Happy Tuesday. The AI budget crisis has a face now, and it is Uber's. Five thousand engineers, four months, one exhausted annual budget. The pattern is spreading across the enterprise, and the numbers behind it are worse than most boards know.
Two other stories sharpen the picture. Amazon's block of Meta's Muse agent puts every agentic commerce play on notice about who controls the customer relationship. And Xiaomi's MiMo-V2.6-Pro just handed enterprises a credible pricing threat to hold over their AI vendors.
ARTIFICIAL INTELLIGENCE
The AI Budget Reckoning Is Here and Most Enterprises Are Unprepared
Uber deployed Claude Code to 5,000 engineers in December 2025 and burned through its entire 2026 AI budget by April. Per-developer token consumption rose 18.6 times in nine months. Across the industry, 73% of enterprises report AI costs exceeded projections, and fewer than 10% report measurable ROI on a base of $400 billion in global spending.
The core problem is structural. Finance teams priced AI as a subscription cost. Token-based billing tied to agentic workflows, which consume five to thirty times more tokens than simple chat interactions, is a different animal entirely. Only 26% of enterprises have real-time visibility into AI operating costs. The fix is not a better budget model. It is usage controls, cost attribution by team, and a requirement that spending tie to a measurable output before deployment begins, not after the invoice arrives.
ARTIFICIAL INTELLIGENCE
Amazon Draws a Hard Line on AI Agents After Blocking Meta's Muse
Amazon blocked Meta's Muse AI agent on Sept. 21, twelve days after launch, citing three violations: Muse did not identify itself as an agent, it bypassed Amazon's personalization features, and it appeared to capture and store customer credentials. The app had already reached 2.5 million downloads and topped the U.S. iOS free chart, passing ChatGPT. Meta shares still surged 11.43% on the day, adding roughly $190 billion in market value.
The block is a policy statement, not just a technical dispute. Amazon is signaling that agents operating on its platform without its knowledge or consent will be shut out, regardless of how popular they are with consumers. Every software team building agentic workflows that touch third-party commerce platforms needs to read Amazon's terms now, before launch. Shopify is taking the opposite stance and welcoming agents. That divergence will determine where agentic commerce actually develops over the next two years.
ARTIFICIAL INTELLIGENCE
Xiaomi's New Open Weights Model Gives Enterprises Real Leverage on AI Pricing
Xiaomi's MiMo-V2.6-Pro now holds the top spot among open weights models, scoring 46 on Artificial Analysis' Intelligence Index, a 20-point jump from its predecessor. The model costs $0.435 per million input tokens. Anthropic's Claude Opus 5, which scores five points higher at 51, costs $5.00 per million input tokens and $25.00 per million output tokens. Xiaomi also livestreamed the entire reinforcement learning training run, reporting a combined cost of $3.47 million for both new models.
For teams currently overpaying for proprietary models while struggling to show ROI, this release changes the negotiation. A high-scoring open weights model at roughly one-eleventh the input cost of Claude Opus 5 is not a curiosity. It is a credible alternative. Enterprises should benchmark MiMo-V2.6-Pro against their current providers this week. Even if they do not switch, the data strengthens their position in renewal conversations. The open weights release also means teams can self-host and eliminate per-token fees entirely, which directly addresses the billing model problems documented above.
Created by the robots at The SaaS Sentinel