BlockBeats reported on Aug. 3 that Citrini analyst Jukan said the continued development of open-source AI models, together with a weakening monopoly held by closed-source model providers, could become a major inflection point for the cloud computing industry. He said the shift may help cloud vendors improve the commercial value of AI infrastructure.
Jukan said the main concern cloud providers faced in the past was that after spending heavily on GPU procurement and data center buildouts, they could end up serving only a small number of closed-model companies, while those model firms kept control of user access and pricing power.
A shift away from the one-model-for-everything approach
According to Jukan, as open-source models improve in performance and come down in cost, AI deployment is moving from a model where one high-performance system handles every task to a layered model-routing structure. In that structure, complex reasoning work would still be handled by top closed-source models, while a large volume of ordinary tasks could be completed by lower-cost small models or open-source models.
He said that as models become more interchangeable, cloud providers stand to gain more control over user access, traffic orchestration, and pricing. In his view, closed-source models may no longer function as a toll gate sitting on top of cloud computing. Instead, they could become another computing resource that cloud platforms can dispatch more freely.
Open-source competition does not necessarily cut hardware demand
Jukan also said growing competition among open-source models does not mean hardware demand will fall. Lower inference costs could lead to a rapid increase in AI calls. At the same time, model compression, inference optimization, intelligent routing, and self-developed ASIC chips may reduce the amount of general-purpose GPU resources required for each token.
He argued that the long-term growth of AI infrastructure depends on whether demand expands faster than efficiency improves. If token usage grows faster than gains in algorithms and chip efficiency, data center utilization and return on capital could remain at high levels, supporting continued investment by cloud providers in computing infrastructure.
The bull case centers on orchestration and vertical integration
Jukan said the real bull thesis for AI infrastructure is not simply that cheaper models are good for the cloud and cloud growth is good for hardware. His conclusion was that open-source models could compress monopoly profits at the model layer, while cloud providers improve monetization efficiency for computing resources through compute scheduling and vertical integration, eventually creating a positive cycle between cloud computing and hardware investment.