"What’s not so mixed is analysts’ reactions to the company’s forecast for the current quarter"
Analysts have yet to grasp that no market has infinite or even just continuous demand. They can't be all that bright.
Hewlett Packard Enterprise has reported a dramatic decline in server revenues as corporate customers cut back on spending amid a more cautious outlook. Compute, HPE’s largest division by far, declined 31 percent year-on-year to $2.585 billion for the three months ended October 31 [PDF], the company’s final quarter of its 2023 …
Also not grasped that the major cloud providers are building their own servers, rather than purchase from the majors. Whilst these are in the main x86 native and increasingly “x86 comparable” ARM, the design of cloud does seem to be following the RAID principle of using lower grade servers in redundant configuration is more cost effective and reliable than expensive proprietary servers. Additionally, with businesses still moving work to the cloud, there is going to be a massive downturn in server shipments.
Major cloud providers have been building their own servers for ~10-15+ years. Assuming of course when you say "purchase from the majors" being HPE/Dell/Lenovo/Cisco/etc (and not referring to the Asian "ODMs"). Just for reference the open compute project is apparently 12 years old, though major cloud companies were building their own long before that as well. Google has been building their own probably since they were founded.
Pulled these links from my retired blog from 2011.
Microsoft Reveals its Specialty Servers, Racks
https://www.datacenterknowledge.com/archives/2011/04/25/microsoft-reveals-its-specialty-servers-racks
Facebook Unveils Custom Servers, Facility Design
https://www.datacenterknowledge.com/archives/2011/04/07/facebook-unveils-custom-servers-facility-design/
It stands to reason that the continued growth of the cloud results from customers moving to the cloud and, therefore, no longer needing to buy servers from Dell/HP. I remember working for one of the big hardware providers and consistently hearing from customers that they were shutting multiple data centres and moving to the cloud, leaving maybe one colo. The writing was on the wall 6/7 years ago, and TBH, I am surprised the hardware vendors have fared as well as they have.
I still need to buy, i still have on prem hardware, we used to be 100% HPE. Paywalled firmware updates for out of support hardware means they can jump in a lake. Why would I pay for ongoing support for hardware thats only fit for lab experimentation after the 3/5 years of support it came with expires?
I realise firmware / driver updates cost HP money but surely this should be built into the original equipment cost as is the case with most other suppliers? Maybe if they wernt "agile" then firmware would arive with less bugs and security flaws?