Snap and Perplexity amicably end $400M AI search integration deal

What Happened

Snap and Perplexity have officially terminated their $400M partnership, originally announced last November. The deal was slated to integrate Perplexity's AI-driven conversational search engine directly into the Snapchat application, likely to augment the "My AI" chatbot or power a dedicated in-app search surface. Both companies have stated that the termination was amicable.

Technical Details

Integrating a third-party AI search engine into a platform with Snap's daily active user base (over 400 million) presents massive scalability and architectural challenges. Perplexity relies on a complex Retrieval-Augmented Generation (RAG) pipeline, requiring real-time web scraping, index querying, and LLM inference. Routing Snap's highly concurrent query volume through an external API introduces significant latency risks, rate-limiting hurdles, and strict SLA requirements.

Furthermore, managing data privacy—specifically ensuring that ephemeral user queries aren't inadvertently absorbed into external training data or persistent logs—requires rigorous, custom data-handling contracts at the API layer. Maintaining state and context across a third-party boundary for millions of simultaneous sessions is a non-trivial engineering burden.

Why It Matters

From an architectural standpoint, this signals a shift in how mega-platforms approach AI features. Relying on an external vendor for a core conversational user experience creates a single point of failure and limits customizability. Snap has historically invested heavily in its own ML infrastructure. By abandoning this deal, Snap is likely pivoting to an internal routing layer for "My AI." This approach allows them to orchestrate multiple foundational models (such as OpenAI models, which they already leverage) alongside their own vector databases, maintaining strict control over latency, inference costs, and user data.

What to Watch Next

Watch for Snap's upcoming AI infrastructure updates, particularly regarding their internal RAG capabilities and model orchestration strategies. For Perplexity, monitor how they adjust their B2B API strategy; losing a flagship consumer integration suggests they may refocus on enterprise knowledge-base integrations or hardware partnerships where latency boundaries and scale requirements are more predictable.