Ruh-Roh. Chuckling aside, this is a really good thing. This is, btw, EXACTLY how enterprises figured out they weren't ready for "The Internet" in the dot com boom. They were spending millions on servers and kit and the ROI was crap.
> through an internal leaderboard ranking teams by total AI tool usage
So you were encouraging people (in fact, practically threatening people) to waste as many tokens as possible. No wonder you blew through the budget so quickly.
Doesn't this just mean that their incentives worked too well? If I budget $1M, for the year, tell everyone to use AI as much as they can, and they spend that million in four months, so what? Next time I just tell them to spend it slower.
My ongoing theory is for small to middle sized businesses AI is incredibly useful as it will help leverage you to grow your business by building more than you could before AI. But for big businesses, like uber, not sure the advantages are the same, they could already build what they want, so it seems the only thing that's likely is cost savings.
As far as I can tell this entire story was derived from a 2 minute segment in this YouTube video: https://www.youtube.com/watch?v=y_mQ6xLcKyc&t=1616s - I linked directly to the relevant section at 26:56 in the video.
If you watch the video snippet it seems a whole lot less of a big deal than all of the headlines that have come out of it.
Let’s be blunt: any CEO or executive that fell for this AI wealth extraction needs to be fired ASAP. These are not the type of people who you want running a company.
My own company hired a new CEO in Q1 and in his first company address he declared that anyone not using AI for their daily duties was putting themselves out of a job and would be unemployable by anyone else. I laughed at that nonsense. 5 months later and we’ve seen zero growth or improvements due to AI. This chucklehead CEO will be the unemployable one.
Feeling nostalgic. There is a new tech. In the beginning everyone will be asked to go crazy on spending as much as you can & just focus on the outcome. After a while, they will see just cost is going up (less on outcomes). Now, start throttling the dollar limits per employee. By this time whole industry will settle on this new tech & reality. Everyone will learn it is not all that fancy as it sounded earlier.
But we need to keep prompting at the Anthropic casino! There is a token slot machine called "Jevons paradox" that promises as the cost of your tokens get much cheaper, you be more productive in gambling even more tokens! /s
Except what Uber is experiencing is that the tokens are not getting cheaper and the employees are gambling even more tokens and using up the budget; racking up expensive bills for Uber. So they are blowing through their budget.
It is at least cheaper to gamble your tokens with DeepSeek's or Xiaomi's casinos and they have just announced a permanent 99% off deal with an allocation of 100 trillion tokens! [0]
That is a great deal, with a small catch. (If you are fine with sending your data and prompts to them)
The AI economic bubble is not because AI is useless. There was a paper recently that highlighted how AI is diagnosing pancreatic cancer patients years before humans do. This is not a one off (and not limited to science). Those who are getting these types of results across many domains understand the potential value (like the people who get a two year head start on their treatments).
The problem is that the entire market right now is set up for resolving things with hardware instead of software. Bigger models, faster chips, data center sprawl etc. Clearly that is not sustainable. But if you use AI inference for reasoning only, and traditional software like databases, web searches, etc. for the things they do well (at a fraction of the cost) the economics flip.
Trying to force expensive training updates and ever larger models that know who the first baseman was for the winning 1939 world series time is useless and where the cost-to-value ratio is broken. Pull what you need from a search api, and send the results to an LLM to analyze. No fine tuning, no super big chips and even small models can produce good results.
16 comments
[ 2.6 ms ] story [ 33.8 ms ] threadWhen you pay by the dead cobra, don't be surprised when people start breeding cobras.
So you were encouraging people (in fact, practically threatening people) to waste as many tokens as possible. No wonder you blew through the budget so quickly.
yep, sounds about right
- https://news.ycombinator.com/item?id=47976415
- https://news.ycombinator.com/item?id=48268871
If you watch the video snippet it seems a whole lot less of a big deal than all of the headlines that have come out of it.
Let’s be blunt: any CEO or executive that fell for this AI wealth extraction needs to be fired ASAP. These are not the type of people who you want running a company.
My own company hired a new CEO in Q1 and in his first company address he declared that anyone not using AI for their daily duties was putting themselves out of a job and would be unemployable by anyone else. I laughed at that nonsense. 5 months later and we’ve seen zero growth or improvements due to AI. This chucklehead CEO will be the unemployable one.
Uber president says AI spending is getting 'harder to justify' - https://news.ycombinator.com/item?id=48277485 - May 2026 (132 comments)
Uber’s COO says it’s getting harder to justify money spent on tokenmaxxing - https://news.ycombinator.com/item?id=48268871 - May 2026 (334 comments)
Uber torches 2026 AI budget on Claude Code in four months - https://news.ycombinator.com/item?id=47976415 - May 2026 (475 comments)
Except what Uber is experiencing is that the tokens are not getting cheaper and the employees are gambling even more tokens and using up the budget; racking up expensive bills for Uber. So they are blowing through their budget.
It is at least cheaper to gamble your tokens with DeepSeek's or Xiaomi's casinos and they have just announced a permanent 99% off deal with an allocation of 100 trillion tokens! [0]
That is a great deal, with a small catch. (If you are fine with sending your data and prompts to them)
[0] https://platform.xiaomimimo.com/docs/en-US/news/v2.5-price-u...
The problem is that the entire market right now is set up for resolving things with hardware instead of software. Bigger models, faster chips, data center sprawl etc. Clearly that is not sustainable. But if you use AI inference for reasoning only, and traditional software like databases, web searches, etc. for the things they do well (at a fraction of the cost) the economics flip.
Trying to force expensive training updates and ever larger models that know who the first baseman was for the winning 1939 world series time is useless and where the cost-to-value ratio is broken. Pull what you need from a search api, and send the results to an LLM to analyze. No fine tuning, no super big chips and even small models can produce good results.