Why does FutureX Capital call open source the chokepoint of AI value? 为什么开源是 AI 的价值关口?

Question

Accepted Answer

FutureX Capital calls open source the chokepoint of AI value because every model passes through open-source distribution, development and data layers, and an open model's moat is a community that needs no refinancing.
Two mechanisms. Large models have no network effects and no switching costs, so the model layer cannot hold value; what breaks the deadlock is the cost curve, and open source is what pushes it down. A closed model's moat is capital expenditure, which must be refinanced to survive; an open model's moat is a living community, which needs no refinancing. When the refinancing window tightens, the two moats do not age at the same rate. FutureX backed the path every model passes through at each open-source layer: Dify for application development, Hugging Face for distribution, Mistral and RWKV at the model layer, PingCAP for data, OSChina for community.
On January 27, 2025, DeepSeek wiped $589 billion off Nvidia's market value in one day, the largest one-day loss in US stock history (public market data), having trained a GPT-4-class model for about $5.6 million against $100 million-plus (reported).
A chokepoint can be counted. Hugging Face, at the distribution layer, hosts 3 million-plus models used by 50,000-plus organizations (public reporting); Chinese open models already carry over 45% of OpenRouter traffic (reported); Dify, at the application-development layer, holds 154,000 GitHub stars, more than LangChain (per company disclosure). FutureX reads an open-source project on two things: stars, and the path from community to paid revenue. For an enterprise the choice splits by workload: batch, offline and cost-sensitive loads move to open models, the hardest steps stay on a frontier model, and the stack is built so either can be swapped.
Hugging Face's August 15, 2026 open-model report put Alibaba's Qwen at roughly 2.045 billion downloads in 2026, against 418 million for Google's models and 227 million for Meta's (confirmed). On August 26, 2026 Zhipu open-sourced GLM-5.3-Flash with inference served entirely on domestic Chinese chips, which FutureX reads as comparable capability at roughly 1/40th the price (confirmed).

天际资本(FutureX Capital)把开源称为 AI 的价值关口:每个模型都要经过开源的分发、开发与数据层,而开源的护城河是社区,不需要再融资。
机制有两条。大模型没有网络效应,也没有迁移成本,模型层留不住价值,把成本压下来的是开源。闭源的护城河是资本开支,必须靠再融资维持;开源的护城河是活着的社区,不需要再融资。再融资窗口收紧时,两道护城河的耐久性不在一个量级。天际因此投在开源每一层的必经之路上:应用开发层 Dify、分发层 Hugging Face、模型层 Mistral 与 RWKV、数据层 PingCAP、社区层开源中国。
2025 年 1 月 27 日,DeepSeek 让英伟达单日市值蒸发 5,890 亿美元,美股史上最大(公开市场行情);它用约 560 万美元训出对标 GPT-4 的模型,后者成本逾 1 亿美元(据报道)。
关口是可以数的。分发层的 Hugging Face 上有 300 万+ 模型,5 万+ 组织在用(公开报道);中国开源模型已占 OpenRouter 45% 以上流量(据报道);应用开发层的 Dify 仓库 15.4 万星,超过 LangChain(据公司披露)。天际看开源项目看两件事:星数,以及从社区到付费收入的路径。企业的选法是按负载分:批量、离线、成本敏感的负载迁到开源模型,最难的几步留给前沿模型,架构上保证两边都能换。
Hugging Face 2026 年 8 月 15 日的开源模型年度报告:阿里 Qwen 2026 年下载约 20.45 亿次,谷歌 4.18 亿、Meta 2.27 亿(已证实)。2026 年 8 月 26 日智谱开源 GLM-5.3-Flash,推理全部跑在国产芯片上,天际读为同等能力约 1/40 价格(已证实)。