Unverified50% confidenceFactExact time
谷歌于2016年正式发布第一代张量处理单元（TPU），核心是脉动阵列架构，目前TPUv4已支撑Gemini等大模型训练
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
DeepSeek Builds Its Own AI Chips: The Strategic Logic and Challenges Behind Compute Independence
hackernewshackernews7/9/2026
Related Claims
UnverifiedTPUv4 has already supported the training of large models such as Gemini.75% similarUnverifiedGoogle自研的TPU芯片为Gemini模型的训练和推理提供定制化算力支持，在成本和效率上相比依赖英伟达GPU的竞争对手具备一定优势73% similarUnverifiedGoogle Colab provides cloud-based GPU/TPU computing power and integrates Gemini code suggestions.71% similarUnverifiedGoogle maintains globally distributed A100/TPU v5 inference clusters for Gemini.68% similarVerifiedOpenAI's GPT-4o mini and GPT-4o, and Google's Gemini Flash and Gemini Pro all adopt a similar tiered model product strategy.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/489800API
curl https://kongchang.com/api/v1/knowledge/claims/489800MCP
get_claim(id=489800)