custom silicon

Alibaba's AI model ran autonomously for 35 hours to optimize code for its own custom chip. This demonstrates AI systems achieving extended autonomous operation in hardware optimization, potentially reducing development cycles and costs for chip manufacturers.

Google plans to deploy nearly two million new AI chips and is collaborating with Marvell for custom chip designs to support its AI infrastructure expansion. This massive chip procurement reflects the enormous computational requirements for training and deploying large language models at scale.

Major tech companies like Microsoft, Google, Meta, and Oracle are designing their own silicon chips, potentially creating demand for foundry services. This trend could transform Intel's foundry business model as hyperscalers reduce dependence on traditional chip suppliers.