The Signal

AWS and NVIDIA announced a major expansion of their AI infrastructure partnership, with plans to deploy 2 million additional NVIDIA GPUs across AWS’s global infrastructure. The collaboration extends beyond accelerators to NVIDIA Vera CPUs, networking, Nemotron open models, data processing and physical-AI technologies. The scale comes alongside NVIDIA fiscal Q2 revenue of US$96.2 billion, up 106% year over year, while Data Center revenue reached US$89.0 billion, up 117%. The signal is broader than GPU supply: hyperscale competition is increasingly about owning an integrated AI stack spanning compute, networking, CPUs, models and cloud capacity.

Google Cloud launched Gemini Enterprise for Financial Services, initially targeting capital markets and corporate banking. The offering combines a Google-managed Financial Research agent, 50+ specialized financial-services skills, enterprise data connectors and access to third-party agents. CME Group is an early user and Deutsche Bank served as a design partner, with BNY, Citi Wealth, Lloyds Banking Group, Macquarie Bank and Signal Iduna also identified as Gemini Enterprise users. The launch signals a move from general-purpose assistants toward AI platforms packaged around regulated industry workflows, specialist knowledge and enterprise governance.

Google Cloud introduced Gemini Enterprise for Legal, a purpose-built agentic platform combining specialized legal skills, enterprise connectors, third-party agents and Gemini Enterprise governance. The preview includes Cleary Gottlieb, Freshfields, Weil and Williams & Connolly, while the ecosystem includes Thomson Reuters, Harvey, Relativity, iManage, NetDocuments and Docusign. Reuters also reported the expansion. Legal AI is therefore evolving from individual copilots toward integrated enterprise platforms where proprietary information, permissions, workflow integration and ecosystem depth may matter more than standalone model benchmarks.

Salesforce and Anthropic expanded their partnership with Claudeforce, connecting Claude reasoning to Salesforce data, workflows, business logic, actions and governance. The initial Salesforce-in-Claude plugin includes 37 prebuilt sales skills, while Salesforce reported Agentforce and Data 360 ARR of US$3.9 billion and quarterly revenue of US$11.35 billion. The partnership shows frontier-model providers and systems-of-record vendors converging around a common objective: giving models secure business context and permission to execute governed transactions rather than merely generate answers.

Thomson Reuters launched Thomson, its first proprietary large language model, after investing US$40 million in talent and compute. The company started with an open-source foundation and specialized it using proprietary professional information and expert supervision; it says less than 10% of its content has been used in training so far. Initial deployment is in CoCounsel Legal, with additional capabilities and sovereign-AI options planned. The development demonstrates a credible enterprise model strategy: organizations with high-value proprietary data can create differentiated domain models without frontier-lab-scale expenditure.

Okta made Agent SSO generally available, extending enterprise identity controls to AI agents. Organizations can register an agent as a workload principal and apply centralized identity and access policies rather than relying on static API keys or repeated user authorization. Okta serves 20,000+ customers, while its research says only 34% of organizations currently apply the same security controls to AI agents as to human workers. Agent identity, permissioning, auditability and lifecycle management are rapidly becoming foundational components of enterprise AI governance.

Mistral AI and Saudi AI company HUMAIN announced a strategic collaboration covering AI infrastructure, advanced model development and AI solution deployment in Saudi Arabia and the broader Middle East. Financial terms were not disclosed. The agreement links a European model developer with Saudi Arabia’s push to build locally controlled AI capabilities, reinforcing a broader shift toward sovereign AI as a full-stack strategy spanning compute, models, infrastructure, data control and applications.

Emerald AI raised an oversubscribed US$150 million Series A at a US$1.05 billion valuation, co-led by Energize Capital and DCVC. The company develops software that allows AI data centers to dynamically adjust electricity demand and says the technology could unlock up to 100 GW of capacity from the existing U.S. grid. It is also working on a nearly 100-MW Vera Rubin AI Research Factory in Virginia, following five global demonstrations, and counts 12 Fortune Global 500 companies among its investors. The financing reinforces electricity as a strategic AI infrastructure constraint alongside compute.

Digital Realty received a provisional 50-MW allocation under Singapore’s second Data Center Call for Application for a new facility at Jurong Island’s low-carbon data-center park. The facility is planned for AI inference, high-performance computing and enterprise workloads and must source more than 50% of its capacity through green-energy pathways. The wider process allocated 200 MW across four operators. Singapore shows how governments in power-constrained markets are increasingly determining where, how and under what sustainability conditions AI infrastructure can expand.

India’s Vice President C. P. Radhakrishnan formally launched Gnani Artha, a sovereign-AI stack combining Evon 3.3 with the Plexus agentic platform. Independent reporting identifies Evon 3.3 as a 30-billion-parameter open-weight model designed for Indic languages and domain-specific applications and capable of running within an organization’s own infrastructure. The launch indicates India’s sovereign-AI push is moving from policy and public infrastructure toward deployable domestic models and agentic platforms.

Uniphore and Tech Mahindra announced an Agentic AI Factory designed to build domain-specific small language models and AI agents for enterprises. Initial use cases include Retail and BFSI workflows such as strategic sourcing, unclaimed benefits and claims processing. Uniphore says it is trusted by 2,000+ enterprises, while Tech Mahindra has 146,000+ professionals across 90 countries. The initiative signals industrialization of agent implementation as systems integrators move from bespoke pilots toward repeatable factory models.

IndiaFirst Life Insurance selected Salesforce Agentforce for an enterprise-wide AI transformation spanning sales, operations and customer experience. Its existing Salesforce environment already supports processes from lead generation through claims settlement, giving AI agents access to workflows much closer to the core insurance operating model than a conventional productivity deployment. Financial value and deployment counts were not disclosed. The strategic question for BFSI is shifting from whether agents can be adopted to which workflows can be safely delegated under identity, audit and human-oversight controls.

Rezolve AI and Tech Mahindra announced a global alliance to deploy agentic-commerce capabilities for large enterprises. The partnership combines Rezolve’s commerce AI with Tech Mahindra’s implementation footprint of 1,100+ clients, 146,000+ professionals and operations in 90 countries. The objective is to move AI beyond customer-service chatbots toward systems that understand purchase intent, recommend products and participate directly in transaction completion. Financial terms were not disclosed. As agents move toward revenue-generating actions, commerce architecture, payments, identity, permissions, fraud controls and auditability are likely to converge.