Microsoft just banned its own engineers from using AI. The tool was literally costing MORE than the humans it was supposed to replace. They lied to you about AI adoption and now the whole narrative is blowing up: Microsoft gave thousands of engineers access to Claude Code six months... ago and encouraged them to use it. Engineers loved it and adoption exploded. But then the invoices arrived. Token-based pricing means every query, every code review, every debugging session costs money. At scale across 100,000 engineers, the numbers became so large that Microsoft issued an internal order to cancel nearly all Claude Code licenses by end of June and force everyone onto their own cheaper tool instead. The company that invested $5 billion in Anthropic just told its own people to stop using Anthropic's product because it costs too much. Uber's story is even worse... Their CTO Praveen Neppalli Naga told The Information that the budget he planned for the full year was "blown away already" by April. Uber had rolled out Claude Code in December 2025. By March, 84% of their 5,000 engineers were using it with 70% of all committed code coming from AI systems. Heavy users were burning $500 to $2,000 per month each. Naga himself spent $1,200 in a single two-hour demo session. The company had even built internal leaderboards ranking engineers by how much AI they used. They literally gamified the spending and then ran out of money. Now look at what Nvidia's own VP of applied deep learning Bryan Catanzaro said to Axios last month. Direct quote: "For my team, the cost of compute is far beyond the costs of the employees." This is a VP at the company that SELLS the chips saying that using AI is more expensive than paying humans. Think about what this means for the entire AI narrative. Every CEO on every earnings call for the past two years has said the same thing: AI will make us more efficient, reduce headcount, and cut costs. The stock market rewarded every company that said it. Fired workers, stock goes up. Announced AI adoption, stock goes up. But the actual companies deploying AI at scale are discovering the math doesn't work. The MORE employees use AI, the HIGHER the bill. Goldman Sachs forecasts a 24x increase in token consumption by 2030 as companies adopt AI agents. Gartner just published a report showing that even though individual token prices will drop 90% by 2030, total enterprise AI costs will go UP because agents consume exponentially more tokens per task than basic tools. Meta built an internal dashboard called "Claudeonomics" to track which employees use the most AI. Amazon started pushing engineers to "tokenmaxx," their internal term for consuming as many AI tokens as possible. Both companies are spending hundreds of billions on AI infrastructure this year alone. And Microsoft, the company that bet its entire future on AI, just told 100,000 engineers to stop using the tool they liked best because the per-token bills got out of control. The companies building AI are telling investors it saves money. The companies using AI are finding out it costs more than the humans it was supposed to replace. And even the company that makes the chips just admitted it through its own VP. This is the gap nobody on Wall Street is pricing in. $725 billion in AI infrastructure spending this year across Big Tech. And the first companies to actually deploy these tools at scale are already pulling back because the economics don't work. What do you think?show more

I have worked with Claude . . . It is good for some things. Others . . . not so good. It changes code in a bunch of areas across a bunch of routines . . . and gets out of control really quickly . . . which of course eats up the token budget very fast. And because it can make errors . . . you are now debugging something that you did not write and may not understand why it made the changes it did. You really have to know when and how to use it . . . When you prompt it, you must be very, very specific with lots of constraining qualifiers . . . otherwise it assumes a bunch of things on its own and runs amuck changing things everywhere and it gets out of control. So, basically you are exchanging one skill for another . . . instead of learning to code . . . you now need to learn how to write effective prompts.

Token maxxing used to be a thing in the good ole days..

Yep , Mo Bitar covered this topis 4 days ago. I think it's really funny they bought into the PR - AI was ALWAYS just a product, no diff imo than a calculator , good for certain functions. The term AI is hype, it's not real intelligence.

No. Stop telling some people what they want to hear for clicks. Microsoft did not "ban employees from using AI" It is a much narrower and more strategic decision. They decided to reduce Claude Code licenses and push teams toward their own tooling, especially GitHub Copilot CLI....obviously. Use Microsoft’s in-house AI stack instead of paying for large-scale external Claude subscriptions. Your post is for idiots. People who cannot think for themselves see this as a collapse of AI adoption overall, and implying Microsoft concluded AI was economically useless. Token-priced coding agents at massive enterprise scale can become extremely expensive. Yes. So say that. Nonsense accounts speaking simple language to simple folk makes this platform annoying.

@SF_Conservative Indian CEOs shocked to find out their H-1Bs they hired for cheap have to rely on AI esclusively to write their code. What happens when nobody knows how to write code except the much smaller pool of AI developers and there is little training data to learn from?

I stand by my Prediction that Most of the planned Data Centers will NEVER be built. For one thing the US does not have the electricity and water for it. And secondly, Oil prices are never going to drop causing a huge reduction in uncertain to begin with profitability.

Let the bubble BURST & all these soulless greedy corporations that did mass layoffs will be forced to hire humans again or perish!

