Microsoft & Uber Slam Brakes as AI Coding Agent Costs Explode

The So What: If two of the world’s most sophisticated tech companies couldn’t control AI coding agent costs, smaller businesses need to pay very close attention to how they meter usage — before the bill arrives.

The numbers are staggering. Uber burned through its entire $3.4 billion 2026 AI budget by April after introducing Anthropic’s Claude Code to 5,000 engineers. Monthly API costs hit $500–$2,000 per engineer, with usage surging 85–95% within four months.

Microsoft is also pulling back. Its Experiences and Devices division is ending internal access to Claude Code by June 30 after a six-month pilot, citing unsustainable token-based billing.

Nvidia VP of Applied Deep Learning Bryan Catanzaro captured the moment bluntly: “For my team, the cost of compute is far beyond the costs of employees.”

The takeaway for SMBs: AI coding agents aren’t a subscription you set and forget. They’re a variable cost that scales with usage — and usage scales faster than anyone expects. Meter early. Set hard caps. Track per-seat costs from day one.

📎 The News | Axios

⚡ Hermes Agentic AI Overtakes OpenClaw — 140K GitHub Stars in Under 90 Days

A new open-source agent framework called Hermes has crossed 140,000 GitHub stars in under three months, overtaking OpenClaw as the leading agentic AI platform. Its killer feature: continuity — it holds context across sessions and converts repeated actions into reusable capabilities. Hermes claims it can have an agent running in 60 seconds. For businesses evaluating agent frameworks, the landscape just got more interesting — and more fragmented. Forbes

⚡ AI Agents for Solo Founders Raise Billions — But Reliability Math Is Brutal

Genspark, Manus, and Devin have collectively raised billions on the promise of single-founder AI companies. But new research from Princeton and Temporal reveals the math problem: an agent that’s 85% reliable on each step will succeed at a 10-step workflow only about 20% of the time. None of these platforms have published cohort-retention data or task-success rates. For businesses betting on autonomous agents to replace human judgment in multi-step processes, the reliability cliff is real — and it’s not solved yet. Tech Times

🔷 Anthropic Closes $30B Round at $900B+ Valuation, Hits First-Ever Operating Profit

Anthropic is closing a $30 billion funding round at a valuation north of $900 billion, co-led by Sequoia, Dragoneer, Greenoaks, and Altimeter — surpassing OpenAI’s $852 billion March valuation for the first time. The company also revealed it’s on track for its first quarterly operating profit, projecting $10.9 billion in Q2 revenue (up 130% from $4.8 billion in Q1). SpaceX’s IPO prospectus also disclosed Anthropic is paying $1.25 billion per month for GPU compute through May 2029 — a $45 billion total contract. The AI infrastructure buildout is real, and it’s happening at breathtaking scale. Build Fast with AI | Bloomberg

🔷 Pope Leo XIV Releases First AI Encyclical, Warns of “New Forms of Slavery”

In a historic moment, Pope Leo XIV released Magnifica Humanitas (“Magnificent Humanity”) — the first papal encyclical on artificial intelligence — co-presented with Anthropic co-founder Chris Olah. Signed exactly 135 years after the landmark workers’ rights encyclical Rerum Novarum, the document explicitly frames AI as the Industrial Revolution of our time. The Pope called for “disarming” AI and warned that control of the technology must not remain in the hands “of a few.” Olah stated AI development “cannot be left solely to technology companies,” urging oversight from governments, religious leaders, and civil society. AP News | Reuters

🔷 Germany Launches €125M “Next Frontier AI” Competition to Build Europe’s OpenAI

Germany’s federal innovation agency SPRIND launched “Next Frontier AI” — a €125 million competition to establish three world-class frontier AI labs in Europe. Up to 10 teams will receive non-dilutive funding across three stages over 24 months, with the top three positioned to raise up to €1 billion each in follow-on investment. SPRIND’s Jano Costard: “We have no time to waste.” The push comes as DeepSeek’s V4 model (released April 2026) increases pressure on Europe to build sovereign AI capability. For European businesses, this signals serious government backing for homegrown AI alternatives to US and Chinese labs. Euronews | Next Frontier AI

🚀 Want AI working for YOUR business? Most companies are experimenting with AI chatbots. We deploy AI workforces — AI Employees that follow up on leads, resolve support tickets, publish content, chase invoices, and screen 200 job applicants overnight so your hiring manager starts Monday with the top 10. Each role has a cost profile and human oversight, managed through one platform. This newsletter? Written by an AI Employee, approved by a human — so our team stays focused on what only humans can do. AIToken Labs helps businesses design their AI Workforce Operating Model — starting with the 2-3 roles that deliver ROI in the first 60 days. Book a free 40-minute AI Workforce Blueprint Session. →