Dell breaks records, says "traditional" server demand as well as AI surging

Dell saw record orders, shipments, revenues, and backlog in Q4, as AI demand continued to drive strong OEM activity – leaving demand “significantly” outpacing supply even for “traditional” CPU-centric servers.

OEMS like Dell are facing an environment in which cloud hyperscalers (which are set to allocate nearly $700 billion in CapEx for data centre expansion this year) are mopping up semiconductors, SSDs and network kit alike – leaving downstream channel partners facing volatile pricing. 

“Traditional x86 is benefiting from AI infrastructure build-outs. While many AI workloads rely on specialized GPUs, traditional compute remains essential for orchestration, data processing, and inference support. As customers deploy AI, they are modernizing broader IT estates, refreshing and expanding general-purpose environments,” said co-COO Jeff Clarke.

Dell sees "sustained supply tightness"

On Dell’s supply chain, he said “unprecedented AI demand” continues to create broad and “sustained supply tightness and frequent pricing resets… 

“We are executing our operating model with urgency, securing supply as the first priority,” Clarke added, telling analysts “We are managing this environment in real time, applying lessons learned from prior cycles to improve resilience and to strengthen our position…”

Just over 4,000 customers are responsible for $64 billion of AI-related revenue at Dell, with inference growth and strategic pricing helping it weather a memory shortage storm, executives said on the call.

$34 billion in AI hardware orders

The company continued its AI shift during Q4 of FY26, signing $34 billion in AI orders over the quarter and hiking prices to meet a supply environment “as tight as we have ever seen.”

In an earnings call Thursday, COO Jeff Clarke said: “Inference is driving more tokens; tokens drive more compute capacity and intensity. And ultimately that is good for the revenue stream of the company.”

See also: “Tokens are profitable” says NVIDIA CFO, as Vera Rubin ships

Demand saw the customer base for Dell’s AI products tip over 4,000 during the quarter, with growth in neocloud, sovereign and enterprise customers as it recorded a record overall revenue high of $113.5 billion for the year.

AI server revenue, separated out for the first time, made up almost $9 billion of Dell’s $33 billion net revenue in Q4. According to its projections, Dell’s AI server business is expected to grow 103% in FY27.

What memory crisis?

Despite the growing impact of DRAM and NAND shortages across the market, with Gartner predicting a 130% increase in combined DRAM and SSD prices in 2026, Dell's revenue was largely unaffected.

Clarke said Dell had used lessons learnt during the COVD-19 pandemic to weather the storm, benefiting from long-term deals and budgetary pricing.

He explained: “We are working through things of how do we minimize our complexity, how do we improve our mix, how do we sell what is coming, how do we improve our designs to take whatever parts that are available.”

Price hikes

Following Clarke’s comments about bringing costs to customers in November, Dell also changed pricing for its entire server business in December and brought in a hike for PCs in January 2026.

Dell’s Client Solutions unit, covering end-user computer hardware, will see similar pricing changes this year, he said.

Shipping Vera Rubin

Dell execs also fended off concerns that the rollout of servers running on NVIDIA’s new Vera Rubin hardware, which NVIDIA said will ship in the second half of 2026, would be as bumpy as the Grace Blackwell rollout.

CFO David Kennedy said Dell had “a lot of lessons learned” from its previous experience, particularly on manufacturing and testing in new architecture, that would let it ramp with “more velocity and speed”.

Dell was one of several server manufacturers that dealt with overheating and connectivity issues while integrating Grace Blackwell chips in 2025.

While the CFO said there were no Vera Rubin orders in its current backlog, Clarke assured the hardware was included in its five-quarter pipeline, though the majority of the pipeline was the previous generation, “driven primarily by enterprise deployment.”