It banned its employees from using a very specific agentic harness tool (i.e., Claude), not from using AI per se. Microsoft literally has Copilot integrated in everything you have installed on your machine. I worked there recently. You cannot not use AI.

They basically gave Claude Code subscriptions to lackluster Indian coders. LOL what did they think was going to happen? Most Foreign Indians scammed their way into America. They are NOT capable coders. So they outsourced ALL of their thinking/work to Claude. #EndH1B

Decades ago, in the 1990s, when I was a service engineer, we had a customer called DEC (Digital Equipment Corporation) in Bangalore who had installed our PABX. After installing, they asked my team to enable national/international calls on every phone on every desk. I said it was a dumb idea, as everyone will misuse it. But they scoffed and said it was a standard in their US offices. So we enabled. Within two months, they called us to disable the STD/ISD on every phone and give it only to senior managers - 😇

@Nobunko1 That’s a man

"More than humans" doesn't survive the math. 100k MSFT engineers cost $30 to 40 billion fully loaded. Claude Code heavy-use runs ~$30/dev/day, ~$660 million annually. ~2% of labor. The cancellation is product substitution, MSFT already owns Copilot.

It's a temporary problem. Compounded by the lack of proper governing and budgeting overlay tools I remember in early days of internet, a poplar term was WWW - World Wide Wait, because slow speed was a big hindrance. Got solved rapidly, now we watch live streaming videos

"Uh oh" That "baked in" Ai build out capex, of $3-$5 TRILLION over the next 5 years, might have just taken a giant dump on the floor A whole lotta downside exposure, just got real "Colossal if Confirmed"

"AI may cost more than human's." But investors knew that, so I don't get it.

@TwinsBus Lowe’s requires employees to use ChatGPT for all emails. I guess because putting together complete thoughts in sentences is now obsolete.

Source for this? Haven't found corroborating evidence. Only that Microsoft told engineers drop Claude, use their own GitHub Copilot CLI. TCO will fall dramatically once AI scaled, moved to space also newer technologies lowering heat/water needs well and increased compute power.

Mismanagement of models is likely a huge culprit. If everyone in the org thinks their planning/chatting phase requires Opus 4.7, then yes it's going to be astronomical in terms of cost. Anthropic charges 5-6x what OAI does in the open market. If employees had no direction or model specific budgets, it's like giving them all black amex's with a mandate of "go shop." The story's irony isn't lost, but mismanagement of rollout and total lack of concern for budgeting will produce the same result in any company.

This isn't even wrong. MSFT switching from Claude code to GitHub Copilot (both based Opus 4.7 paid for by enterprise API usage) isn't a cut. The payment to Anthropic doesn't change. The decision was made to shift customers to a Microsoft harness (the control layer around an AI model).

🧐Grok takes a look into AI costs

The issue isn’t using AI, it’s how you use it. If you incentizinze volume, not quality interventions then yes you will get low productivity gains and high bills This is a story about misaligned incentives and human behavior, not AI

Why are companies not using local models? They have big servers available, so they can easily deploy large local models and use them unlimitedly. The gap between local and closed models is narrowing for coding and agentic tasks. Companies can easily do 80% coding tasks with local model and 20% with claude for reviewing and improving code.

And then there is this :

How interesting. The Logistics company I work for have also rolled out use this AI and github etc. Over 6 months ago. No useful results yet and I am sure this is costing tons of money! I am in that program and it is retarded! I eventually get frustrated and just ask Grok in general terms, no company details, how to extract the info I need to make reports. After 10 minutes with Grok on my cell phone next to my company computer, the report query etc is done! Hahahaha! JUST USE @grok

They can't get spell check to work, for God's sake.

Thing is- if the productivity went up 10x, but the cost only increased 2x- this is simply a budgetting outcome- not an indictment of AI.

"Gartner just published a report showing that even though individual token prices will drop 90% by 2030..." Does anyone actually believe that?

Claude be like: Claude: I fixed all files, they are bug free. Me: Log in button doesn't work. Claude: You're right! Let me check files....Checking... You've hit your limit...

give it a couple of months and OpenCode (open source version of Claude Code) will run on a $10k server and be as good as current Claude Code is.

While I agree, this take doesn’t need to be anti-ai to work. The angle people might be missing is that token economics and revolutionary technology of AI can both be true. It is a budgeting argument, and budgeting eventually comes for every technology. OSS models can help combat these huge bills but whether a valuation correction comes with the budgeting is yet to be seen.

Cost will eventually drop significantly IMO. Like any new tech. Many of the AI buildouts will lose lots of money and fail as the low cost tool will win. Likely too much AI data center supply being built. Tech improvements will reduce the need for some of the AI data centers.

And you best believe if they can’t get more data centers? They’re going to all lose their minds and start doing shit illegally obtained access to waterways and such. Do not trust the data centers. And no matter what the fuck they say they’ll offer you!?, once it’s built they’ll say fu

@MsLaToshaBrown For now… AI innovation will catch up and become cheaper.