Tanj, are you saying that there aren't enough engineering resources? In what areas? Can you expand?
Imagine that you are in the business of kitting out data centers - the dominant market for servers. You, and your competitors, have spent the last 10 years balancing your resources so that you have just enough engineers to design the boards, chassis, network, racks, power, etc. that allows you to have a full catalog of VMs and container types for the market. You have been lulled by 6 years of crawling progress at Intel which has allowed you to actually reduce the number of new designs your teams make because, heck, what is new.
Then AMD comes along and you get busier. First you sample Rome. Then, darn, you realize it is for real a good CPU and AMD has some momentum. You scramble to hire a few more, and bring them up to speed. Maybe you tinker with Ampere or in-house ARM, maybe RISC-V. Darn, you are spending a lot on engineering all of a sudden, and ouch it is hard to recruit for hardware.
But that barely compares to the horror of realizing that there is now a land grab going on in AI, both training and inferencing, and the machines you desperately need to stake a claim in this new world are nothing like the ones you were just strolling along with last year. You need a whole bunch of Nvidia but you want to have some edge so you pillage your prior projects to put all your aces on figuring out how to have an advantage. Cooling, networking, host servers, reliability, memory, storage, more networking, privacy, partitioning, assigning and tearing down VMs in entirely new ways. Finding the sites, deciding which other assets they need to work with. And your hiring is almost all going to be backfill because it is not like there is a pool of engineers who know both the AI modules and know your cloud - no your best bet is to reassign some of your best people, maybe even shelve a project or two to get some entire teams who know how to work together. You may find a few great hires to add to them - probably recruiting from competitors like they recruit from you, because where else? The rest of the new hires are going into patching up the traditional server projects you had going, though a few of those are rationalized below the cut list.
And as for that capex you were going to spend on the new ordinary servers, well you need capex for the hot newness in AI and inferencing and LLM and .. oh, thankfully the customers are more interested in those too.
To be clear, I have been out of that world for nearly 3 years now, while AI was just beginning to curve up but had to fight for budget. I don't ask my old buddies what they do now, no insider knowledge. Just guessing how it goes when $10Bn or more of new stuff is suddenly the highest priority on a finite org with a serious need for competent engineers.
