By S&P Capital IQ
Publication Date: 2026-01-30 03:05:00
Perplexity, which primarily used Amazon Web Services (AWS) previously, will now access models from OpenAI, Anthropic, and xAI through Microsoft’s Foundry service. The company said it has not yet shifted spending from AWS, which remains its main cloud provider, and plans to expand its existing AWS partnership. Perplexity’s CEO, Aravind Srinivas, has previously emphasized a strong reliance on AWS, which has highlighted the company as a notable AI customer.
The deal aligns with Microsoft’s strategy to position Azure as a leading platform for AI application development, offering models from multiple vendors. Implications, context, and why it matters. The $750 million commitment’s fine print needs to be checked.
The near-term value of Perplexity’s three-year, $750 million commitment will be determined by the exact model access terms through Microsoft’s Foundry service. The status of partner models such as Anthropic should be confirmed, since general availability versus a limited ?preview? affects whether they can be used for live workloads.
The regions eligible for pay-per-token billing should be checked, since model providers can limit where usage is allowed. From these terms, it will be clear whether capacity is available now or mostly deferred to later. Multi-cloud AI can create a costly data-egress problem that private interconnects aim to address.
Perplexity uses Amazon Web Services (AWS) plus Microsoft Azure, which can add network complexity and raise spend.
