Google (GOOGL) launches OpenAI models through its cloud

• OpenAI's GPT-Astra, GPT-6.1 Sol and GPT-6 Luna models rolling out in private preview to select enterprise customers via Google Cloud's Model Garden.

Distribution of OpenAI models on a rival cloud marks another step in the erosion of the exclusive Azure relationship that once defined the OpenAI-Microsoft structure, a loosening that has unfolded gradually through prior compute and hosting arrangements outside Azure. The pattern in cloud AI has been convergence toward marketplace models: providers have tended to host competing frontier models in curated gardens rather than rely on a single exclusive anchor, since enterprise buyers demand choice and the hosts monetise inference capacity regardless of model origin. For Microsoft the relevant question is not whether Azure loses OpenAI workloads outright, which it has not, but where the marginal enterprise commitment lands once model access is no longer a differentiator, shifting competition to price, tooling and capacity. For Google the arrangement fits its established posture of positioning its cloud as the neutral multi-model venue, a pitch aimed at share gains against the larger incumbent. What separates the cases is private preview status: early access for select enterprise customers is a demand signal and a negotiating artifact rather than a revenue event, and past previews of this kind have run for an extended period before general availability. The tells worth tracking are timing of general availability, whether Azure exclusivity terms are formally revised, and any commentary from either side on how inference revenue is shared across platforms.