Post

Rick Beato had a video on how this sort of thing could happen with AI and music. Telecoms had a monopoly over physical infrastructure. Tech companies pushing their luck on AI pricing don't have that luxury. Especially as on device AI solutions emerge:

just all in a mad dash for the trough without thinking things through, as this weeks Google and Meta fuckups attest

DeepSeek is currently providing their models at an insanely low cost for exactly this reason, i'm sure it's costing them a fucking fortune but the benefits are twofold: 1.) possibly hasten the demise of overleveraged American firms 2.) scoop up a ton of extremely valuable American data

in hindsight, very clever of openai and anthropic to trick the biggest companies in tech into bleedint themselves out via free infrastructure deals par-ai-sitic

I’m not sure it’s an apt comparison, the per user cost of telecom and AI have different considerations. Tokens aren’t transparent though, and user behavior can really shape costs for an organization at scale.

Pigs get fat but hogs get slaughtered.

This is wild to me. What a hose job.

But this isn't going to be a natural monopoly, there are going to be choices, including on-device localized open source and international AI software

Rapacious capitalism can never backfire.

It gets. All those values in dollars are even more unrealistic if you don't earn in dollars. Like, you are easily talking about pounds of beef in tokens.

Ya - they demo'd this; reminds me of Rational Rose, frankly.

This whole thing is going to collapse because users can’t afford the fees necessary to find both operational and capital costs needed for AI.

Even funnier? Claude, ChatGPT, Gemini... They'll all help you do it. So you really can do the whole "Buy a 3d printer, print a 3d printer, return your original 3d printer" Only its "Rent an agent. Have the agent setup an agent. Stop renting the agent"

It's the natural conclusion of this multi trillion dollar version of the prisoner's dilemma. If hyperscalers and neoclouds had waited and built out AI infrastructure slowly and thoughtfully, maybe costs could be contained. But the idea of being left in the dust forced a rapid and expensive buildout.

poppity pop pop pop

Cloud computing is another one where companies often found themselves facing unexpectedly huge usage bills.

It can take years to just figure out a container for a product. It has to not be destroyed in transit, have a logo and info printed on it, be visibly distinct on a shelf, potentially protect whatever is inside from extreme heat or cold. For AI apparently we are just meant to shoehorn it in and pay

Well, given the ridiculous amounts of money their investors are into them for, I’m sure there’s tons of pressure to do just that

I'll make some popcorn and watch it all go down.

yeah though for your average pleb, who has like zero media literacy and has had what the software is and can do broadly misrepresented to them by terrible people, it gets more complicated

All the PMs and VPs read from the same playbook.

The banks are looking at projections and holding out their hands for some of that 800 billion back. None of these are within 72 months of b/e let alone ROI. So they are desperate to show positive earnings. Turning the screws while the S is still on the bottom half will cause defaults.

They are pricing this stuff like clueless, non-creators. Like if you sold me a kit-Ferrari, I prolly wouldn't know it from the real one (I'm not that into cars). Likewise, if one's not into coding, the quality you deliver using an LLM will royally suck Same w/ writing, art, science, everything 2/2