They're doing about a billion per month in revenue by running proprietary models on GPUs like these. Unless they're selling inference with zero/negative margin, it seems like a business model that could be made profitable very easily.
Revenue != profit, and you don't need to become net negative margin to be net unprofitable. Expensive researchers, expensive engineers, expensive capex, etc.
Inference has extremely different unit economics from a typical SaaS like Salesforce or adtech like google or facebook.
Currently, selling LLM inference is a red queen race: the moment you release a model, others begin distilling and attempting to sell your model cheaper, avoiding the expensive capitalized costs associated with R&D. This can occur because the LLM market is fundamentally -- at best -- minimally differentiated; consumers are willing to switch between vendors ("big labs", as you call them, but they aren't really research labs) to whomever offers the best model at the lowest price. This is emphasized by the distributors of many LLMs, developer tools, offering ways to switch the LLM at runtime (see https://www.jetbrains.com/help/ai-assistant/use-custom-model... or https://code.visualstudio.com/docs/copilot/customization/lan... for an example of this). The distributors of LLMs actively working against LLM providers margin provides an exceptionally strong headwind.
This market dynamic begets a low margin race to the bottom, where no party appears able to secure the highly attractive (think the >70% service margin we see in typical tech) unit economics typical of tech.
Inference is a very tough business. It is my opinion (and likely the opinion of many others) that the margins will not sustain a typical "tech" business without continual investment to attempt to develop increasingly complex and expensive models, which itself is unprofitable.
I don't disagree but you're moving the goalposts. I never said that they could achieve the profits of a typical tech business, just that they could be profitable. Also, the whole distilling problem doesn't happen if the model is proprietary.
> I don't disagree but you're moving the goalposts. I never said that they could achieve the profits of a typical tech business, just that they could be profitable. Also, the whole distilling problem doesn't happen if the model is proprietary.
In the absence of typical software margins, they will be eroded by providers of "good enough" margins (AWS, Azure, GCP, etc.) who gain more profit from the bundled services than OpenAI does from the primary services. This has happened multiple times in history, either resulting in smaller businesses below IPO price (such as Elastic, Hashicorp, etc.) or outright bankruptcy.
Second, the distilling happens on the outputs of the model. Model distillation refers to the usage of a models outputs to train a secondary smaller model. Do not mistake distillation for training (or retraining) to sparse models. You can absolutely distill proprietary models. In fact, that is how DeekSeek-R1-Distill-Qwen and the DeepSeek-R1-Distill-Llama are trained. This also happens with Chinese startups distilling OpenAI models to resell [2].
The worst part is OpenAI is already having to provide APIs to do this [1]. This is not ideal, as OpenAI wants to lock people into (as much as possible) a single platform.
I really don't like OpenAIs market position here. I don't think it's long term profitable.
Indeed. And even if that revenue is net profitable right now (and analysts differ sharply on whether it really is), is there a sustainable moat that'll keep fast-followers from replicating most of OpenAI's product value at lower cost? History is littered with first-movers who planted the crop only to see new competitors feast on the fruit.
These kinds of phrases are...eerily similar to the phrases heard right before...the .com bust. If you were old enough at the time, that's exactly what the mindset was back then.
The classic story of the shoeshine boy giving out stock tips...and all that.
Crazy that this (random blog post with no points and zero comments) is the only discussion of a major Ethereum upgrade on here. Even if the average commenter is firmly against blockchains this is still a very interesting and impressive feat.
No, being unemployed is the coercive factor here. It's not fair to treat at-will employment as non-coercive unless non-employment is actually zero. Non-employment currently stands at about 7.7%: https://www.richmondfed.org/research/national_economy/non_em...
Why would anyone turn down a chance to make a living if a job is offered? Why do you think the fed ensures that there are never enough jobs for everyone? Why do you think the fed and the business world talks about the economy in terms of "jobs" and "unemployment" when these are metrics largely unrelated to stuff like "am I actually getting a fair wage" and "is housing priced anywhere near rationally"? etc—the non-coercive labor market is a complete illusion.
Not only is Western Union significantly less reliable and significantly more expensive, as someone who gets paid in crypto the most expensive and painful parts of my finances are those that touch the traditional banking system at all.
> significantly less reliable and significantly more expensive
Going to need a citation on those two claims. You can lookup fees for both, right now BTC fees are about one USD more than a WU transfer. WU is subject to banking laws and if you're going to claim it's "unreliable", you're going to need to provide some evidence of that. It's literally their business, and a heavily regulated one at that.
Yeah, doesn't even make sense for something like this to be decentralized, it should be run by Rolex so they can catch edge cases and correct them by fiat
Yup. If your father dies, and leaves you a Rolex, but does make a way to give you access to the keys to the nft, would the Rolex stop being authentic?
When we are talking about digital proof of ownership of physical goods, any solution that can fail synchronization of the record due to entirely foreseeable events like death, ownership disputes, etc is a failed solution.
Good remark. I think it is simpler to just use a public blockchain that having Rolex properly build "open" cloud services to do that and maintain it themselves over the long run.
>It is certainly possible the court will disagree with the SEC, but I wouldnt bet on it.
I would (not a lot, but I would). They lost on the Grayscale trust, they lost on XRP being a security (extremely relevant here), they're seemingly settling with Binance... the court has been disagreeing with the SEC a whole lot on the matter of blockchain regulation lately.
No, you aren't, and it isn't vocal fry. Overtone singing is a distinct technique in Tuvan throat singing, and comparing it to a bandpass filter is accurate — as a whole, Tuvan throat singing is a set of techniques designed to induce vocal sounds with extremely rich harmonics, which can then be modulated and selected for by shaping the mouth.
What you're thinking of is called "Kargyraa", a particular subset of Tuvan techniques that involves singing with the vocal cords as normal, but also tightening the voicebox such that your "false vocal cords" (some flaps somewhere in your throat) are struck at a frequency an octave below the sung note — for every full cycle your vocal cords complete, the false vocal cords complete half of 1. It creates a rich sound which can be useful for the "bandpass" technique, but is fundamentally something different.
Take this with a small grain of salt, I came to this technique through the modern beatboxing community who independently discovered it as the standard "throat bass" and only bothered researching the Tuvan equivalent a long time ago.
There is a bass singing technique called "subharmonics" that uses something similar (identical?) to vocal fry to create interference with a sung note for a similar effect.
Are there actual perverse incentives that come with "it being married to crypto" — keep in mind that there are demonstrable positive incentives — or is it just aesthetic distaste?
There is a demonstrable disincentive in that approximately everything related to crypto is a scam, so by being married to crypto they demonstrate at least a lack of good judgement.
In which Theodore Blackman, CTO of the Urbit Foundation, responds to the "Plunder and Urbit" piece that was submitted here yesterday (https://news.ycombinator.com/item?id=37057471).
How is this published? On a single server somewhere? What country would you put it in? A friendly one I hope. Also one that won't block access to other countries that it might have petty disagreements with in the future. It might be a good idea for it to be replicated across a number of servers across the internet. We call those "nodes" in the blockchain space.
