Five AI stories from the week of April 20–April 26, 2026 that actually mattered — what changed, why it matters, and where to read more.
In this issue
- Google commits up to $40B to Anthropic in cash and compute
- OpenAI releases GPT-5.5 with stronger agentic coding and research
- Google signs multi-billion deal to supply compute to Thinking Machines Lab
- Sergey Brin leads DeepMind strike team to close AI coding gap with Claude
- US states pass wave of AI legislation in a single week
TechCrunch
Google commits up to $40B to Anthropic in cash and compute
Google will invest $10 billion in Anthropic immediately at a $350 billion valuation, with up to $30 billion more contingent on performance targets, plus 5 gigawatts of Google Cloud compute capacity over five years. The arrangement follows Amazon’s separate $5 billion commitment earlier in April and gives Anthropic more than 11 gigawatts of total contracted compute across major cloud providers. For practitioners, the deal cements a model where frontier labs rely on competing clouds simultaneously rather than single-vendor lock-in, reshaping how infrastructure contracts are structured across the industry.
OpenAI
OpenAI releases GPT-5.5 with stronger agentic coding and research
OpenAI released GPT-5.5 on April 23, targeting enterprise agentic workflows in coding, computer navigation, scientific research, and knowledge work; the model is available to Plus, Pro, Business, and Enterprise ChatGPT users and in the API as of April 24. OpenAI describes it as a faster reasoner per token than GPT-5.4, with benchmark improvements over Gemini 3.1 Pro and Claude Opus 4.5 on coding and technical research tasks. API availability makes it directly adoptable for production systems without a waitlist.
TechCrunch
Google signs multi-billion deal to supply compute to Thinking Machines Lab
Google Cloud signed a single-digit-billion agreement to provide Mira Murati’s Thinking Machines Lab with Nvidia GB300 GPU infrastructure, training support, and deployment services — the startup’s first formal cloud partnership, following an earlier investment arrangement with Nvidia. The non-exclusive deal specifically covers reinforcement learning workloads used by Tinker, Thinking Machines’ custom model-building tool. The arrangement illustrates the competitive dynamic in which cloud providers bundle compute access with strategic relationships to attract frontier labs before they commit to a single vendor.
Sherwood News (original reporting: The Information)
Sergey Brin leads DeepMind strike team to close AI coding gap with Claude
The Information reported on April 21 that Google co-founder Sergey Brin has assembled a focused research group within Google DeepMind — led by pretraining researcher Sebastian Borgeaud — with an explicit mandate to improve Gemini’s coding capabilities to match Anthropic’s Claude. Brin is coding daily himself and has required all Gemini project engineers to use internal agentic tools for multi-step tasks; the team will train on Google’s internal codebase of more than 2 billion lines of proprietary code. The move reflects an internal assessment at Google that AI coding ability has become the decisive competitive axis in the enterprise market.
Read at Sherwood News (original reporting: The Information) →
Transparency Coalition
US states pass wave of AI legislation in a single week
Tennessee’s legislature passed the CHAT Act (Curbing Harmful AI Technology) on April 21 by a 90-0 House vote after a 31-0 Senate vote, adding to Nebraska’s chatbot safety law signed April 17 and four AI bills Maryland sent to its governor following its session close. Arizona faced an April 25 adjournment deadline with three additional AI bills in active reconciliation, California intensified committee hearings on dozens of proposals covering chatbot safety and employment transparency, and Colorado advanced chatbot safety legislation through a House third reading on April 21. The simultaneous activity across multiple states signals a shift from federal inaction to an emerging patchwork of state-level compliance requirements that practitioners deploying consumer-facing AI products will need to track.