Amazon Plans AI Chips to Cut Costs 30% and Challenge Nvidia’s 80% Market Share
NVDA•Amazon Web Services is in talks to develop and sell custom AI inference chips aimed at generative AI workloads, targeting volume shipments by late 2027. The move could reduce AWS’s per-inference costs by up to 30% and challenge Nvidia’s over 80% share of the data center AI accelerator market.
1. AWS Pursues In-House AI Chips
Amazon Web Services has entered discussions with semiconductor partners and its internal design teams to create AI inference accelerators optimized for large language models and vision workloads, with a target to begin volume production by late 2027.
2. Cost and Market Share Implications
AWS aims to deliver per-inference prices up to 30% lower than current GPU offerings, directly targeting Nvidia’s position, where the company holds more than 80% of the data center AI accelerator segment.
3. Competitive Landscape and Industry Response
Analysts expect this initiative to pressure Nvidia’s growth and spur similar moves from Microsoft and Google, intensifying competition across the roughly $75 billion AI hardware market and potentially reshaping cloud procurement strategies.
