Dell Technologies has become one of the most closely watched indicators of the AI infrastructure boom, and its latest quarterly numbers only reinforce the view that demand is still running far ahead of supply. The company reported a record AI backlog of $95 billion in unfilled orders, with quarterly earnings showing more than 50% year-over-year growth in AI demand. For Dell, the challenge is no longer about generating interest; it is about getting enough servers, storage systems, memory, and networking gear into the hands of customers who are waiting.
The scale of the backlog reflects a broader industry reality: AI workloads, especially agentic AI systems that can reason, plan, and take actions on their own, require enormous amounts of compute, memory, and storage. The need is not limited to flashy graphics processing units, or GPUs. It extends across the entire data center stack. Dell Chief Operating Officer Jeff Clarke said on an earnings call that supply constraints begin with servers and storage and reach “just about every product going through a leading node.” He added: “We are doing everything we can to get more supply. In today’s environment, that’s a very difficult task.”
A glimpse of infrastructure demands ahead
Dell said that for its financial quarter ending July 31, revenue came in at $47 billion, a 58% increase year over year. Revenue from Dell’s Infrastructure Solutions Group, a key measure of data center demand, climbed 89% to a record $31.8 billion. Much of that growth came from server sales, including traditional CPU-based servers that are increasingly being used to support agentic AI workloads. Dell said earnings from those traditional servers rose 122% year over year.
The backlog figures are even more striking. Dell booked nearly $61 billion in AI server orders during the three months ending July 31. Over the last 12 months, it has taken in more than $130 billion in AI server orders. Clark said Dell converted $131.7 billion of demand into orders over that period, and the demand is broadening well beyond the largest cloud providers. Enterprises, neoclouds, and sovereign cloud providers are all competing for the same constrained supplies.
One telling example of this broadening demand is Dell AI Factory, the company’s platform for building AI workflows. Dell said it has more than 6,500 customers using the platform, with 3,300 joining in the last three quarters. By comparison, it took Dell about two years to sign up its first 3,200 customers after the platform debuted in May 2024. That pace of adoption suggests that AI infrastructure investment is becoming a mainstream priority.
Agentic AI is reshaping the data center
The term “agentic AI” describes systems that do more than generate text or analyze data. These AI agents can break down goals into tasks, use tools, and execute workflows with limited human supervision. As their use grows, so does the demand for data center resources. Inference, the process of generating outputs from trained models, is especially resource-intensive because agents may need to run many inference calls in sequence or in parallel before completing a single task.
“Agentic demand is reshaping the data center,” Clarke said. He described inference as “pure demand in our industry.” Dell’s internal forecasts point to a dramatic surge in token usage, the basic units of text and data that AI models process. The company expects 3,600 quadrillion tokens will be in use by 2030, representing an 87-fold increase from today. Over the same period, training demand is expected to grow to 850 zettaflops, a fivefold jump. If those projections are even close, the pressure on infrastructure supply will continue for years.
Dell also predicts that enterprise agentic AI will become the single largest data center workload by 2028 and will account for 75% of all data center demand by 2030. That has profound implications for companies planning their IT budgets. Many are realizing that their existing server fleets were not designed for the scale of data movement, real-time inferencing, and model orchestration that agentic AI requires.
Enterprises clamor for traditional servers
Although much of the conversation around AI has focused on accelerated computing, Dell sees a surge in demand for traditional CPU-based servers as well. Customers need “meaningful CPU compute capacity” to support AI endpoints, enterprise applications, data pipelines, and agent workloads. The demand has been strong enough that Dell generated nearly as much revenue from traditional servers and networking in its last two financial quarters as it had in any prior full year in company history.
Most of that growth comes from existing customers who are accelerating investments in traditional IT environments to refresh aging hardware and modernize data centers. They are looking for better performance, efficiency, and resiliency. Dell expects “significant and durable” refresh cycles ahead, driven not only by AI but also by higher security requirements and the need to handle expanding data volumes.
“AI requires modern, disaggregated architectures that keep data accessible and in motion across compute, storage, and networking,” Clarke said. He noted that modern AI deployments require much more than assembling components. Customers may need more than 50 unique server designs as they optimize for performance, power, cooling, and physical space. This engineering complexity adds to Dell’s challenges because every design must be manufactured, tested, and delivered even as component shortages persist.
