Application Refactoring · IT Operations Automation · Real-time Analytics
After unifying its core data estate, PayPal needed an AI platform that could power global commerce use cases, let developers choose the best model for each workload, and elastically handle training and inference without reverting to monolithic release cycles or custom infrastructure for every project.
Reported impact
PayPal transformed its post-migration data estate into a flexible AI platform that speeds releases, supports tens of thousands of pipelines a day, and lets the business scale agentic commerce capabilities without locking teams into a single model vendor or architecture pattern.
PayPal built its next-generation AI stack on Google Cloud by combining BigQuery, Google Kubernetes Engine, Gemini Enterprise Agent Platform, and Cloud GPUs into a unified model and infrastructure layer. BigQuery provides the AI-ready data foundation following PayPal's migration of more than 300 petabytes, while GKE gives teams an elastic architecture that replaced slower monolithic release patterns and now supports 25,000 pipelines per day. On top of that runtime, PayPal created a unified LLM stack: open-source models run on the GKE inferencing layer, while proprietary models such as Gemini run through Gemini Enterprise Agent Platform and Model Garden. Cloud GPU capacity provides the compute horsepower for training and high-performance workloads. This architecture lets developers move from idea to production quickly while preserving security, compliance, and model-choice flexibility across PayPal's 439 million-user commerce business.
Enterprise data lands in BigQuery and is consumed by PayPal's data scientists and AI services. Open-source models run on the GKE inferencing layer, proprietary models such as Gemini are served through Gemini Enterprise Agent Platform, and Cloud GPUs supply the training and inference capacity needed to scale AI products into commerce workflows.
PayPal built a unified large language model (LLM) stack, which runs open-source models on top of the inferencing stack on GKE and close-sourced models like Gemini on top of Gemini Enterprise Agent Platform.
With Google Cloud, PayPal built a unified AI platform to architect the next era of commerce for over 439 million global users.
after GKE, our release life cycle went from weeks to minutes.
developers have access to more than 200 enterprise-ready models through Model Garden.
migrating over 300 petabytes of data to BigQuery
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Google Cloud Sample Archived · Aug 29, 2026 WeSpeak Building that product required infrastructure that would never go down. Early on, WeSpeak ran on a virtual private server that could fail without warning, taking the entire platform offline. For a hotel waiting on a… 70% of hotel guest inquiries automated end to end Cloud Run · Gemini · Gemini Flash Google Cloud Sample Archived · Jun 6, 2026 Foodpairing In the food and beverage industry, launching new products successfully is a challenge. More than 80% of product launches fail, due in large part to the sector’s traditional reliance on slow, expensive consumer testing. Achieves more than 75% correlation with real consumer panels BigQuery · Cloud Run · Cloud SQL Google Cloud Sample Archived · May 28, 2026 IKS Health Their technology goals centered on a cloud-first strategy with Google Cloud, aiming to modernize from legacy on-premise, monolithic systems. Key objectives included migrating to a unified, scalable cloud-native platform… Achieved 30% infrastructure cost savings by migrating to Google Cloud BigQuery · Cloud Build · Cloud Logging 