> Gosh, I have a coworker who acts like gaming GPUs are the only hardware that matters.

> I've tried explaining that one or two AI data center clients for Nvidia dwarfs the entire gaming GPU market, but he just doesn't get it.

I have a feeling that the different judgements come from the fact that the coworker thinks that the AI bubble will soon burst - thus, in his judgement, the AI data center sector of Nvidia is insanely overvalued, and will collapse. What will "save" Nvidia then will be the gaming GPUs. Thus, in his opinion, this is the sector that matters most for Nvidia, since it will become Nvidia's lifeline when (not "if"! - in your coworker's judgement) things will go wrong in AI.

You, on the other hand believe AI data centers are here to stay (which is a bold assumption: it could happen that AI will move more to the edge), and no big competition will arise for NVidia for "big AI ASICs" (another bold assumption). Your judgment is based on these two strong assumptions about the future, while your coworker's is based on different (possibly similarly bold) assumptions.

His coworker simply has no idea that datacenters provide 93% of Nvidia's revenue, and a single B200 GPU sells for $40k.

> His coworker simply has no idea that datacenters provide 93% of Nvidia's revenue, and a single B200 GPU sells for $40k.

Currently. :-)

Real beats imaginary, though.

The company will unlikely make it through a collapse like that. It’s like Netflix going back to DVDs via mail.

The market is so tiny that their capex investments into AI stuff would catch them with massive debt that the gaming revenue couldn’t support and they would have to go through bankruptcy

Data centers and enterprise will always dwarf consumer hardware. AI has nothing to do with it.

You're silly if you think otherwise.

> Data centers and enterprise will always dwarf consumer hardware.

Before the current AI hype, except for some rather specialized applications, people had rather little use for GPU acceleration (GPGPU) in data centers.

Again, you're focusing on AI and GPU. Datacenter contracts have always been the bulk of sales for companies like Nvidia.

One or two large datacenter contracts will make more money than all consumer hardware for companies like Nvidia. The margins are completely different.

Consumer hardware is where the innovation slowly starts though. Without the years of people messing with GPGPU on consumer hardware we might not have got the AI revolution on GPUs.

Around 15 years ago in college, I took the opposite position that datacenter workloads mattered more to Nvidia’s future than gaming and no one believed me. It is amazing how times have changed.

I used to work for Oracle in their datacenter hardware module.

The scale of the orders that datacenters put in is insane. It's literally 1000:1 vs consumer stuff.

While building a single server I could easily handle 2-3 million dollars worth of hardware. We'd roll out 50 of those in a week. That was just my shift and we had several.

Gamers and consumers simply dont understand the scale that datacenters work at.

Sort of reminds me of ios vs macos. I'm pretty sure the ios market dwarfs macos.

EDIT: brief search says last year apple sold 300m ios vs 20m macos devices.

AI makes a lot of money. But games are what matter