DRAM, NAND, and the long tail of shortages
When Dell executives talk about supply constraints, they often point to memory and storage shortages first. “DRAM, DRAM, DRAM, followed by NAND, NAND, NAND,” Clarke said. Dynamic random-access memory, or DRAM, is critical for AI inference and training because it stores the data and model parameters that processors need to access quickly. NAND flash provides high-speed storage for the enormous datasets used in AI pipelines.
There are also “spotty” shortages of CPUs and disk drives. The constraints stretch down the entire supply chain, from microcontrollers to transistors. That means Dell cannot simply wait for one type of component to become available; it must manage a complex web of dependencies. In many cases, the company is optimizing the components it has to maximize production.
“We’ll continue to focus on trying to get more supply, and take the supply we have and optimize the output,” Clarke said.
Storage products have seen particularly strong growth as enterprises prepare, manage, and protect huge volumes of data. Dell reported robust demand across its PowerFlex, PowerStore, PowerProtect, and PowerVault product lines. “Demand remains broad based; enterprises continue to modernize their storage environments as data growth increases the importance of keeping data available and secure,” Clarke said.
How customers respond to shortages
Customers facing long lead times and higher prices are changing their purchasing behavior. Clarke acknowledged that modern configurations come with higher core counts, more DRAM, and more storage, and “those configurations cost more than they did last quarter, and the quarter before, and the quarter before.” The price increases are straining IT budgets that were not always flexible enough to absorb multiple quarters of inflation in key components.
Some customers are deferring purchases until they can realign budget dollars. Others are placing orders much further in advance than they did in previous years to ensure access to constrained supplies. Large enterprises and multinational corporations, Clarke said, “would prefer to have products now if we had the supply.” He added that Dell is “supply constrained in the sense of what we can build in any given quarter.”
The response among Dell’s largest customers is different. They are collaborating closely with Dell to forecast future needs and reserve capacity ahead of time. That kind of forward planning was rare before the latest AI surge. “That is a new phenomenon,” Clarke said. Dell is also working to reduce lead times, though the company admitted that building more gear remains difficult when the supply of memory, storage, CPUs, and networking components is so tight.
Demand across every customer segment
One of the most significant takeaways from Dell’s report is that AI infrastructure demand is no longer confined to hyperscale data centers. Mid-sized enterprises, government agencies, financial institutions, and healthcare organizations are all deploying AI workloads. Sovereign cloud providers, which operate cloud services within a specific country to meet data residency and regulatory requirements, are also driving new orders.
Dell AI Factory’s customer base illustrates this shift. More than 6,500 organizations now use the platform, and the rapid growth suggests that many enterprises are looking for an integrated way to handle AI deployment rather than stitching together pieces from multiple vendors. This kind of adoption can put more pressure on supply because new deployments create demand for backup systems, test environments, and future capacity expansions.
The infrastructure crunch is also spreading to the physical data center itself. More compute density means more power consumption and more heat to remove. Enterprises are being forced to rethink their power distribution, cooling systems, and rack layouts. Dell’s engineering effort includes working with customers on the total environment in which servers operate, not just the server configuration itself. That added complexity, combined with record order levels, makes it difficult to shorten lead times.
There are early signs that memory manufacturers are planning new capacity, but it will take time to build new fabrication plants and qualifying products. In the meantime, AI workloads continue to grow in scale and sophistication. Agentic AI systems, in particular, tend to be more dynamic than earlier batch-processing models. They may run continuously, pause for user input, spawn subtasks, and write to storage as they go. That behavior makes them even more dependent on low-latency memory and high-throughput storage fabrics.
Dell’s record numbers are therefore best understood not as a one-time spike but as an indication of a long-term structural trend. The company’s revenue and order figures point to an industry that is still under-building when measured against the future demands of AI. Servers, storage systems, memory devices, processors, and networking hardware are all running at capacity. While Dell works to improve its own supply situation, the flood of orders suggests that the infrastructure crunch is far from over.
Source: Network World News