AI's Economics Don't Make Sense
Had the writer paid for her actual token burn, and had each of the times it got “stuck” resulted in $15 in token charges, I don’t think she’d be quite as forgiving of these fuckups.
You cannot guarantee whether an LLM will do a particular action, or whether it will present you with an outcome based on reality. You cannot for certain say how much a particular task — even one you’ve done many times using an LLM in the past — might cost, nor can you be sure when a model might go berserk and delete something, or simply not do something yet claim that it did.
Oooh oui, Ed marque vraiment des points, là. Quand je vois les montants quotidiens remontés par l'extension VS Code Claude Code Usage(ex : 85$ aujourd'hui pour une journée pas particulièrement intense), si je devais payer moi-même, ou même faire payer à mon entreprise ces montants quotidien, jamais je ne jugerai que ça vaut le coup.
On a troqué la barrière du temps et des compétences (je ne fais pas car je n'ai pas le temps et les compétences) contre une barrière financière. Mais pour le moment cette barrière financière est complètement invisible, derrière un abonnement vendu à perte.
[...] as the majority of AI users perceive its real or imagined benefits entirely through the lens of being able to burn anywhere from $8 to $13.50 for every dollar of their subscription in tokens.
À vue de nez, ça me parait pas déconnant comme ratio. Encore une fois, sans être un acharné.
The real thing to look out for — a true pale horse — will be a major AI lab like Anthropic or OpenAI moving all of its subscribers to token-based billing. Once that happens, you’ll know it’s closing time.
À suivre.
Sure, sure, you’re gonna say that engineers are “shipping code faster” or some such bullshit, I get it, but how much faster, and how much money are you making or saving as a result?
Franchement, c'est dur à estimer ! Je fais beaucoup plus de choses que sans, ça oui. Mais je fais aussi beaucoup de choses que je n'aurais jamais pris le temps de faire. Beaucoup plus de détours, beaucoup plus de tests. Quand à savoir si ce que je fais apporte une valeur réelle à l'entreprise, cette question est bien compliquée.
For the most part, everybody you’ve read gooning over the many possibilities of generative AI has experienced it without having to pay the true costs.
Vrai
Anthropic Head of Growth Amol Avasare said last week that its Max subscriptions were built for heavy chat usage rather than whatever people are doing with Claude Code and Cowork, and made it clear that Anthropic is now looking at “different options to keep delivering a great experience,” which is another way of saying “we’re going to change the prices at some point.”
I’m not sure people realize how expensive these tokens are, [...] What happens if they go over budget, or if they literally can’t afford to spend the money necessary to finish their work?
[The Economics of AI Data Centers And Compute Do Not Make Sense]
Pour la faire court:
- On installe des GW et des GW de datacenter qui sont difficilement profitable pour plein de raisons
- On en installe tellement qu'on ne sait pas qui va les payer
- L'économie des Data Centers d'IA repose à 50% sur Anthropic et OpenAI
- Le business nécessaire pour rentabiliser tout ce qu'il est prévu d'installer est beaucoup trop important (
$1.18 trillion in annual revenues) et ne sera pas réalisé.
By selling users a product on a monthly rate and creating habits based on its availability, companies like Anthropic and OpenAI have misrepresented their businesses in such a way that most of their users are interacting with products and building workflows based on products that are unsustainable and impossible to maintain in their current form.
Generative AI is an insult. It is unreliable, its economics don’t make sense, its outcomes don’t justify its existence, and the perpetrators of its con are boring, oafish and avaricious men disconnected from society and anybody who would ever disagree with them. It requires stealing art from everybody, destroying the environment, increasing our electricity bills, the constant threat of economic annihilation, the endless cacaphony of “everything fucking sucks now because of AI,” all to push software that can only be justified by people willing to ignore basic finance or sense.
The assumption, if I had to imagine, is that I’m simply being alarmist, and that “the demand will absolutely be there.”
You’d better hope you’re right.
Extraits cités issus de ce lien visité le , crédits: Ed Zitron