Sundar Pichai took the stage in Las Vegas. Google Cloud Next 2026 kicked off with a bang. The CEO laid out numbers that demand attention: $70 billion in annual revenue. Forty-eight percent year-over-year growth. A $240 billion backlog, more than double last year’s $155 billion. Customers outpacing commitments by 30%. Billion-dollar deals in 2025 alone topping the previous three years combined. And capex? $175 billion to $185 billion this year, nearly double 2025’s spend.
Pichai calls it “a fundamental rewiring of technology and an accelerant of human ingenuity.” Boom. The agentic era isn’t coming. It’s here. Google positions its cloud as the operating system for enterprises where AI agents run the show—autonomously handling operations, chatting across platforms, dipping into commerce and search. Full-stack control gives them the edge: custom silicon like Ironwood TPUs, frontier Gemini models, the cloud platform itself, Workspace, and Search reaching three billion users.
Gemini metrics stun. The app hits 750 million monthly active users, up 100 million in a quarter. AI Overviews serve two billion users in 200-plus countries, sparking 10% more queries, triggering on 48% of them—up from 31% in February 2025. Gemini API cranked 85 billion requests in January 2026, a 142% jump from March 2025’s 35 billion. Eight million paid Gemini Enterprise seats across 2,800 companies. Thirteen million developers on generative models. Gemini 3 Pro? Fastest adoption ever.
Search Transforms. Agents Take Over.
Search evolves from retrieval to “agent manager.” Pichai envisions it dispatching tasks, running multiple threads. “A lot of what are just information-seeking queries will be agentic in Search. You’ll be completing tasks,” he said on a podcast. Shopping agents? He’s tired of being “an indecisive shopper” and awaits agents bridging discovery to purchase.
Enter Universal Commerce Protocol. Open-source, co-built with Shopify, Etsy, Wayfair, Target, Walmart. Backed by Adyen, American Express, Best Buy, Flipkart, Macy’s, Mastercard, Stripe, Home Depot, Visa, Zalando. UCP runs on REST and JSON-RPC. It weaves in Agent2Agent protocol, Model Context Protocol, Agent Payments Protocol. Stores become programmable services. Merchants keep the records. Agents buy on your behalf.
Agent2Agent matures fast. Now at version 1.2, governed by Linux Foundation’s Agentic AI Foundation. In production at 150 organizations—not pilots. Microsoft, AWS, Salesforce, SAP, ServiceNow run it. A2A handles agent-to-agent talk across boundaries. Complements Anthropic’s MCP for tool connections. Google rolled managed MCP servers for Maps, BigQuery, Compute, Kubernetes. Apigee bridges APIs to agent tools.
Thomas Kurian, Google Cloud CEO, slams rivals: “Competitors are handing you the pieces, not the platform.” Google’s stack: eighth-gen TPUs split for training (Broadcom) and inference (MediaTek). Gemini 3 Pro, 3 Flash, 3.1 Pro. Gemini Enterprise Agent Platform—ex-Vertex AI. Workspace Studio for no-code agents. Search and Android for reach. Microsoft Copilot lags with 3.3% paid adoption in 365, negative NPS. Google’s eight million seats grow quicker.
Hardware Heats Up. Partners Pile In.
New TPUs target agentic workloads. TPU 8t for training: 3x faster, 9,600 chips per superpod, 121 exaflops. TPU 8i for inference. Virgo Network links up to 134,000 TPUs at 47 petabits per second. A5X with NVIDIA Vera Rubin cuts inference costs 10x, boosts throughput per megawatt. Google Cloud snagged NVIDIA as AI Global Tech Partner and Infra Modernization Compute winner at Next.
A $750 million agentic AI partner fund launched. Targets value ID, prototyping, building, deployment, upskilling. Accenture, KPMG, others join. KPMG hit 90% Gemini Enterprise adoption in a month. GE Appliances runs 800-plus agents in manufacturing. ServiceNow inks $1.2 billion, five-year deal. Cognizant drops Agentic Retail CX on Gemini for omnichannel retail. Salesforce integrates for cross-platform agents in Slack, Workspace.
Security gets agentic too. Threat Hunting, Detection Engineering agents probe malware, auto-generate rules. Wiz AI Application Protection scans code to runtime, multi-cloud. Google Unified Security ties it together. Agent Gateway enforces policies on MCP, A2A traffic. Agent Identity uses crypto for traceable auth. Model Armor blocks prompt injection, leaks.
Seventy-five percent of Google’s new code? AI-generated. Up from 25% late 2024, 50% last fall. Agents sped a code migration sixfold. Engineers review; some teams use Claude Code. Internally, it proves the point. Externally, Gemini Enterprise Agent Platform unifies: Agent Studio (low-code), orchestration for delegation, registry, observability (OTel-compliant). Project Mariner browses web. Workspace Intelligence fuses Gmail, Docs, Drive for auto-reports, actions.
Capex breaks down: 60% servers, 40% data centers, networking. Big tech AI spend nears $700 billion. Constraints linger through 2026—power, land, supply chains. Google Cloud’s 11% share trails AWS (31%), Azure (25%), but 48% Q4 growth and profitability shine. Backlog covers three-plus years revenue. Thirteen product lines over $1 billion each.
Agentic Data Cloud shifts data from repository to reasoning engine. Cross-Cloud Lakehouse queries AWS, Azure data via Iceberg—no moves needed. Knowledge Catalog, Smart Storage extract context from PDFs, images. C4N/M4N VMs handle agent traffic. Fluid compute across Engine, Kubernetes.
And the code stat lingers. Seventy-five percent AI-written. Humans check. But as agents scale, review bottlenecks loom. Google bets its stack—from chip to inbox—captures value beyond share. Rivals fragment. Google integrates. The race? Not models alone. Agents. Interop. Governance. Production at scale.
Pichai’s closing vision: Search as conductor. Agents as orchestra. Enterprises transformed. Supply tight. Demand exploding. Google Cloud rides the wave.
WebProNews is an iEntry Publication