An Interview with Google Cloud CEO Thomas Kurian About the Agentic Moment

Google Cloud CEO Thomas Kurian framed Google Next around AI agents, citing 16 billion tokens per minute on Gemini, up from 10 billion in late 2024/early 2025, and 40% sequential growth in Gemini Enterprise. He previewed new TPU 8t and 8i chips, additional storage and compute offerings, expanded cybersecurity tools, and a cross-cloud lakehouse/Knowledge Catalog aimed at enterprise data integration. The interview signals strong momentum in Google Cloud’s AI stack and a broader monetization strategy across chips, models, security, and partner ecosystems.

Analysis

The market is still underestimating how much Google is trying to convert AI from a product feature into a systems-of-record tax. If the agent stack becomes the control plane for enterprise workflows, the winner is not just the model provider but the company that owns identity, policy, observability, storage, and compute economics in one bundle. That makes GOOGL's advantage more durable than a pure-model narrative: it can monetize both internal demand and external inference, while also using partner models as load-fillers for the same infrastructure.

The second-order bull case is that Google is quietly broadening AI demand beyond hyperscaler training budgets into latency-sensitive verticals where switching costs are high and procurement is messier. If TPU deployments move into finance, energy, and on-prem/third-party venues, that expands TAM without requiring customers to move their data, which is the right answer to egress friction and data sovereignty. This also supports a more stable utilization curve for silicon and storage assets, improving returns on capex even if consumer AI usage remains diurnal.

The main risk is execution complexity: the more Google sells a fully integrated platform, the more any gap in model quality or agent reliability becomes visible, because customers will benchmark the whole workflow, not just the model. A failure mode over the next 1-2 quarters is that enterprise adoption remains real but deployments stay shallow, while developers standardize around third-party model interfaces and Google gets compressed into infrastructure margin. The best tell will be whether the announced ecosystem expansion converts into sustained partner attach rates and not just keynote optics.

Contrarian view: the consensus is too fixated on who wins the frontier-model headline, and not enough on who captures the operating layer around agents. If agents become the interface to enterprise data and permissions, then the economically important asset is the orchestration layer, where Google has a credible chance to re-enter as a platform rather than just a cloud vendor. That makes GOOGL the cleaner long, while ORCL, CRM, and SAP are more exposed if agent-driven workflows start bypassing parts of their application surfaces.

Key Decisions for Investors

• Add to GOOGL on any post-event dip over the next 1-3 sessions; target a 10-15% upside over 3-6 months if enterprise agent adoption and TPU monetization keep compounding.
• Pair long GOOGL / short ORCL for 3-6 months: thesis is that agent control-plane economics accrue to the integrated AI platform, while legacy application vendors risk being disintermediated at the workflow layer.
• Buy a starter long in ARM for 3-6 months if the market extrapolates TPU/AI infrastructure spend into broader inference demand; risk/reward improves if Google's ecosystem narrative lifts non-Nvidia silicon adoption.
• Use CRM and SAP as relative shorts vs GOOGL over 6-12 months if customer budgets shift from app-seat expansion to agent automation spend embedded in cloud platforms.
• For higher beta, consider GOOGL call spreads 6-9 months out: limited downside, positive convexity if the market starts valuing Google as both AI platform and infrastructure monetizer.