Google is set to roll out specialized inference processors designed to run AI models post-training, marking its most aggressive push yet into semiconductors. These chips aim to deliver high throughput and lower power consumption compared with general-purpose GPUs, directly targeting Nvidia's core revenue stream in data centers and cloud services.

Cerebras Systems has submitted paperwork for a public listing under ticker CBRS, highlighting $510 million in 2025 revenue, up 76% year-over-year. The company’s wafer-scale engine architecture and blue-chip customers such as OpenAI, Amazon and Meta position it as a credible competitor in large-scale AI training and inference markets.

Anthropic secured a fresh $5 billion investment from Amazon, bringing total funding to $13 billion, in exchange for over $100 billion of AWS usage commitments over the next decade. The agreement covers capacity on Amazon’s Trainium2 through Trainium4 AI accelerator chips, with options for future chip generations as they become available.