That's my take, at least.
Nemotron is TERRIBLE, and purposefully so. It must be.
They cannot be THAT BAD at training AI models. I don't believe it.
Smaller customers are also less able to develop their own hardware and threaten NVidia's business.
I think NVIDIA does want small open models running on-prem to explode as a market! Lots of smaller GPU installations for companies who wisely want on-prem inference.
Of course NVIDIA will also keep making a ton of money selling to hyper scalers, but not forever: Chinese chips are getting better, Google, Microsoft, Amazon, etc. designing their own inference chips.
NVIDIA is handling this brilliantly.
No one wants to pay the Nvidia tax
NV's long-term strategic incentive in funding a semi-open model provider like Reflection.ai is to ensure competitive frontier models which fully leverage the NV proprietary stack (chips, interconnects, servers, CUDA) continue to be widely available and continue to offer performance worth a higher price to the most profitable market segments.
NV's ~75% margins on hardware(!) at >$100B/yr scale are historically unprecedented and still increasing, creating tectonic pressure on NV's largest customers (hyperscalers and frontier labs) to escape the "NV Tax" by gaining access to competitive frontier chips, servers, and/or middleware at lower margins. Why would NV help create semi-open models that threaten their best customers? Because as Jeff Bezos famously said, "Your margin is my opportunity" and that's turning NV's biggest customers into their largest existential threat.
At the moment, NV's moats blocking significant competition are almost unimaginably deep, but on a decadal time-scale, literal trillions of dollars are at stake. That's enough to get people thinking the unthinkable, making NV the biggest target in modern business history. It's to the point that it's almost "Everyone against NVidia" which is forcing all the big companies into playing 3D strategic chess on multiple time horizons at once, simultaneously working with, investing in and hedging against each other. It's a 'co-opetition' (https://en.wikipedia.org/wiki/Coopetition) race where the smaller players are grouping into tactical alliances and uneasy truces while the biggest players are spending billions to 'commoditize their complements' (https://gwern.net/complement) as NV is doing with Reflection.ai. This can create strange bedfellows overnight. I wouldn't be surprised to see some of NV's biggest customers, who compete fiercely against each other, pooling resources with NV's competitors to create a viable alternative to NV. This is the stuff Jensen has nightmares about, waking up in a cold sweat in his black leather pajamas.
NV doesn't need Reflection to be better than the best models or even be profitable. They just need to ensure a viable alternative to frontier lab's proprietary models: A. Remains widely available at low enough cost for all NV's other customers to buy, B. 'Works best on NVidia', and C. Stays close enough in price/perf to prevent any single proprietary model becoming as dominant in models as NV is in hardware. The Chinese semi-open models have been strategically convenient for NV but it'd be foolish to count on the Chinese govt continuing to subsidize them or Chinese models not getting blocked or limited by some governments. If it only costs a few billion, Reflection.ai being "good enough," especially for a semi-open, (near-)free, US-based model, is a cheap strategic hedge against long-term threats to NV's (near-)monopoly, especially when Jensen has trouble finding room to store the mountains of cash NV is piling up. When your margins are ~75%, there's literally no better place to put money except toward extending your dominance.
Right now US economy is basically a centrally planned economy.
More on-topic: I remain amazed by how the past quarter century of American legislation has included one overwhelming boon for corporations after another. Once corporations could give an infinitly higher amount of political contributions when compared to individual citizens, it became clear that the overall well-being and happiness of corporations was the only real political focus. Anything that doesnt directly or indirectly benefit a group of corporations is almost always designed to manipulate the electorate through invented/perceived/embiggened/non-cromulent "social issues".
What's the mechanism for this? Genuine question: I read an overview over Sarbanes-Oxley and it seems like a reasonable idea.
> The major film studios owned the theaters where their motion pictures were shown, either in partnerships or outright. Thus, specific theater chains showed only the films produced by the studio that owned them. The studios created the films, had the writers, directors, producers and actors on staff (under contract), owned the film processing and laboratories, created the prints and distributed them through the theaters that they owned: In other words, the studios were vertically integrated, creating a de facto oligopoly. [...] Ultimately, this issue of the studios' then-alleged (and later upheld) illegal trade practices led to all the major movie studios being sued in 1938 by the U.S. Department of Justice.
https://en.wikipedia.org/wiki/United_States_v._Paramount_Pic....
Nowadays it seems like "vertically integrated" is something most companies openly aim for.
The investment was out of necessity, not interest in monopolies.
Writers and directors craved an audience, so they appealed to studios for funding. That’s opportunity meeting, not monopolistic practices.
Did you know that 151 companies were founded by SpaceX's alumni? Compare to 5 they've acquired (not including XAI)
similar to how eventually AWS started making their own chips for data centers, and Apple did that for their hardware, it's not a ridiculous thing to plan for the AI companies to start making their own chips to optimize for their use cases and cut out the middleman for margins.
But, not really at the current technologies. Kimi and GLM are fucking awesome, but I don’t have 3TB of VRAM to run them, and I don’t expect to even when ram prices drop.
So now you’re back to the scaling issue before talking about power and compute distribution.
They can't ride the hyperscaler gravy train forever; at some point between Google, AMD, and Apple NVIDIA is going to lose its monopoly on serving large customers.
At that point, it would be useful if a few open models existed which were only a couple of months behind the frontier.
But it's very important that the open models never be TOO good, because the AI companies are buying compute on the assumption that their software will add value. If it becomes a commodity business with frontier open models, NVIDIA won't be able to get away with such a crazy markup.
Of course they would. You need hardware to run the model. nVidia sets the floor.
I can see one of Nvidia's biggest fears is the inference hardware becoming commoditised.
are they actually suppressing the western open models?
china doesn't give a fuck either way
imo, they see the weakness emerging at the intersection of all the labs, everybody knew there was no moat, so they're gonna control its direction and basically tell the Jev guys what they want them to work on
They are just trying to grow the pie because they have nobody else competing for slices.
My guess is they want as many frontier models using their chips as possible. The only threat to their business is companies making their own chips which Google does and the others are working toward. The last thing they want is only 3 frontier labs who are all not buying NVIDIA.
Hugging Face - distribution for model that you can run on your local Nvidia Spark
Neoclouds - Nvidia setup a 500 billion investment fund with Wall Street so Nvidia can sell chip and this news about buying a LLM start up. it seem like another customer for Nvidia.
correct me if i'm wrong but i remember i saw an interview with Jensen where he want more company to have their own model and country to have their own LLM model.
Nvidia doesn't make money from the gold rush. Nvidia made money from selling the shovels.
Just from basic maths, the GeForce line (except 5090) has and continues to be extremely ‘subsidised’ at street prices compared to their DC lineup; in terms of margins, and $ they’d make instead of using the same fab or memory allocation for e.g. a RTX Pro or higher; which still have massive demand.
If you want 32GB and CUDA, 2x 5060ti 16gb and tensor parallelism 2 is pretty easy to set up in any ATX PC case.
They are a greedy mega cap business like everyone else. But if you rationalise it a bit, if they make geforces too compelling for scalers and neoclouds with more VRAM, the margin shortfall would be in the hundreds of billions.
At least they’re not exiting the consumer market altogether like other players, and at least they continue to deliver on software.
But we’re getting ahead of ourselves