Dell Technologies

Where we land

Dell has become one of the main suppliers of the computers that power artificial intelligence. Its AI servers, machines built to train and run AI models, now account for roughly a third of sales and helped lift total sales 49% in a year. The share price has climbed steeply too, but it asks for much less than the business has recently delivered: earnings growth of about 7% a year, against a recent pace nearer 34%.

The price leaves ample room for a slowdown. That does not make a slowdown unlikely or harmless. The real questions are how durable the growth is and how much of it reaches profit. We award Dell a silver medal. Among computer and hardware makers, whose own neighbours are rated even more highly, it ranks bronze.

What the business is

Dell makes three kinds of things. The largest is general computer hardware, chiefly PCs and laptops for businesses and households, at 45% of revenue. Servers, the data-centre machines that now include the AI systems driving the story, bring in 39%. Storage systems, which hold the data those servers work on, make up the remaining 15%.

This is a high-volume, thin-margin business. Of every dollar of sales, about 20 cents is left after paying for components and assembly. About 10 cents remains as operating profit after sales, research and administration. Gross margin has been remarkably steady, sitting around a middle value of 22% over five years and typically moving only a point either side.

Why now

The surge comes from one source. Cloud companies, specialist AI cloud providers and large enterprises are buying whole racks of servers built around Nvidia's chips, and Dell is one of the companies assembling and delivering them. In the quarter to July, Dell took 60.9 billion dollars of new AI server orders, more than its entire revenue for the period. It delivered 16.4 billion dollars' worth, double the figure a year before. It reports a backlog of 95 billion dollars still to deliver and has raised its full-year sales forecast to 192 billion dollars, from 167 billion. A year ago Dell's growth looked unremarkable against the wider market; it now ranks among the faster growers.

The case for

The first argument is the scale of what has already happened. Sales grew 48.9% over the past year, against 17.4% a year over three years, so the AI build-out has sharply accelerated an existing trend. The extra sales have also reached the bottom line. Twelve-month operating profit doubled, from 7.5 billion dollars to 15.0 billion. Net profit, what is left for shareholders after interest and tax, rose from 4.8 billion to 11.4 billion. AI servers carry thinner margins than Dell's older products, yet volume has more than made up for it.

The second is visibility. Most hardware makers sell into a future they can only guess at. Dell's backlog, built up in a quarter where orders outran total revenue, gives unusual confidence about sales well into the coming year. The shares rose again in early October on further backlog news, a sign that investors see the order book as the heart of the story.

The third is that forecasters have moved in one direction only. Over the past 45 days, earnings forecasts were raised 50 times and cut not once, and the consensus now expects earnings per share to rise 19.4% next year. That unanimity cuts both ways: when every forecast has gone up, so has the bar for the next set of results.

Where the growth could come from

The engine is the AI server order book, which sits inside the server line. Last quarter AI servers delivered 16.4 billion dollars of the 47.0 billion in total sales. For this to keep working, new orders need to keep matching or beating shipments so the backlog holds or grows. The open question is profit. AI servers earn less per sale, so earnings will grow more slowly than sales unless Dell does one of two things. It can attach more of its own software, networking, storage and services to these deals, or it can hold AI server margins steady as volumes climb.

The challenge comes from the ordinary, non-AI side of servers and from storage. Barclays recently cut its view of Dell to the equivalent of a hold. It argued that strong sales to businesses owed more to high memory prices than to companies genuinely moving to AI. Memory is a large part of a server's cost, so when it is expensive the price of the finished machine rises with it. Those gains reverse as memory cheapens and companies spend cautiously. Barclays downgraded Hewlett Packard Enterprise and NetApp the same day, so this is a view about the whole industry rather than Dell alone. Our figures do not separate growth that came from higher prices from growth that came from more machines sold, which is exactly why it matters.

PCs, the biggest single line, are a steady contributor rather than the growth engine. Demand depends on companies replacing ageing machines, helped by newer versions of Windows and by PCs built to run AI tasks. Costlier memory and cautious business spending pull the other way. PCs carry more of Dell's profit than their growth suggests. If Dell cannot pass on higher component costs, or the replacement cycle stalls, earnings would suffer even while AI servers boom. The sign of health is business PC sales growing while their margin holds.

