Meta runs trillions of tokens weekly on Microsoft’s Azure Foundry AI marketplace, making it one of Microsoft’s top AI clients. This highlights a strategic use of external AI models alongside Meta’s internal capabilities.
- Meta runs trillions of tokens weekly on Azure Foundry AI models.
- Microsoft Foundry resells models from providers including OpenAI.
- Meta uses external AI to supplement and evaluate its proprietary models.
Market signal
Meta’s hundreds of millions in annual spending for AI model access on Microsoft’s Azure Foundry indicates increasing commercial demand for third-party AI services even among major technology operators with significant internal AI capabilities. This demand reflects the need for diverse AI models and platforms to accelerate software development and evaluation.
Microsoft’s Foundry marketplace aggregates AI models from multiple vendors, excluding Microsoft’s own models, and recently has seen rapid growth—both in customers and token processing volume. Meta now ranks near the top alongside other major buyers like ByteDance, positioning the marketplace as a significant industry platform for AI procurement.
Operator impact
Meta’s approach to combining internal frontier AI models with rented external models represents a pragmatic development strategy. Using Microsoft’s marketplace allows Meta developers to cross-validate model outputs, improving quality control. This includes using OpenAI models rented via Foundry to benchmark Meta’s own model generation.
The substantial operating cost incurred by Meta for external AI access is notable given its large capital expenditures on internal infrastructure. It signals that cloud-based AI model rental is seen as a complementary resource rather than a replacement, helping fill capacity gaps and speed innovation cycles without requiring full in-house model deployment.
What to watch next
Meta has publicly acknowledged potential plans to build its own AI cloud compute rental business, which could compete directly with Microsoft’s Foundry marketplace. Monitoring how Meta evolves from a large customer to a competitor will be critical for cloud operators and AI model providers.
The growth trajectory of Microsoft’s AI revenue, heavily driven by Foundry resales including OpenAI’s models, will influence cloud services pricing and partnerships. The expanding token processing volume and customer base also suggest broader enterprise AI adoption patterns that operators and buyers across sectors should carefully track.