Dell booked $16.4 billion in AI systems revenue in the quarter ended July 31, double the year-ago figure, and ended the period with $95 billion in AI bookings — a backlog 8.1 times larger than a year ago. The numbers come from the company's second quarter of fiscal 2027 results, and they reframe Dell as an AI infrastructure vendor rather than a legacy server vendor that AI forgot.

The headline projection is larger still. COO and vice chairman Jeff Clarke told analysts: "We are expecting AI to be 75 percent of all datacenter demand by 2030, adding 200 GW of power over that same time frame." Dell believes half of that power buildout sits with its customers — neoclouds, sovereigns and enterprises — and sizes the opportunity at more than $1 trillion.

How fast is the AI customer base growing?

Clarke offered a clean measure of acceleration. It took eight quarters to reach 3,200 customers who had bought an "AI factory." The next 3,300 closed in only three quarters, bringing the total to 6,500 organizations running AI systems under the PowerEdge brand. The buyer mix has widened from Tier 2 clouds and neoclouds to large enterprises that now ask their traditional server supplier for inference clusters as well.

What is happening to the traditional server business?

The quieter story may matter more. Traditional servers plus networking hit an all-time quarterly high of $10.53 billion, up 2.2X year on year. The driver is a long-postponed refresh: over 1.2 million servers in Dell's installed base are at the PowerEdge 14G generation or older, machines announced in 2017 and ramped in 2018. Clarke pegs consolidation at six to eight vintage servers per 17G system, and twelve to fourteen per 18G machine, which arrives next month.

These replacement systems are heavily configured and inherently more expensive, and Dell is passing through rising DRAM, flash and CPU costs. Based on the Infrastructure Solutions Group forecasts, traditional servers would reach roughly $25.1 billion in Q4 and eclipse AI server sales — just as Nvidia's Vera-Rubin systems ramp to volume.

Group-level, ISG posted $31.78 billion in revenue, up 89.2 percent, with operating income of $4.78 billion, up 3.25X. Company-wide revenue reached $46.97 billion, up 57.7 percent, with operating income of $5.36 billion and net income of $4.13 billion — 8.8 percent of revenue and, excluding two divestiture-inflated quarters, Dell's best profitability since before the Great Recession. Margins have moved from the 2-to-3 percent territory of grocery chains and ODM rivals toward 8-to-10 percent.

What did Clarke say about Project Lightning?

Dell also updated its two-year-old Project Lightning parallel file system, aimed at AI workloads with many small I/Os against flash block storage — not Isilon's OneFS, which powers PowerScale. "It is in beta at several customers," Clarke said. "We are in runoffs against other competitors with the product." He named no dates. The competition — DDN's Lustre, VAST Data, WEKA — is probably not worried yet.

Can the debt and the opportunity both grow?

Dell's balance sheet carries the usual tension. Debt is rising again, though far below the post-EMC peak, and cash and debt are climbing at roughly the same rate, so net debt is roughly stable. Dell holds $14.25 billion in cash and investments and spent $2.2 billion on buybacks and dividends in the quarter. Michael Dell and his wife own 41 percent of shares and collect the same share of payouts. Wall Street, so far, does not object; revenue is growing faster than debt.

For the full year, Dell guides to about $19 billion in AI system sales in Q3 and $22.5 billion in Q4, for roughly $74 billion total. Clarke's claim that "inference is past training" and that enterprise agentic workloads become the largest category by 2028 rests on projections, not bookings. The $95 billion backlog, by contrast, is signed — and it is the number Supermicro, IEIT Systems, Quanta and Foxconn will have to take apart.