Storage could turn high-volume AI sales into faster earnings growth. Companies running AI need large amounts of fast storage for the data feeding their models, and Dell's storage systems earn better margins than AI server hardware. Our figures do not show how much of storage is tied to AI today. Storage returning to steady growth, with AI customers adding it to their server orders, would confirm the link.

Further out, Dell has signed a non-binding agreement with the Japanese power company JERA and RHAELM for a 400-megawatt AI data centre in Chiba, planned to begin operating in 2028. Access to electricity has become a bottleneck for AI build-outs. Pairing power with Dell's computing could open large, long-dated orders beyond the big US cloud customers. For now it is an option rather than revenue. The first real sign would be a binding contract with a named server order, or similar deals elsewhere.

The case against

Start with margins. Hardware peers weighted towards storage do far better than Dell's 20 cents of gross profit per sales dollar: Seagate keeps 46 cents and Western Digital 49. Among its peer group only Foxconn Industrial Internet, a contract manufacturer, earns thinner margins. Dell's profit depends on volume rather than pricing power, and that dependence grows as AI servers take a larger share of sales.

Worse, gross margin has slipped to 19.7%, below the 22% around which it had barely moved for five years. That is the AI mix showing through: each extra dollar of AI server sales adds less profit than a dollar of the older business. If the slide continued, the gap between the sales story and the earnings story would widen.

Then there is the price. Dell trades at 26 times earnings, meaning investors pay 26 dollars for each dollar of annual profit. Western Digital trades at 14 times, Canon at 12 and Sandisk at 18. Against the wider market, the shares have become markedly more expensive over the past six and twelve months, and they have risen faster than almost any others. Shares priced like that have little room to disappoint.

Finally, the mood has shifted. Headlines in early October reported more than 1 billion dollars of share sales by company insiders. On 15 October, after the run to all-time highs, the stock fell 8% in a single day. Our evidence does not say why, but alongside the Barclays downgrade these mark a change from the confident rally of September.

What the price assumes

One way to judge a share price is to ask what it already takes for granted. Suppose the price stayed where it is for five years, and Dell then traded at 18.6 times earnings, the middle value for every profitable listed company we cover. For that to be fair, earnings would need to grow about 7% a year, reaching 24.26 dollars a share.

Now suppose the recent trend in operating profit, driven above all by the AI build-out, simply continued. Earnings would grow about 34% a year to 73.30 dollars a share, 302% of what the price requires. That is the company's own recent results carried forward, not a forecast. A once-in-a-generation buying spree will not run at that pace for five years. But it shows how much slowing the price can absorb.

That sits alongside the shares looking expensive, and both are true. Against today's earnings and against hardware peers, Dell is dear. Against the earnings the AI business has recently produced, the price asks little. The shares have risen 204% over the past year and 131% over six months, so much of the good news has been paid for. What the price has not assumed is growth anywhere near its recent rate.

The other side of the medal

The risk we would keep in view is that part of the surge in ordinary servers and storage rests on buyers paying up for costlier memory. That lift fades as memory prices settle. It matters most because it attacks the part of the story that looks like a normal, durable business. AI orders come with a visible backlog; traditional server and storage sales do not. If their strength came from higher prices rather than more machines, it would unwind with no warning from the order book.

It bites harder because, measured against its assets, sales and earnings, Dell is dearer than most of the companies we rate. Growth inflated by component prices would therefore be felt in the share price as well as the results. Three signs would show the risk arriving. Memory prices would ease while ordinary server sales slow. Management would describe customers as buying ahead of price rises rather than for AI or to replace old equipment. And traditional server growth would come from price alone, with no rise in the number of machines shipped.

What to watch

The most important signal is new AI server orders against shipments. While fresh orders keep pace, the 95 billion dollar backlog holds or grows and the visibility underpinning the case stays intact. If orders fall behind, the backlog is being consumed rather than replenished, and that would show in the order figures well before it reached reported sales.

Alongside that, watch memory prices and the volume of ordinary servers shipped. If memory gets cheaper and Dell still sells more machines, the Barclays argument weakens; if sales slow as memory cheapens, it gains ground. Gross margin is the check on all of it. The case would be wrong if the backlog converted slowly or saw cancellations. It would also be wrong if gross margin kept sliding below 22%, so that surging sales stopped turning into the profit growth forecasters are now pencilling in.

The numbers

The same industry, for comparison

Research for information, not personal investment advice or a recommendation to buy, sell or hold any security: see Important information. Data as at 18 October 2026.