has long undervalued Dell, treating it as a legacy PC maker that mostly assembles parts — which is why it traded at a lower price multiple than other AI-linked companies, even below the S&P 500 average. But this quarter's numbers pushed back against that skepticism. Dell turns out to be the company physically supplying the servers, storage, and networking gear that run AI infrastructure — for both hyperscalers and enterprises alike.

There are roughly three reasons companies are bringing AI back in-house. The first is data. Sensitive corporate data — contracts, patient records, telemetry — already sits on a company's own servers, and it's cheaper, faster, and safer to bring AI to where that data lives than to ship the data out to the cloud. The second is control. Banks, hospitals, defense contractors, and government agencies are wary of letting sensitive data leave their premises or cross borders. Michael Dell said in May that the real risk lies in losing control over data, cost, security, intellectual property, and speed. The third is cost. Training a model can be done with short-term rentals, but running AI agents around the clock is a different story — continuous rental ends up costing more, so companies are opting to install their own servers instead.

Amazon CEO Andy Jassy made a similar point in a recent earnings call, saying "85% of global IT spend is still on-premises." Cloud may look like the dominant model, but most enterprise data still lives on companies' own servers — and the demand to turn that data into a competitive AI advantage is flowing straight into Dell's order book.

Unlike people, AI agents don't eat or sleep — they work continuously, and every task they run generates more data that needs to be stored and protected. That cycle has helped Dell surpass 6,500 AI enterprise customers, 3,300 of them added in just the past three quarters. Traditional server revenue grew 122%, and storage revenue grew 26%. Operating margin in the infrastructure segment expanded 620 basis points to 15%, a level Morgan Stanley called unprecedented. Because Dell sells servers, storage, networking, PCs, and services all under one roof — and buys components together as well — it secured better purchasing leverage than rivals even during a period of memory shortages.

Dell was founded in 1984 by a 19-year-old Michael Dell out of a college dorm room. In 2013, the company fought off activist investor Carl Icahn over a plan to take Dell private — and won. Three years later, in 2016, Dell acquired EMC for $67 billion, the largest acquisition in tech industry history at the time. That deal was a bet that companies would ultimately want to own their infrastructure rather than rent it. Exactly a decade later, that bet has materialized in the form of the on-premises AI shift.

Most of the rivals that started out alongside Dell are gone now. Compaq, the world's largest PC maker in 2000 with $42 billion in revenue, was absorbed into HP — which itself later split into four separate companies. Digital Equipment followed a similar path. Data General, once bigger than Dell, was sold to EMC in 1999 — and when Dell bought EMC outright in 2016, Data General effectively became part of Dell too. Dell didn't just outlast its competitors; in some cases, it ended up owning them.

Part of why Dell has managed to stay standing is that Michael Dell has retained effective control over the company through a dual-class share structure. While rivals got stuck as their eras ended, Dell's governance let it keep pivoting its business lines.

In the end, this quarter confirms one thing: the view of Dell as merely a server assembler has now been contradicted by the numbers, and Dell sits squarely at the center of the shift as companies move AI out of the cloud and into their own buildings. Training AI can still be done with short-term rentals, but in an era where agents run nonstop, the math increasingly favors owning your own servers — and that math is what keeps filling Dell's order book.