Dell's AI Server Revenue Jumps 757% as Backlog Hits $95 Billion

Dell’s quarterly results on September 1 put a number on how absurd AI infrastructure demand has become: AI server revenue hit $16.1 billion in the quarter, up 757% year over year, per CNBC’s coverage of the report. Total revenue came in at $46.97 billion against $44.92 billion expected (up 57.7% year over year), adjusted EPS of $7.04 crushed the $4.92 consensus, and the stock jumped roughly 9% after hours. Dell raised its full-year fiscal 2027 revenue guidance to approximately $192 billion on the strength of the AI business.

The Order Book Is the Real Story

The quarter’s headline revenue is less important than what is stacked behind it. Dell booked a record $60.9 billion in AI server orders during the quarter and now carries an AI backlog of approximately $95 billion, which means the revenue surge everyone is watching is only the portion of demand that has already been built and shipped. That backlog covers multiple quarters of production at current rates and explains why Dell can raise full-year guidance by tens of billions mid-year. For anyone modeling AI capex, the interesting quantity is no longer what hyperscalers say they will spend; it is how far the supply chain is booked out, and right now the answer is deep into fiscal 2027.

What It Says About Where Compute Is Going

The composition of that backlog also says something about how the buildout is maturing. Dell is the archetype of the second wave of AI infrastructure: not a hyperscaler designing its own silicon, but a systems integrator assembling racks of accelerators for enterprises, governments, and neoclouds, the same buyer class behind deals like the neocloud financing rounds around Groq’s Nvidia-cluster business. A 757% growth rate at Dell’s scale implies the buyer set has broadened well past the frontier labs; there are not enough frontier labs to order $95 billion of servers. Watch the mix of that backlog over the next two quarters: sovereign AI buyers, enterprises standing up private clusters, and inference-focused operators are where the incremental orders are coming from, and that mix determines whether the current guidance survives contact with 2027.