Unverified50% confidenceOpinionExact time
The spending of large-model companies mainly flows to three areas: computing power procurement (especially GPU clusters), the electricity and bandwidth costs of model training and inference, and the high salaries of top research talent.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/20/2026
First Seen
Valid until: 1/18/2027
Sources
Related Entities
Related Claims
VerifiedAI订阅服务的成本主要来自大规模GPU集群的推理算力开销74% similarUnverified大模型API的成本结构主要由GPU算力折旧（60-70%）、电力和冷却成本（15-20%）、带宽和运维开销（10-15%）构成72% similarUnverifiedGPU compute costs are the underlying driver of inference service pricing.70% similarUnverifiedThe official APIs of top large models are all pay-as-you-go, with high inference costs.70% similarUnverifiedThe cost of training frontier models is a massive sunk cost, and recouping that investment relies primarily on the pricing of inference services.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1077793API
curl https://kongchang.com/api/v1/knowledge/claims/1077793MCP
get_claim(id=1077793)