Today in AI — 29 May 2026
Today's top AI news — curated links and commentary on the stories that matter for product builders.
A $65 billion raise with memory chip manufacturers as strategic partners. Futures contracts for GPU compute. Custom silicon on the agenda at two AI labs. Today's AI news reads like a commodities desk: the supply chain is financialising, and the companies that control it may end up with more leverage than the ones training the models.
Capital follows the silicon
Anthropic's $65 billion raise brought Samsung to the table, and its involvement may extend to foundry manufacturing, giving Anthropic a path to custom silicon. Mistral is exploring its own chip design while targeting €1 billion in revenue. The pattern from the oil industry applies: when a resource becomes valuable enough, producers and consumers start integrating vertically.
- Anthropic raises $65 billion at $965 billion valuation, eclipsing OpenAI for the first time — TechCrunch
- Samsung may extend Anthropic partnership from memory chips to foundry manufacturing — SamMobile
- Mistral explores designing own chips as it targets €1 billion in revenue — CNBC
Compute gets a trading floor
Shanghai Futures Exchange, CME Group, and Intercontinental Exchange are separately designing financial instruments tied to AI compute. GPU rental prices surged 48% between mid-February and mid-April 2026. When a resource becomes volatile enough to need hedging, it has become a commodity.
Claude Opus 4.8
Anthropic released Opus 4.8 with agentic improvements and a 3x cheaper fast mode. Simon Willison's review flagged mid-conversation system messages as the most notable new capability, useful for agentic loops where context needs updating without breaking cache.
- Claude Opus 4.8 launches with agentic improvements and 3x cheaper fast mode — Anthropic
- Simon Willison reviews Claude Opus 4.8 as 'a modest but tangible improvement' — Simon Willison
Enterprise AI: plumbing first, results second
Merck and Mastercard are both reporting real agentic AI results, but with the same caveat: infrastructure came first. Oracle's shift to monthly security patches, citing AI-accelerated vulnerability discovery, tells a parallel story from the defensive side.
- Merck and Mastercard report real agentic AI results — both say infrastructure came first — VentureBeat
- Oracle shifts to monthly security patches, citing AI-accelerated vulnerability discovery — Oracle
The consumer layer fractures
Meta is testing AI subscription tiers while DuckDuckGo installs surged 30% as users rejected Google's AI Search overhaul. YouTube is building automatic detection for AI-generated content. Apple's leaked iOS 27 screenshots show a revamped Siri with new on-device AI features. Some platforms are pushing AI into everything; others are building their brand on keeping it out.
- Meta launches Instagram, Facebook, and WhatsApp Plus subscriptions and begins testing AI tiers — TechCrunch
- DuckDuckGo installs surge 30% as users reject Google's AI Search overhaul — TechCrunch
- YouTube rolls out automatic AI content detection alongside improved creator disclosure labels — YouTube Blog
- Bloomberg leaks iOS 27 screenshots showing revamped Siri and new on-device AI features — Bloomberg
For builders: model capability is no longer the binding constraint. Chip supply, compute cost, and data infrastructure are.