Is pricing due to AWS massively overcharging because they know Enterprise might pay it due to ease of use with purchasing? Or is this realistic pricing for LLM coding assistance and AWS has decided to be upfront with actual costs?
The future is not that AI takes over. It's when the accountants realize for a $120K a year developer, if it makes them even 20% more efficient (doubt that), you have a ceiling of $2000/mo. on AI spend before you break even. Once the VC subsidies end, it could easily cost that much. When that happens... who cares if you use AI? Some developers might use it, others might not, it doesn't matter anymore.
This is also assuming Anthropic or OpenAI don't lose any of their ongoing lawsuits, and aren't forced to raise prices to cover settlement fees. For example, Anthropic is currently in the clear on the fair use "transformative" argument; but they are in hot water over the book piracy from LibGen (illegal regardless of use case). The worst case scenario in that lawsuit, although unlikely, is $150,000 per violation * 5 million books = $750B in damages.
I don't think businesses sees it this way. They sort of want you to be 20% more efficient by being 20% better (with no added cost). I'm sure the math is, if their efficiency is increased by 20% then than means we can reduce head count by 20% or not hire new developers.
In many workplaces this is true. That means an "ideal" workspace is 20% of the size of its current setup, with AI doing all the work that the non-core devs used to do.
Source? Dario claims API inference is already “fairly profitable”. They have been optimizing models and inference, while keeping prices fairly high.
> dario recently told alex kantrowitz the quiet part out loud: "we make improvements all the time that make the models, like, 50% more efficient than they are before. we are just the beginning of optimizing inference... for every dollar the model makes, it costs a certain amount. that is actually already fairly profitable."
I suspect that the answer is almost all of the training data, and none of the weights (because the new model has a different architecture, rather than some new pieces bolted on to the existing architecture).
So then the question becomes, what is the relative cost of the training data vs. actually training to derive the weights? I don't know the answer to that; can anyone give a definitive answer?
GPT-5 is a bellwether there. OpenAI had a huge head start and basically access to whatever money and resources they needed and after a ton of hype released a pile of underwhelming meh. With the pace of advances slowing rapidly the pressure will be on to make money from what’s there now (which is well short of what the hype had promised).
In the language of Gartner’s hype curve, we’re about to rapidly fall into the “trough of disillusionment.”
Whereas others are willing to loose a bit to get ahead in a competitive space like GenAI, AWS has been freaking out since if the cloud gets driven to a low margin commodity then Amazon is in big trouble. This they’ll do everything possible to keep the margins high, which translates to “realistic” pricing based on actual costs in this case. And thus yes seemingly hoping enterprises will buy some expensive second rate product so they can say they have customers here and hoping they don’t notice better cheaper offerings are readily available.
This is also a signal what what the “real cost” of these services will be once the VC subsidies dry up.
Sooooo true. Waiting for AI’s “uber” moment aka when uber suddenly needed to actually turn a profit and overnight drivers were paid less while prices rose 3x.
GCP is only big three that is much cheaper, clearly a play to get customers since they are in distant third.
Amazon / AWS might not want (or need) to play that game.
Obviously, they are not taking VC money but using revenue from other parts of the company.