Monday, April 27, 2026

Start of archive·May 16

Top story

The NumbersAI Business

Google Could Invest Another $40B in Anthropic

Google doubling down on Anthropic represents a major capital reallocation strategy within the AI race, positioning it against OpenAI's Microsoft backing and signaling that tech giants are willing to deploy massive sums to own frontier model development rather than compete independently.

Google potential $40B additional investment in Anthropic

The briefs

GitHub Copilot, one of the highest-volume AI inference endpoints in production, is moving from unlimited to metered pricing. This signals that even at massive scale, per-token economics of LLM inference are forcing product model changes—and hints at broader margin pressure across the AI-as-a-feature market.

GitHub Copilot moving from unlimited/all-you-can-eat to metered billing model

DeepSeek's V4 release—open-source, cost-optimized, and running on alternative (Huawei) inference hardware—signals a shift in the global AI competitive landscape away from closed Western models and NVIDIA-dependent infrastructure.

DeepSeek released two versions of V4

Meta's $2B AI acquisition faces regulatory blockade in China, signaling intensifying geopolitical restrictions on cross-border AI deals and raising questions about whether Western AI companies can still access or invest in emerging Chinese talent.

Meta $2B acquisition blocked by China

Core Scientific's pivot from crypto to AI compute represents a major shift in how data center capacity is being repurposed for large-scale model training and inference. Gigawatt-scale facilities are becoming the currency of competitive advantage in the AI economy.

Core Scientific converts Bitcoin mining operation to AI compute farm

Microsoft is shifting GitHub Copilot from flat-rate to consumption-based pricing, signaling a broader industry move toward metered AI billing as inference costs climb. This mirrors how cloud infrastructure scaled and foreshadows pricing models across the entire AI product stack.

GitHub shifts to metered AI billing model

Microsoft's non-exclusive IP license and multi-cloud strategy signals a fundamental shift in OpenAI's business model—moving from a captive partnership to competitive cloud neutrality. This restructuring carries major implications for Azure's AI moat and OpenAI's independence as Musk litigation adds external pressure.

Microsoft IP license converted from exclusive to non-exclusive

DeepSeek's V4 preview demonstrates significant capability advances (extended context window, world model improvements) in direct competition with US-based flagship models, signaling accelerating Chinese AI competitiveness and reshaping model benchmarking hierarchy.

DeepSeek V4 preview released Friday

A landmark legal ruling on OpenAI's corporate structure could reshape the entire AI industry's path to profitability and set precedent for how AI companies can be organized. With an IPO pending, the outcome affects investor confidence, regulatory frameworks, and competitive dynamics across the sector.

Trial scheduled for week of April 27, 2026 in Northern California

Microsoft's exclusive cloud partnership with OpenAI is officially ending in 2032, freeing OpenAI to distribute through competing cloud providers. This reshapes the cloud AI infrastructure landscape and signals shifting power dynamics between the two companies.

Microsoft-OpenAI exclusivity ends in 2032