As AI gets more expensive, popularity of Chinese AI models like DeepSeek is growing fast, even though these models are not as good as ChatGPT, Claude and Gemini. Industry experts believe the cost pressure is real and most companies will move to a hybrid model for AI use.
Oh don’t mistake me, they are not consumer friendly.
They are just trying to sell enterprise GPUs directly to “consumer” businesses and the cloud providers they use, instead of through literally fraudulent middlemen like OpenAI.
This is what pretty much everyone with hardware is doing, including Huawei, Tenstorrent, Cerebras, even AMD. Maybe I misinterpreted you, but hardly anyone cares about individual self-hosters.
Apple does, though. MLX is actually getting pretty cool. But they’ll always be quite insular, anti-consumer in other ways, and they still seem detached from what the community is largely doing.
My view is that we’re basically in the mainframe era of AI, but local models are already getting good enough to do useful stuff. Qwen 3.6 in particular is very capable, and you can do real work with it. So, extrapolate this into a couple of years into the future and it’s almost certain that we’ll be able to run models that perform as well as current frontier models locally. And that means companies are going to be much more likely to self host as well. In fact, I think you’re completely right that the immediate target will be business customers that want to self host their own models before this tech really gets to consumer grade.
Yeah. I mean, I have a Ryzen desktop and a 2020 GPU, and Mimo 2.5 is a bit faster and mind bogglingly better than frontier models from like… two years ago? And frontier models are plateauing, I think.
Still, my worry is that we consumer won’t HAVE any hardware. Many don’t even own a laptop these days, and it feels like they’ll just drop desktops (and work will just use thin clients) if they’re too cost prohibitive for people to buy.
I guess gonna have to hope that Chinese companies ramp up production soon. Might have to smuggle that hardware in though at the rate things are going.
Of what, though? Huawei NPUs are datacenter hardware.
As much as we hate it, Nvidia gaming GPUs are ultimately cheap consumer devices, and they’re very good at hybrid CPU+GPU inference.
I think Intel has the best chance of pulling a rabbit out of a hat with Arc. They have a usable platform already, hardware “close enough” to Nvidia that LLM compatibility isn’t a nightmare. And they have nothing to lose, no illusion of “protecting datacenter cards” like AMD has.
Chinese companies are very much ramping up production fo consumer devices right as we speak. I expect we’ll see the same thing we saw with stuff like solar panels and EVs in the coming years. https://www.techspot.com/news/112529-china-first-credible-gaming-gpu-sells-30000-units.html
Doesn’t matter(for this, specifically) if it’s not performant on LLM inference engines.
And I’m not just talking about CUDA. Even GGUF Vulkan (for example) has all sorts of vendor quirks that can absolutely trash performance. VLLM is often a joke on AMD, with certain models, on certain cards, even with dev support.
Sure, but try extrapolating 2 or 3 years into the future here. Models are going to become more efficient and hardware is going to improve. Right now Chinese companies are just starting to put out GPUs, but once that process is ironed out, I don’t see why they wouldn’t put out chips that work well with Chinese models. This kind of stuff is happening already, it’s only a matter of time till it makes it to consumer market. too https://lushbinary.com/blog/deepseek-v4-huawei-ascend-ai-infrastructure-strategy