In the first week of June 2026, Bloomberg reported that Uber had capped each employee's AI tool spending at $1,500 a month. When the story hit Hacker News, it drew 349 comments. One commenter argued that "if that's the ceiling at Uber's scale, our team is under-investing." Another multiplied $1,500 by headcount and arrived at "$25.5 million a month, company-wide." The number itself grabbed attention, but the thread's real staying power came from something else: a major corporation had, for the first time, attached a public dollar figure to AI tool spending. That's the part of this story worth actually reading.

The Moment Spending Controls Kick In

Let's sit with Uber's number for a moment. $1,500 a month works out to $18,000 a year — roughly ₩24 million in Korean won — per employee. Underneath that ceiling sit tools like GitHub Copilot, Claude Pro, ChatGPT Plus, Cursor, and Perplexity, most priced between $20 and $200 a month. You could subscribe to seven or eight of them at once and still not hit the cap. Which suggests $1,500 isn't really a constraint that bites today — it's a boundary drawn in advance, on the assumption that costs are headed up.

A commenter who identified themselves as a former Google engineer put it this way: "When a big company sets an AI budget cap, it means whoever owns the budget has started looking at the logs. AI tools used to get filed on a different line than software licenses — now they're starting to land on the same one." The more AI tools become part of daily work, the more their cost gets absorbed into routine overhead. Uber's decision is one of the first times that shift has played out in public.

Not everyone agreed the number was right, though. Some developers pushed back, arguing that "in an era where AI tools write hundreds of lines of code a day, a $1,500 cap is too stingy relative to the productivity it buys." With software engineer salaries running $150,000 to $250,000 a year, $18,000 in annual AI tool spend is just 7 to 12 percent of labor costs. Even a 40 percent cut in coding time would more than pay for it. Some companies have in fact published internal data showing release cycles shrinking 30 to 40 percent after adopting coding assistants. That's why the thread was full of comments like "it should be $3,000, not $1,500" and "even $5,000 wouldn't be too much." The debate ended up splitting three ways — too little, too much, and just right — with no consensus on where the ceiling itself should sit.

Where the Money Goes When Nobody's Tracking It

Uber's cap put two questions on the table at once: how much should a company spend on AI tools, and are the tools it's already paying for actually doing anything?

The first question looks simple, but most organizations still don't have a clear answer. AI subscriptions start with a few clicks — sometimes an employee just puts it on a personal card, sometimes it gets folded into an IT budget line. Without a company-wide tracking system like Uber's, even knowing the actual total is hard. One informal survey of Korean startups found that only 18 percent of teams said they knew exactly how much they were spending on AI tool subscriptions. The other 82 percent were either guessing or not checking at all.

The second question is the more uncomfortable one. In the same Hacker News thread, a startup CTO confessed: "We rolled out seven AI tools for the team, and six months later we were actually using two of them every day. The other five were just switched on because we figured we'd get around to them." A lot of readers will recognize themselves in that admission. It's a reminder of the gap between using a lot of tools and using them well.

When you open a line-item and can immediately explain why that cost exists, you end up making different financial calls than someone who can't. If the reason you're paying $10 a month for Notion AI is "it's just on," you're missing the fact that the same $10 spent elsewhere might be saving someone 30 minutes of work. The discipline of reading spending and results as numbers — sorting out which expenses actually produce outcomes and which just keep running on inertia — applies just as directly to AI tool spending. What matters isn't how much you spent, but how much you spent and what you got for it.

A Solo Founder's AI Bill, Korean Style

Let's be honest: $1,500 a month isn't a realistic ceiling for most solo operators in Korea. You won't find many solo PMs or one-person agencies spending ₩1.5 million a month on AI tools. But there's still something to learn from the number.

The fact that Uber landed on $1,500 means someone there had already wrestled with what "the right amount" looks like. Most of us, by contrast, haven't even started that conversation. Add up what a typical Korean solo operator is already paying: ChatGPT Plus ($20), Claude Pro ($20), Perplexity ($20), Cursor ($20), Notion AI ($10), Gamma ($15), Midjourney ($10) — that alone comes to $115 a month, or $1,380 a year. Throw in two or three specialized tools for a particular workflow, and it's not hard to clear $2,000 a year. Suddenly Uber's $1,500 ceiling doesn't feel so far removed from your own numbers.

The trouble is that these charges sit quietly, scattered across separate card statements. One month you add a new tool thinking "let's give this a try." Another month it just renews without you noticing. Another month you spend the whole time debating whether to cancel it. Somewhere in all that, $2,000 a year can slip by quietly. Uber was probably in a similar spot before it set the cap. The difference is that Uber decided to actually measure it.

Setting a boundary starts with a list. Laying out every AI tool you currently use across four columns — name, monthly cost, last used, main purpose — takes about 15 minutes. Then fill in, honestly, how often you actually used each one in the last 30 days. Anything under five uses in that window is a candidate for reconsideration. Add one more question — "what would I have done without this tool?" — and you start to see where that money could be better spent.

A big company has budget owners, and it has a CFO. A solo operator doesn't have anyone else to fill that role. Uber's cap is an internal control mechanism. The equivalent for a solo operator is a self-imposed "monthly AI budget limit." It looks like a constraint, but what it really does is force a decision about where to concentrate your investment.

I don't think this is simply a matter of cutting costs. AI tool spending has already climbed from $0 to tens of dollars a month, and some specialized tools now run past $200 a month. Products like Claude Workspace or GitHub Copilot Enterprise, sold only at the organization level, are starting to appear too. Those price tags are still steep for a solo operator today, but that could change in two or three years. AI tool spending is moving toward sitting on the expense sheet right alongside office rent and software licenses.

To avoid being caught off guard when that moment arrives, it's worth taking a look at your own spending pattern now. The question Uber's $1,500 ceiling raised holds regardless of company size: which tools are worth keeping, and which ones should get cut this month? That call plays out differently depending on whether you've already set a standard for yourself or not.