[sitkack]

The fact that the market can still do this (reduce supply to operate in the high margin quadrant) is a temporary blip due to the x86 duopoly. Soon anyone will be able to supply cycles/$ at a better rate than the entrenched players.

[cypress66]

On the gpu side there is no "x86 duopoly" and the situation is arguably worse than cpus. So that's probably not it.

[wlesieutre]

Nvidia and AMD have been a GPU duopoly just as much as Intel and AMD have been for CPUs.

Intel's new Arc product is the first competitor from another company in a very long time.

[cypress66]

But anybody's free to enter, there is no x86 IP in the way, as parent mentioned.

[szatkus]

It would take a lot of time and money to create anything competetive with Nvidia and AMD's offerings. It took Intel over a decade and their dGPUs still aren't very good.

GPUs are insanely complex and the drivers are full of hacks (even per game basis, a driver could swap shaders with ones written by Nvidia/AMD if it detects you're playing the newest popular title). The threshold to enter the market is incredibly high.

[philistine]

This is far from the truth; oftentimes the joke with GPUs is that you're buying a hardware driver with a device to unlock it.

Intel is hitting that problem head-on; no game is written to take advantage of its new GPUs, and has to suffer a ton for it. It's not a guarantee Intel will ever be able to escape this problem.

[sangnoir]

From my limited outsider perspective, it appears like there are way more CPU and FPGA startups than GPU startups - why is that when GPUs seemingly have lower barriers?

[jmalicki]

Most of the related startups are targeting the much higher-margin deep learning training accelerator portion of the GPU system market, and there are a lot there:

https://www.ai-startups.org/top/hardware/

[touisteur]

It's also a quite explored and already deeply optimized area.

Companies like TensTorrent, Graphcore and Lightmatter (for example) try to hit different spots of the architectural landscape (seemingly all optimising for 'cheap to tapeout' in all sorts of clever ways, but you still have to code for it (even though you often get python APIs and deep learning tensorflow/torch support, but that's probably not getting you to to max perf). Very interesting to watch, and hopefully one can get their hands on that kind of hardware and build a community around it.

[garaetjjte]

As I understand it GPUs are patent minefield.

[echelon]

Nvidia has a pretty tight stranglehold on deep learning thanks to CUDA.

We really need the forces that be to change this.