Consumer harm comes when there is intention to control consumer choice. Since Linux isn’t the biggest platform, but a major choice: It is therefor harmful for Nvidia to not support it properly.
Check your bias.
Consumers don't generally rent server space. It would be difficult to establish consumer harm on the basis of server prices.
> Furthermore plenty of people run Linux desktops that would be affected by this
Right, this is the insignificant bit. Inconveniencing 2 or 3% of the market is not a valid antitrust claim [1][2].
> Consumer harm comes when there is intention to control consumer choice
No, it comes when you can prove prices were raised, output reduced, innovation diminished or customers were "otherwise harmed" [3]. To the degree intent is considered in the enabling case, it's in reading the intent of the Congress, not the defendant [4].
[1] https://en.wikipedia.org/wiki/Usage_share_of_operating_syste...
[2] https://gs.statcounter.com/os-market-share/desktop/north-ame...
[3] https://www.americanactionforum.org/insight/why-the-consumer...
You are also not counting handheld use such as the Steamdeck. There is a reason Steamdeck doesn't use NVIDIA graphics.
GNU/Linux desktop is a different story. Yes it's accurate to say that's an insignificant market share. And I said "GNU" to differentiate from the most popular Linux desktop OS, ChromeOS.
I'm not interested in the OS fan wars and neither is Nvidia, but if you want to consider Linux server users the same as Linux desktop users instead of isolating them, you can count me on that side. I have an RPi, a PowerEdge, an Android phone, and yes an Alpine Linux Docker container on my Mac.