My feeling was that even if OS will win, real winners will be cloud providers and those who sell hardware to them to run all of that.

It's kinda similar to elastic search and similar OS that is just being run and managed on AWS.

Companies who train OS or proprietary also really heavy on computer provided by cloud infrastructure providers. Open AI relies on Microsoft.

Stability on Amazon. Adobe most probably also used AWS for Firefly. No public information though.

Now going back to arguments of OS vs proprietary

I do think OS has a chance. OS have a leverage where they become a common platform and standart on which all players collaborate.

One such story for me is Java Applets vs Flash vs Silverlight vs HTML

Where HTML is last man standing

And companies like Adobe with Flash lost because most of smart people worked elsewhere. Adobe was kinda alone against everyone(400 companies in web consortium) pushing web standarts. And it was open so that Apple could come in and adopt it to their new small touch screen form factor.

Same argument was made in that Google paper about moats. It was saying that winner is Facebook because LLAMA became platform for everyone to experiment, improve common platform, adapt to niche use cases. And it was happening faster then evolution of Google models.

Something similar happening with Stable Diffusion with things like ControlNets and other stuff.

Yes Firefly is impressive. But what will companies like Canva will use down the road? Dall-E or Stable Diffusion? Well they already are using Stable Diffusion.

So for me it just boils down to a phrase "Most smartest people are working elsewhere". Meaning that there is more smarter people outside the single company then inside, and if there is open platform that orchestrated their efforts then it's hard to compete.

Not sure how it all will evolve in thus case though.

There are questions of laws, regulation, legality of data sets and hardware requirements.