Alphabet Inc. opened its annual Google Cloud Next conference in Las Vegas today, unveiling a suite of advanced artificial intelligence tools centered on autonomous enterprise agents and the next generation of its Gemini model. Google Cloud CEO Thomas Kurian detailed the company’s transition from assistive AI to agentic systems capable of executing complex workflows with minimal human oversight. The keynote highlighted significant upgrades to the Vertex AI platform and the introduction of specialized hardware designed to support massive-scale inference.
Central to the announcements was the debut of Gemini 3.0, the latest iteration of Alphabet’s multimodal large language model. According to Kurian, Gemini 3.0 features a 5-million-token context window and improved reasoning capabilities for industrial applications. The model is now integrated across the Google Cloud stack, including a new Agent Builder tool that allows corporations to deploy autonomous agents for customer service, software development, and supply chain logistics. Kurian stated that over 70% of generative AI unicorns are currently utilizing Google Cloud infrastructure.
On the hardware front, Alphabet introduced the Tensor Processing Unit (TPU) v7, its most powerful custom AI accelerator to date. The TPU v7 is reported to deliver a 2.5x increase in training performance and a 4x improvement in inference efficiency compared to the previous generation. This hardware powers the new AI Hypercomputer architecture, which integrates liquid-cooled systems and high-bandwidth networking to support models exceeding 10 trillion parameters. The company also announced the general availability of its Axion processors, custom ARM-based CPUs designed for general-purpose cloud workloads.
Alphabet also expanded its global footprint, announcing three new cloud regions in Riyadh, Warsaw, and Bangkok to meet increasing demand for data sovereignty. These regions will feature Sovereign AI capabilities, allowing government and highly regulated enterprise clients to maintain local control over data used for model training. The company confirmed that its cloud division reached a $48 billion annual revenue run rate in the most recent fiscal quarter, driven by high demand for AI-optimized infrastructure.
In a joint presentation, Alphabet CEO Sundar Pichai and Kurian showcased partnerships with several Fortune 500 companies. Mercedes-Benz announced the deployment of Gemini-powered agents across its global manufacturing facilities to optimize assembly line efficiency, while Goldman Sachs detailed its use of Google’s Vertex AI for automated financial reporting. The conference, which continues through April 22, is expected to draw over 35,000 attendees, marking the largest gathering in the event's history.