"Speaking with analysts on Intel's Q4 earnings call Thursday, CFO David Zinsner admitted the company was caught with its pants down after it misjudged demand for its datacenter products, leading to a capacity crunch during the quarter. Zinsner said six months ago "every hyperscale customer" was sending signals they planned to order a smaller number of high-core count chips. They soon changed their tune and demand for Intel's Xeon products increased considerably over the third and fourth quarters."
"To meet that demand Zinsner said Intel is " shifting as much as we can over to the data center." But don't worry, the exec promises the x86 giant won't abandon its client business "completely" to chase AI revenues in the datacenter. "Within the client, we're focusing on the mid-and-high end, and [we're] not as focused on the low end. To the extent we have excess [capacity], we're pushing all of that into the datacenter space," the CFO said."
Intel is reallocating foundry capacity from client chips to Xeon processors to meet surging AI datacenter demand. The company misjudged datacenter demand, producing a capacity crunch after hyperscale customers increased orders for high-core-count Xeon chips. Xeon 6 commonly serves as host CPUs in GPU systems like Nvidia DGX B200/B300 and in AMD Instinct-based GPU boxes. Intel is shifting as much capacity as possible to datacenter products while focusing client efforts on mid- and high-end Core-series parts and deprioritizing low-end processors. Major memory vendors are also redirecting capacity to high-margin DRAM and high-bandwidth memory for AI, driving steep consumer memory price increases.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]