Uber CTO Praveen Neppalli Naga says the company's surging use of AI coding tools has maxed out its full-year AI budget just a few months into 2026
Uber's surging use of AI coding tools, particularly Anthropic's Claude Code, has maxed out its full year AI budget just a few months into 2026 …
The InformationLaura Bratton
Context & Ripple Effects
Uber’s use of AI coding tools has become large enough to exhaust a full-year AI allocation within months, with Claude Code identified as a major tool in that usage. Related coverage later shows Uber moving from broad access toward per-employee token limits, indicating that adoption quickly became a cost-control problem as well as a developer-productivity initiative.
The story also sits alongside coverage of Claude Code’s emergence as a leading AI coding product. That makes Uber a visible enterprise example of how rapidly agentic coding consumption can scale once a tool is widely adopted.
First-order effects
Uber must constrain, reallocate, or increase spending on AI coding access after usage outran its annual budget; later reporting indicates a $1,500 monthly token limit per tool for employees.
Anthropic benefits from heavy Claude Code consumption at a major customer, while Uber’s engineering users face usage governance rather than effectively open-ended access.
Second-order effects
Uber’s procurement and engineering leaders have incentive to measure which coding workloads justify token spend, favoring teams and tools that can demonstrate useful output per dollar.
Other large software organizations adopting coding agents are likely to treat token caps, allowances, and centralized purchasing as standard controls rather than leaving spend entirely to individual teams.
Third-order effects
If enterprise agent use continues to scale this way, AI coding adoption will be governed less by seat count and more by inference consumption, making FinOps-style controls part of software-development operations.
The leading coding-model vendors may increasingly compete on controllability and cost transparency alongside model capability, because high utilization can turn successful deployment into a budget-management issue.
The trend: Agentic coding tools are moving from experimental developer perks to metered enterprise infrastructure that requires active inference budgeting.
This is why the cost of intelligence can't be going to zero. Demand is too high and supply too low. As a result, AI costs will approach the cost of human labor. Unless we solve the energy problem I don't see a way around that.
@altcap
Brad Gerstner on x
Most CEOs set 2025 people budgets before they saw the banger results of the new models / coding / co-work. Companies now making real time adjustments to lower tech headcount growth for 2026 & increase spend on tokens / intelligence. 🧐🧐
@ericvishria
Eric Vishria on x
Seeing this all over. The efficiency and rationalization push is coming. Initially it was just getting everyone to use AI all over. “Opus everything!” Now the spend is material and companies will rationalize.
@cto_junior
@cto_junior on x
This is the conversation happening in every executive boardroom right now You can rein in AI spend and sleep well tonight But if your competitors don't, they'll outship you by an order of magnitude
@cdpetty
Clayton Petty on x
A this point something is becoming clear: just as there is massive waste in human labor & work inside enterprises, there will be massive waste in agent / token spend.
@tunguz
Bojan Tunguz on x
We are already futuremaxxing.
@anissagardizy8
Anissa Gardizy on x
Uber's CTO told @LauraBratton5 that AI coding tools—particularly Anthropic's Claude Code—has already maxed out its 2026 AI budget 📈 “I'm back to the drawing board, because the budget I thought I would need is blown away already,” Neppalli Naga said. https://www.theinformation.com…
@clementdelangue
Clem on x
let's go open-source and local models!
@chamath
Chamath Palihapitiya on x
ummm so tokenmaxxing didn't increase operating margins?!?!? of course it didn't. its just that the cto of uber had the courage to say the quiet part out loud.
@virtualmilin
Milin Desai on x
AI pilled. But tokenmaxxing ≠ value creation. More companies are about to learn this and work through this.
@edzitron.com
Ed Zitron on bluesky
Neppalli refused to say how much Uber was actually spending on tokens, but this is the beginning of organizations admitting that AI is a massive resource drain. With Anthropic moving all enterprise to API pricing we'll find out in the next few months if AI is really “worth it” […
@selectric.space
Liz Ten Eleven on bluesky
even the most cash-strapped school department gets further along in the year before the point where they have to start running photocopies on the backs of already-printed papers lmao [embedded post]