According to BlockBeats, on August 3, Citrini analyst Jukan published an article stating that the continued development of open-source AI models and the weakening of the monopoly of closed-source models may become a significant turning point for the cloud computing industry, and will help cloud vendors enhance the commercial value of AI infrastructure.

Jukan points out that the core concern faced by cloud vendors in the past was that after investing heavily in GPU procurement and data center construction, they might end up only providing infrastructure for a few closed-source model companies, while these model companies would control user access and pricing power.

However, as the performance of open-source models improves and costs decrease, the AI application model is shifting from "a single high-performance model solving all tasks" to "hierarchical model routing." Complex inference tasks are still handled by top-level closed-source models, while a large number of ordinary tasks can be completed by low-cost small models or open-source models.

He believes that with increased model substitutability, cloud vendors will gain more user access, traffic allocation rights, and pricing power. In the future, closed-source models may no longer be "paid channels" on top of cloud computing, but rather a computing resource that can be freely allocated by the cloud platform.

Furthermore, competition among open-source models does not necessarily mean a decrease in hardware demand. Jukan stated that lower inference costs could drive rapid growth in AI calls, while model compression, inference optimization, intelligent routing, and self-developed ASIC chips will reduce the general-purpose GPU resources required per token.

He believes that the long-term growth of AI infrastructure depends on whether the rate of demand growth exceeds the rate of efficiency improvement. If token usage grows faster than algorithm and chip efficiency improvements, data center utilization and return on capital can remain high, driving cloud vendors to continue investing in computing infrastructure.

Jukan concludes that the true logic behind the AI infrastructure bull market is not simply based on "cheap models benefiting the cloud, and cloud growth benefiting hardware," but rather on open-source models reducing the monopoly profits at the model layer, and cloud vendors improving the efficiency of monetizing computing resources through computing power scheduling and vertical integration, ultimately forming a positive cycle between cloud computing and hardware investment.