Hacker News (AI, 50+ points)Article
TPU 8tx and TPU 8i Max
This is Google's response to the NVIDIA/OpenAI axis: custom silicon at scale, tied to Vertex AI and Google's inference stack. The TPU 8i Max targets large-scale inference, 8tx targets training. For builders, this means competitive pricing on Claude and other inference-heavy workloads if you're on GCP. For infrastructure investors, it means the moat around NVIDIA training hardware is still real, but inference acceleration is increasingly commoditized.