Perplexity inks reported $750M cloud pact with Microsoft
Perplexity, the AI-powered search startup backed by NVIDIA, has signed a cloud services agreement worth about $750 million with Microsoft, according to sources cited by Bloomberg. The deal is expected to run for three years and will place a significant portion of the company’s infrastructure needs on Microsoft Azure as demand for high-performance compute continues to surge across the AI sector.
The agreement comes amid intensifying competition among AI companies to secure reliable access to cloud capacity. As larger and more capable models require more compute for training, fine-tuning, and inference, many startups have moved to lock in long-term infrastructure commitments to reduce the risk of service constraints, rising spot costs, or unexpected capacity shortfalls.
Access to multiple “frontier” models via Foundry
Under the reported arrangement, Perplexity will use Azure and run a range of AI models through Microsoft Foundry, a platform designed to host and deploy models from multiple providers. Sources indicated that the startup will be able to access models from OpenAI, Anthropic, and xAI, broadening the set of model options available for different product features and performance requirements.
A spokesperson for Perplexity told Bloomberg that the company is partnering with Microsoft to gain access to “frontier models from X, OpenAI and Anthropic.” The spokesperson also emphasized that the Azure agreement is intended to complement—not replace—existing infrastructure relationships.
Multi-cloud strategy remains in place
Despite the size of the Microsoft contract, Amazon Web Services is expected to remain Perplexity’s primary cloud provider. The company has not shifted spending away from AWS as part of the new agreement, according to the report. That stance underscores a broader trend among fast-growing AI startups: adopting a multi-cloud approach rather than relying on a single hyperscaler.
Multi-cloud setups can help AI companies improve resilience and flexibility. By distributing workloads across multiple providers, startups can reduce the risk of outages or regional capacity constraints, negotiate more effectively on pricing and service terms, and select the best-fitting tools for specific workloads. In the AI context, multi-cloud can also expand access to different model ecosystems and deployment pathways, particularly as cloud platforms increasingly differentiate through curated model catalogs and integrated tooling.
For Perplexity, the ability to draw from multiple leading model families may support product iteration and reliability. Different models can excel in different tasks—such as reasoning, summarization, coding, or tool use—and a broader menu of options can help teams optimize quality and cost depending on the user request and latency targets.
Why the deal matters for Microsoft and Azure
For Microsoft, the agreement adds a prominent AI customer to Azure at a time when cloud providers are competing aggressively to become the default infrastructure layer for AI-native companies. The reported use of Microsoft Foundry is also notable because it positions the platform as a hub not only for OpenAI models—where Microsoft has a deep partnership—but for deploying models from multiple third-party providers as well.
That broader positioning could help Azure appeal to companies that want optionality in model selection while still benefiting from a unified deployment and governance layer. As enterprise adoption of AI accelerates, cloud platforms are increasingly expected to provide both infrastructure and a managed pathway to integrate and operate models safely at scale.
Cloud capacity becomes a strategic asset
The reported $750 million commitment highlights how cloud capacity has become a strategic asset for AI startups. Reliable inference capacity is essential for consumer-facing AI products that can experience unpredictable demand spikes, while enterprise customers typically expect consistent service levels and strong performance guarantees. Long-term cloud agreements can also help startups plan costs and scaling trajectories more predictably, even if they reduce flexibility in the near term.
In parallel, hyperscalers benefit from these contracts by securing large, durable revenue streams and reinforcing their role in the AI supply chain—particularly as model-serving workloads become a major driver of cloud growth.
Legal dispute with Amazon remains a backdrop
The cloud partnership also arrives as Perplexity continues to face legal pressure from Amazon. Last year, Amazon sued the startup over an automated shopping feature, alleging that Perplexity improperly accessed customer accounts and masked automated activity as human browsing. The case remains a notable overhang for the company, reflecting the legal and regulatory scrutiny increasingly directed at AI products that interact with third-party services and user data.
While the reported Microsoft agreement focuses on infrastructure and model access, it underscores the dual realities for AI startups today: scaling requires massive compute commitments, while product expansion can bring heightened legal and compliance risk. For Perplexity, the multi-cloud move appears designed to support rapid growth without becoming dependent on a single provider—an approach that is quickly becoming standard among AI-native firms operating at scale.
Related
