It is great to have an option to enable certain features for a fee after the purchase. I'd like to buy the most basic CPU and then enable features as needed instead of buying the most expensive CPU right away just in case. If they see it as economically viable, then what's the issue?
This is a genuinely good question.
I personally do not object to this practice in principle. What I do object to, is when that is done in order for one market segment to subsidize sales in another market segment: that play only works when there is not enough competition in the market, or when the sellers collude against customers. In my experience, that play usually also involves skirting the law and copyrights... I do not believe Intel fits in that category at all, and they must base their plan on something different.
For complex equipment, the design and manufacturing costs can easily be such that splitting the product (with run-time configuration the only difference) makes business sense for both the vendor and the customer. Simply put, the larger volume allows price such savings that selling any of the sub-products alone would not be (as) profitable.
We see this often in high-end test gear and scientific equipment. Intel already does this type of splitting, by testing and binning its silicon, and selling them as different models. Now that AMD is once again manufacturing processors that are competitive with Intel ones in both price and performance, I don't see how Intel is going to benefit from this.
(Interestingly, if we look at Nov 2021
Top 500 supercomputers, the top five use processors from Fujitsu, two IBM (Power9), one SunWay, and one AMD Epyc. There are two Intel Xeon and four AMD Epyc based ones in the top 10. At these scales, the contracts with the vendor are obviously special, and we shouldn't draw any hasty conclusions. Indeed, the
Nov 2021 highlights tells us that 81.6% of Top500 still runs on Intel (but down from 86.4% since May 2021), and 14.6% run on AMD (up from 9.6% since May 2021). So, when I am talking about "AMD is once again competitive", I mean the situation has significantly changed in 2021, from say 2020 or 2019.)
This leaves me puzzled.
Every submodel has to be competitive (in terms of both price and performance) with respect to the competitors, especially AMD, possibly ARM-based ones too, or customers will just use the competitors processors instead. I do not see how Intel intends to pull that off in practice. It's not like people by server farms
first and then seek use cases for them; the need exists before the purchase, in this business domain.
My guess is that a convincing salesdrone has hoodwinked the Intel board to try this, with some fancy calculations on how this will let them somehow subsidize Top 500 or some other server sector purchases, and buttress their business against competition somehow, and will somehow be profitable.
Perhaps; I just don't see how. Those salesdrones are much better at getting paid and moving on before the actual results are realized.