Unverified80% confidenceFactExact time
OpenAI、Google DeepMind、Anthropic等头部公司同时争夺有限的H100/H200 GPU供应
1
Sources
80%
Confidence
Long-term
Relevance
5/14/2026
First Seen
Sources
Claude变笨真相：Anthropic算力短缺引发开发者信任危机
rss5/14/2026
Related Entities
Related Claims
UnverifiedOpenAI和Anthropic使用NVIDIA H100/A100等高端GPU集群，硬件折旧和电力成本极高76% similarUnverified微软为OpenAI建造了一台配备约1万张英伟达V100 GPU的超级计算机69% similarPartially VerifiedNVIDIA的MIG技术可在A100/H100等高端GPU上实现硬件级别的资源隔离67% similarUnverified英伟达H100/H200 GPU均依赖CoWoS将计算芯片与HBM3内存集成，CoWoS产能瓶颈一度制约了H100的整体供货量67% similarUnverifiedllama.cpp 支持 CPU+GPU 混合推理，通过 --n-gpu-layers 参数控制卸载到 GPU 的层数，并内置兼容 OpenAI API 格式的 HTTP 服务端66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/10938API
curl https://kongchang.com/api/v1/knowledge/claims/10938MCP
get_claim(id=10938)