Analysts flag AI consolidation wave after recent big‑cap deals
- Alphabet agreed on April 24 to invest up to $40 billion in Anthropic, days after Amazon expanded its own Anthropic commitment to $25 billion. - Anthropic said it will spend more than $100 billion on Amazon Web Services over 10 years while keeping Claude available on AWS, Google Cloud, and Azure. - The back-to-back deals shifted focus from model launches to cloud control, chip supply, and distribution leverage. (reuters.com)
Alphabet’s April 24 agreement to invest up to $40 billion in Anthropic landed four days after Amazon expanded its Anthropic commitment to as much as $25 billion. (reuters.com 1) (reuters.com 2) Anthropic said Google will put in $10 billion now, with another $30 billion tied to performance milestones, while Amazon said it will invest $5 billion immediately and up to $20 billion more later. (reuters.com) (anthropic.com) The Amazon side of the deal came with a bigger operating commitment: Anthropic said it will spend more than $100 billion on Amazon Web Services technologies over the next decade. (anthropic.com) (aboutamazon.com) That agreement secures up to 5 gigawatts of Amazon capacity for Claude, spans Trainium2 through Trainium4 chips, and adds inference expansion in Asia and Europe. (anthropic.com) (aboutamazon.com) Anthropic also said Claude remains available on the three biggest cloud platforms: Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry. The point is not exclusivity; it is that cloud companies now want equity stakes, chip commitments, and sales-channel access at the same time. (anthropic.com) That is why analysts are talking about consolidation even without a formal merger. Control is moving toward the companies that own the data centers, custom chips, enterprise contracts, and marketplaces where customers buy model access. (cnbc.com) (reuters.com) OpenAI’s infrastructure ties show the same pattern from the other side of the market. OpenAI said in July 2025 that Oracle would help develop 4.5 gigawatts of additional Stargate data-center capacity in the United States. (openai.com) By April 28, investors were also treating those contracts as balance-sheet risk after The Wall Street Journal reported OpenAI had missed internal revenue projections, sending Oracle and chip shares lower. (cnbc.com) The recent deals leave fewer truly independent paths to frontier-scale artificial intelligence. The labs still write the models, but the hyperscalers increasingly own the pipes, the power, and the checkout counter. (anthropic.com) (openai.com)
