Custom AI Chips to Capture Over 33% Compute Market by 2030, Google’s TPU Rising
ARK Investment forecasts custom AI chips to capture over one-third of the compute market by 2030, highlighting Google's TPU and Amazon as rising competitive forces. Amazon’s expanded $100 billion compute deal with OpenAI includes 2 GW of Trainium capacity, underscoring intensifying chip competition.
1. ARK Predicts Custom Silicon Adoption
ARK’s research team projects that custom AI chips—distinct from GPUs—will account for more than one-third of global compute capacity by 2030, signaling a shift away from traditional server architectures and intensifying competition among chipmakers.
2. Amazon’s Trainium Agreement
Amazon’s multi-year partnership with OpenAI has expanded by an additional $100 billion in compute commitments, with OpenAI slated to consume approximately 2 gigawatts of Trainium chip capacity, underscoring Amazon’s growing influence in custom silicon.
3. Google’s TPU Competitive Position
Google’s Tensor Processing Units have emerged as a leading alternative to GPU-based acceleration, with major AI developers such as Meta reportedly leasing TPUs for advanced model training, reinforcing Google’s position in the evolving AI hardware market.