惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

月光博客
月光博客
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
阮一峰的网络日志
阮一峰的网络日志
罗磊的独立博客
T
Tailwind CSS Blog
博客园_首页
博客园 - 司徒正美
Google DeepMind News
Google DeepMind News
Hugging Face - Blog
Hugging Face - Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
V
V2EX
J
Java Code Geeks
量子位
D
DataBreaches.Net
MongoDB | Blog
MongoDB | Blog
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Microsoft Azure Blog
Microsoft Azure Blog
P
Proofpoint News Feed
C
Check Point Blog
V
Visual Studio Blog
H
Help Net Security
Recent Announcements
Recent Announcements
Engineering at Meta
Engineering at Meta

PYMNTS.com

BNY Names New Head for Payments/Trade Client Platform Treasury Calls for Programmable Financial Enforcement Across Crypto DeepSeek Seeks $20 Billion Valuation as Tech Giants Weigh Investment Google Accelerates Agentic AI Shift With New Enterprise Platform OpenAI Begins Briefing Governments on Cybersecurity Capabilities DeFi Security Suffers New Blow With $3 Million Volo Exploit Uninvited Users Access Anthropic’s Mythos AI Model Block and Uber Expand Partnership Across Several Global Markets OpenAI Pledges $1.5 Billion to PE Enterprise AI Project Podcast: Inside the $9 Billion DeFi Hack That’s Shaking Crypto’s Foundations Synchrony CFO Flags Momentum in Spending and Credit Banks Risk Slowing the Emerging Middle Market Firms Driving Growth Paysafe Expands Digital Wallet Availability Across 18 European Markets Bad Data Can Break Good AI in Payments 50% More Digital Shopping Days Put Parents at the Center of Retail’s Shift 65% Call Insurance Essential. Why Most Spending Isn’t So Clear-Cut Amazon Recasts Marketplace Fraud as a Broader Trust Problem Capital One’s Q1 Shifts Attention From Spending to Strategy Lawmakers Question JetBlue About Surveillance Pricing Allegations Small Businesses Stop Chasing Amazon on Delivery Speed Google Embeds AI Into Chrome for 3.5 Billion Users Adobe Plans Outcome-Based Pricing for New AI Product Suite UnitedHealth Spends $1.5 Billion on AI and Wants Double Back MiCA Forces Crypto Firms to Get Licensed or Get Out Prediction Market Kalshi Targets Crypto Perpetuals New York Sues Coinbase and Gemini Over Prediction Markets Amazon and Anthropic Deepen Ties With Investment and Hardware Pact Agentic B2B Is Here. Are Your Contracts and Invoices Ready? Apple Hardware Leader John Ternus to Succeed CEO Tim Cook The Web Is Gaslighting AI Agents and Nobody Can Tell
Rising AI Adoption Is Driving Up Enterprise Costs
PYMNTS · 2026-04-17 · via PYMNTS.com

AI adoption Claude Code

A few months into 2026, Uber’s chief technology officer found himself back at the drawing board.

“The budget I thought I would need is blown away already,” Praveen Neppalli Naga said in an interview with The Information. The cause was Claude Code, Anthropic’s AI coding assistant, whose use inside Uber’s engineering teams had surged past every projection the company had made.

Neppalli Naga’s comments illustrate the challenge for companies trying to take advantage of new AI tools while keeping costs under control.

AI coding tools don’t behave like traditional software. The cost isn’t fixed. It rises with use. Each interaction consumes compute, measured in tokens, and the cost of those tokens add up quickly. At Uber’s scale, usage didn’t grow steadily. It surged. And costs followed faster than anyone expected.

Usage Scaled Faster Than Anyone Planned

Uber uses Claude Code to write and improve code across its back-end systems, the infrastructure that matches drivers and riders, sets prices, and launches new product features. The company also used it to build its recently launched Earner Assistant, an AI agent that helps drivers find high-paying trips.

Neppalli Naga told The Information that roughly 11% of live updates to Uber’s back-end systems are now written by AI agents built primarily with Claude Code, up from a fraction of a percent three months ago.

Advertisement: Scroll to Continue

Uber created internal leaderboards ranking engineers by AI usage, actively pushing adoption. Claude Code usage soared from the end of last year while usage of a competing tool, Cursor, plateaued.

Uber’s research and development expenses rose 9% to $3.4 billion in 2025. The company said in a recent securities filing it expects that figure to keep climbing. Neppalli Naga wouldn’t disclose exact AI spending figures.

Anthropic Moves to Charge by Consumption

The pricing structure enterprises had been working with has since changed. Anthropic moved enterprise customers to usage-based billing in recent weeks, as PYMNTS reported. Customers of Claude Enterprise now pay for computing capacity consumed plus a flat monthly per-user fee. Previously, they paid up to $200 per licensed user monthly for a set amount of discounted usage.

Fredrik Filipsson, co-founder of Redress Compliance, a firm that helps businesses negotiate software licensing agreements, told The Information the changes could double or triple costs for heavy users. Several IT executives told the publication they’re monitoring whether the new pricing will produce substantially higher bills. Anthropic said the changes don’t apply to businesses with fewer than 150 users.

The pricing shift extends beyond enterprise accounts. Anthropic also began charging Claude Code subscribers more to use the tool with third-party software, as PYMNTS reported. As of April 4, subscribers can no longer use their Claude subscription limits for third-party tools including OpenClaw, an open-source personal AI assistant. They must now pay through a separate pay-as-you-go option. Anthropic said the policy applies to all third-party tools connected to Claude Code.

Boris Cherny, Anthropic’s head of Claude Code, said the company’s subscriptions “weren’t built for the usage patterns of these third-party tools” and that it was attempting to manage growth sustainably.

What Enterprises Are Now Recalculating

AI tools that embedded themselves inside engineering workflows as productivity aids are now line items with variable, consumption-driven costs. Enterprises that adopted them at scale without modeling actual usage are finding invoices that don’t match budgets.

Organizational readiness remains the most cited barrier to AI adoption at large companies. Some 71% of executives at companies with at least $1 billion in annual revenue identify it as the primary limit on AI performance, according to PYMNTS data. Only 11% said the technology itself is the barrier.

Neppalli Naga said his vision is for AI agents to eventually take on full software engineering roles, writing, testing and deploying code without human involvement. Uber said it continues to calibrate hiring based on broader business needs.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.