NVIDIA and SK hynix signed a multiyear agreement to co-develop advanced memory for Vera Rubin AI supercomputers, RTX Spark PCs, and Jetson Thor robotics platforms — securing supply at the design stage across three distinct NVIDIA markets, not just at procurement. The deal landed during Jensen Huang's Seoul visit that swept in four more Korean partnerships — Naver's 55 MW sovereign AI data center, SK Telecom's gigawatt-scale cloud by 2027, LG humanoids, and Hyundai autonomous manufacturing — marking Korea as a central node in NVIDIA's global AI infrastructure buildout.

OpenAI is merging ChatGPT, Codex, and partner integrations — Canva, Booking — into a superapp in the biggest overhaul since the 2022 launch, with CPO Thibault Sottiaux describing the target as a personal agent handling tasks across work and life. Codex's sixfold growth to 5 million weekly users since February, and business tools now at 40% of revenue targeted at 50% by year-end, mark this as product consolidation already underway — not a roadmap promise.

Perplexity's Search as Code (SaC) replaces fixed search API calls with agent-generated Python scripts that run in a sandbox — on an internal task tracking 200 critical CVEs, token usage fell 85% and the approach outscored Anthropic and OpenAI systems on four of five self-reported benchmarks. Agents on Perplexity's Agent API or Perplexity Computer can now write custom retrieve-filter-deduplicate-rerank pipelines, running parallel queries and scripting their own filters instead of accepting whatever a fixed API returns.

Devin Desktop wraps Windsurf with a multi-agent dashboard — ACP, an open protocol, lets Codex, Claude Agent, OpenCode, and custom-built agents share context in one workspace via project-grouped 'Spaces' that persist across pull requests. Engineering teams running multiple AI coding tools can consolidate them under one interface today; Cognition has not published pricing, deployment terms, or independent benchmarks for the claimed 30% efficiency gains in the rebuilt Devin Local.

SpaceX signed a $920M/month compute deal with Google for ~110,000 Nvidia chips through June 2029 — capacity Musk originally built for xAI, which lagged, now leased to a second hyperscaler alongside SpaceX's existing $1.25B/month Anthropic contract. The two-tenant structure converts a stranded infrastructure bet into locked-in hyperscaler revenue, and both deals surface in SpaceX's IPO S-1 ahead of its June 12 Nasdaq debut at a $1.75T target valuation.

Uber opened a UK interest list for Wayve autonomous rides — Ford Mustang Mach-Es with human safety operators, at standard fares, launching in the coming months pending regulatory approval — as Waymo tests roughly 100 Jaguar I-PACEs across 100 square miles of London, making this the first city where Uber's two AV relationships compete directly. The $300M tranche of Wayve's $1.5B raise, contingent on London deployment, reframes Uber as an equity stakeholder whose payout depends on Wayve taking ground Waymo is already testing.