> a model that runs on my own machine will never have the capacity of a model that runs in a datacenter.
I don’t think so. A local run model only needs to serve one or a few people. It seems possible to run a DeepSeek v4 model at full capacity on a server costing 200k usd. Very expensive but not impossible.
Factor in hardware and software improvements over time, and the fact that most people may just need to run a smaller and quantized model, it should take a pc at 10k usd scale.
> Releasing open weights seems to be an effort to slow down competition.
Why? I don’t get it. Open weight models enabled a lot more foundational model trainings. I believe R1 was benefited from llama.
The proliferation of alternative, open weight models in turn put heat on the leading labs and forced them to make better models, or squeeze more out of slowing improvements on model capabilities through innovation on harness.
And eventually we the common people benefit, exactly from these competitions
Instead we got home solar system become very affordable over the past decade.
And driving out US manufacturers isn’t even the main goal for China. They know their huge risk on reliance on petroleum and was doing everything they can to mitigate that. Building out a huge solar manufacturing base is their answer. Now they are reducing petro imports YoY.
But the US was a leader in manufacturing solar panels and could have facilitated the same price decline trajectory. Selling a product in a market below manufacturing cost specifically to bankrupt domestic manufacturers is an illegal trade practice under the WTO called dumping.
If you see this as people just want free stuffs you are missing the point. There is an entire different level of danger to allow a few entities to completely own a day to day technology.
Imagine personal computers and open source operating systems completely don’t exist. What you can do on a pc today, you have to always do it on a thin client always connected to a main frame run by a selected few. Everything you do is recorded and subjected to at least 30 days of retention and inspection.
Imagine every car is completely not customer serviceable and have to be connected to one of the three manufacturers in order to operate. Everywhere you go is recorded. The manufacturer may decided the place you want to go is inappropriate for you to go.
Most of the cloud platforms are open source. Linux, container, k8s… it’s entirely possible for someone to build and deploy their private cloud if they have the resources.
> and eliminate egress fees
What does it mean? If I sign up for cloud service I am only bound to the contract terms. If I am PAYGO I can switch anytime.
> Most of the cloud platforms are open source. Linux, container, k8s…
Linux isn't a cloud platform and neither is Docker.
Kubernetes was created specifically to create a way in against AWS' de facto public cloud monopoly. The Cloud Native Compute Foundation is a classic "alliance of smaller players uses open-source and interoperability as a wedge against an incumbent that threatens hegemony".
> it’s entirely possible for someone to build and deploy their private cloud if they have the resources.
What is there for that, really? Basically just OpenStack?
I never say these pieces are the equivalent of a cloud platform. What I said is that with them, it is entirely possible for someone to build their own.
Why someone wants to do that is not my concern. But not being able to is.
More like Anthropic’s priorities are not everyone else’s priorities. They are in the consistent culture of being in absolute control and dictating what is good and bad, while taking any opportunity to trash and crush potential competitors (open source models happened to be mostly developed in China). All these in the name of safety and anti-authoritarian.
The day self hosted models catch up with Anthropic’s capabilities is when they will fully lose their shit. This day can’t come soon enough
If price is fully discovered right after ER then you will see price stabilized right after ER. But in fact post ER prices can wildly differ from the next minute, next day and next week price. It’s speculation and anticipation.
You cherry-picked examples. Counter examples would be:
Crowdstrike up 1067%
Cloudflare up 1408%
Robinhood up 148%
For an index fund, I would take a break even or even a slight loss on AirBnb to get those Crowdstrike and Cloudflare returns. I do agree with the overall sentiment that the foundation AI companies are overvalued but the whole point of an index fund is not have to analyze each individual company.
You examples are great… at supporting my points. CRWD ipo’ed at 7bn and NET at 5bn. Unlike the mega IPOs like abnb and uber and now 10x worse spacex. There is very limited upside left after high valuation milking.
By the time they drop and being kicked out (if they do) the insiders already dumped their shares. Not to mention now all the index fund holders will rush to sell creating even more price pressure.
Shorting (itself being a bad idea for regular investors) also breaks the mantra of passive investing, 401k or otherwise. It’s almost impossible to short right after IPO because of low float and high margin risk.
These mega IPOs are just using passive investors as backstop.
I don’t think so. A local run model only needs to serve one or a few people. It seems possible to run a DeepSeek v4 model at full capacity on a server costing 200k usd. Very expensive but not impossible.
Factor in hardware and software improvements over time, and the fact that most people may just need to run a smaller and quantized model, it should take a pc at 10k usd scale.
reply