Oct 28, 2026 earnings call — Gemini 4 launch and Cloud TPU scaling
Google Cloud revenues grew 103% year-over-year, driven by AI infrastructure, solutions and TPU system sales.
Revenue $127.6B (est $128.0B) · EPS $2.12 (est $3.01)
Key quotes
"Our first-party model APIs are now processing nearly 32 billion tokens per minute, 40% more than last quarter and more than 4x what we processed a year ago."
"Google Cloud's backlog was $561 billion in the third quarter, with the majority of the backlog related to typical GCP contracts."
"In Q3 alone, Gemini models helped advertisers generate over 1.5 billion campaign assets, up over 7x year-over-year."
Analyst Q&A 7
AI productivity and Search ads
Q (Douglas Anmuth · JPMorgan): He asked Sundar to explain why 2027 could be a golden year for productivity with the new models. He also asked Philipp why query growth is fastest in 5 years but Search growth slowed, and if AI Overviews/AI Mode show fewer ads.
A: Sundar said recent models (3.8 Flash, Argon) created a step change in productivity for engineers and product work, and Carbon is a further step-up now being deployed internally. Philipp said Search grew 12% year-over-year to $63 billion and that FX, vertical dynamics, and lower monetization in AI Mode contributed to the deceleration; he added they are scaling ads in AI Overviews and AI Mode.
"But more importantly, people are tackling massive problems as well as we are beginning to see innovation on the product and the consumer experience side from deploying those models." — Sundar Pichai
Agent rollout tech limits; Search compares
Q (Brian Nowak · Morgan Stanley): He asked what technical hurdles remain to ship agentic products to all users (model efficiency, capacity, security, etc.). He also asked Anat to clarify the search performance comparison factor and how to think about Q4 compares.
A: Sundar said they are making distilled Carbon models to deploy at 1 billion user scale while keeping compute, latency and quality efficient, and they need a robust harness and secure runtime (GCP); enterprise rollout came first and consumer rollout will follow. Anat said FX headwinds will be a headwind to Q4 and they are lapping growth acceleration from late 2025, leaving about 3 points of growth to lap into Q4.
"Later this year, we will bring that same experience to consumers, obviously, starting with the 1 billion users in our Gemini app." — Sundar Pichai
Compute allocation and TPU strategy
Q (Eric Sheridan · Goldman Sachs): He asked how the company prioritizes compute as it becomes available across businesses. He also asked how Google Cloud thinks about applying compute to customer needs and external TPU usage.
A: Sundar said compute allocation is handled via long-range planning and disciplined ROIC across businesses, and models are the foundation across products. Thomas said they have over 100 significant TPU customers, have built TPU systems (hardware, software, networking), designed new silicon (8t for training, 8i for inference), optimized software like PyTorch, and are broadening distribution including on-prem and partners.
"So we have 8t for training, 8i for inference." — Thomas Kurian
TPU system sales and CapEx effect
Q (Ross Sandler · Barclays): He asked what percent of TPU system costs are recognized as revenue and the margin flow-through. He also asked whether customers building their own infrastructure will lower Alphabet's future CapEx intensity.
A: Anat said they do not give product-level financials but explained TPU systems include hardware, software, storage and networking so they sell more than a chip. She said sales show up first in backlog and inventory is built as they procure components; cash goes out to build inventory and revenue is recorded when they deliver and bill, and more system sales are expected through 2027.
"When we signed an agreement with the customer to sell TPU systems, once we sign that agreement, you'll first see that showing up in our backlog and then we start building inventory." — Anat Ashkenazi
Agentic effects on Search and TPU margins
Q (Justin Post · BofA): He asked whether the shift to agentic search could be a headwind and whether agentic could drive more monetization long term. He also asked if TPU system sales are a good business and why.
A: Philipp said agentic experiences are additive, designed to automate grunt work while preserving discovery, and they expect to build compelling advertising models and other monetization (subscriptions, payments). Sundar called agents an expansion opportunity that opens new use cases for Search. Thomas said they don't disclose product margins but see benefits from TPU systems' price performance and scale, and inventory is procured after contracts are signed.
"So given that there's almost never only one perfect option for a decision, we're actually confident that we can actually provide really interesting, compelling advertising models for agentic use cases." — Philipp Schindler
Early agentic signals; Cloud growth details
Q (Mark Shmulik · AllianceBernstein): He asked if anything in early agentic use cases surprised them, including consumer productivity. He also asked Cloud growth excluding TPU sales and how much capacity constrained growth, and whether third-party capacity coming online will accelerate Cloud.
A: Sundar said they see agentic queries in the Gemini app task tab and that repeat use requires handling long tasks well; capability needs to cross a quality threshold before it takes off. Thomas said Google Cloud revenues were up 103% to $30.7 billion in the third quarter, GCP drove most of that growth, Cloud growth excluding TPU system sales was very strong, demand was broad, and they can serve more demand but are managing capacity.
"Google Cloud segment delivered record growth in the third quarter with Cloud revenues up 103% to $30.7 billion." — Thomas Kurian
Working with competitors; safety and trust
Q (Michael Nathanson · MoffettNathanson): He asked why Google works with companies that could be competitors and whether that strategy is a long-term risk. He also asked what they've learned about consumer concerns on safety and trust for agentic tools and how they protect privacy and trust.
A: Sundar said they distinguish products from platforms, and when building platforms they aim to make them work for everyone; TPU systems are being built as a platform and exposed to market forces. On safety and trust he said users are clearly concerned, and decades of building safe consumer products plus extensive testing and real-time monitoring of models and agents will be a key differentiator that users will reward.
"I mean, to the earlier question, and I think I should have emphasized, people really -- when people are approaching these agentic experiences, we clearly see people are concerned about safety and trust." — Sundar Pichai