Productivity and Search Growth
Q: Doug Anmuth with JPMorgan asked about the golden age of productivity for Google with new models and the disconnect between record query growth and search growth deceleration.
A: Sundar Pichai noted that internal deployment of Argon has significantly improved software engineering productivity, with more gains expected from Carbon. Philipp Schindler explained that search growth was impacted by FX headwinds, specific vertical dynamics like retail and media spend, and the lower monetization of AI mode queries, though they are now scaling ads in AI mode globally.
Agentic AI and Search
Q: Brian Nowak with Morgan Stanley asked about technological hurdles to shipping agentic products to the full user base and the search performance comparison factors for coming quarters.
A: Sundar Pichai explained that they are working on distilled versions of Carbon models for efficient deployment and building a secure runtime via GCP, with consumer rollouts starting later this year. Anat Ashkenazi clarified that search growth faces continued FX headwinds and the lapping of 2025 growth acceleration, which will pressure results in Q4 and Q1.
Compute Allocation and Cloud Demand
Q: Eric Sheridan with Goldman Sachs asked about priorities for compute allocation and how the view on cloud customer needs is evolving regarding external TPU systems.
A: Sundar Pichai stated they use a disciplined, long-range planning process for compute allocation across their consumer and cloud businesses. Thomas Kurian added that they are broadening distribution of TPU systems to customer data centers and other providers like Oracle to meet strong demand from AI labs and enterprises.
TPU Systems Financials
Q: Ross Sandler with Barclays asked what percentage of a TPU system sale is recognized as revenue, the margin profile, and if handing off infrastructure to AI labs reduces Alphabet's CapEx intensity.
A: Anat Ashkenazi explained that TPU systems include hardware, compute, storage, and networking, providing significant added value. She noted that inventory is built ahead of revenue recognition based on signed contracts, and while this impacts cash flow, it is not traditional CapEx, with more revenue expected through 2027.
Agentic Commerce and TPU Business
Q: Justin Post with Bank of America asked how the company will manage the search transition to agentic use cases, if it will drive more monetization, and why the TPU business is a good use of capacity.
A: Management stated that agentic experiences are additive and allow for new advertising models, such as offering choices in shopping results. Regarding TPUs, they emphasized that designing and manufacturing these systems allows them to scale the product, improve economics, and address markets where data cannot move to the cloud.
Agentic Use Cases and Cloud Growth
Q: Mark Shmulik with Alliance Bernstein asked about surprises regarding early agentic use cases and the growth rate of Google Cloud excluding TPU system sales.
A: Sundar Pichai noted that users are actively using the task tab in the Gemini app, and they are learning from these interactions to improve outcomes. Thomas Kurian stated that GCP accounted for the vast majority of revenue, and growth excluding TPU systems was very strong, driven by Gemini usage, compute, storage, and cybersecurity.
Platform Strategy and Safety
Q: Michael Nathanson with MoffettNathanson asked about the philosophy of working with competitors on models and cloud, and what has been learned about safety and trust in agentic consumer behavior.
A: Sundar Pichai explained that they distinguish between products and platforms, and they aim to make their platforms, like TPU systems, world-class by competing for demanding workloads. Regarding safety, he noted that users are concerned about trust, and Alphabet's history of safe, scaled deployment is a key differentiator.

