Urgent.News

What's breaking now, across thousands of outlets.

AI

Uber Spent Its Entire 2026 AI Budget in 4 Months. It Just Found a Way to Fix It.

Uber Spent Its Entire 2026 AI Budget in 4 Months. It Just Found a Way to Fix It.

So far, in 2026, Uber has maintained flat AI spending while usage of its AI agents has surged. Weekly demands to its AI agents increased more than nine times since February, yet the overall bill did not follow suit. Each thousand requests now costs 34% less than its April peak, with a single session running 52% cheaper than in June.

This cost reduction was achieved by delegating simple tasks to cheaper AI models and reserving the pricier ones for more complex jobs. Uber also implemented a cap on each session's AI usage and provided engineers with real-time cost tracking. This strategy allowed Uber to spend its entire 2026 AI budget in just four months. However, this move was necessary as the company's net margin dropped from 27% to 17.3% last year, and the stock fell after the last earnings report due to investor concerns over margins.

Moreover, Uber is set to invest over $10 billion in self-driving cars, making cost-saving measures crucial. Currently, AI agents make up more than 70% of Uber's code-change submissions, and engineers run over 30,000 agent tasks daily. This approach enables Uber to maintain growth without hiring as many people, as engineers now produce double the code each.

This AI cost discipline may also protect Uber's profits as it continues investing in autonomous vehicles.

Written by urgent.news from Yahoo Finance's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at finance.yahoo.com →

More in AI

The guardrails that matter are code, not prompts

I built an agent that coordinates eldercare for my mother in India, and the interesting part was not the agents. It was the four times I stopped and decided what the system was not allowed to do, and…

  • Framework enforces safeguards to prevent agents from acting beyond limits
  • Deterministic layer and severity dictionary block prompt manipulation
  • Compliance with legal boundaries and tamper-proof record-keeping

175 Items Marked Done, Zero Delivered: An AI Agent Queue Incident

Originally published at aideazz.xyz — cross-posted here with canonical link. A newly added data source ran on schedule for a full day, logged a healthy count every hour, and delivered nothing at all…

  • 175 items marked as processed before actual handling
  • Processing cap applied after deduplication, causing flaw
  • Commit 36e985c fixed issue, implemented August 30, 2026

More from Sunday 30 August →