AI partnerships, under watch.

Microsoft-OpenAI restructured partnership (OpenAI PBC, Microsoft ~27%/$135B, $250B Azure commitment)

Recapitalization of OpenAI into a PBC controlled by the OpenAI Foundation. Microsoft IP rights extend to 2032 (including post-AGI models with safety guardrails), Azure API exclusivity continues until AGI, OpenAI commits $250B in incremental Azure purchases, Microsoft loses right-of-first-refusal on future compute.

Anthropic expands Google Cloud TPU usage (up to 1M TPUs, >1 GW online in 2026, tens of billions)

Large compute commitment from Anthropic to Google Cloud (up to 1M TPUs including Ironwood, >1 GW capacity in 2026). Builds on Google prior equity investments. Anthropic explicitly frames as part of a three-platform compute strategy (TPU + AWS Trainium + NVIDIA GPUs), with AWS remaining primary training partner.

Claude becomes preferred Agentforce model for regulated industries with deep Slack integration

Three-pillar expansion: Claude becomes a foundational model in Agentforce 360 with all traffic kept inside the Salesforce trust boundary/VPC (Anthropic is the first LLM provider fully integrated there); deep bidirectional Claude-Slack integration via MCP; Salesforce deploys Claude Code firmwide for engineering. Bundles ISV-embed, co-marketing, and customer-of-each-other dynamics.

Deloitte deploys Claude to 470,000 employees globally, Anthropic largest enterprise deal

Largest Anthropic enterprise deployment to date. Combines internal workforce rollout (470K seats) with a Claude Center of Excellence, certification of 15,000 Deloitte practitioners, and co-developed regulated-industry solutions for financial services, healthcare, and public sector. Functions as both seat deal and channel/services play.

OpenAI-Oracle 4.5 GW Stargate data center expansion (5+ GW total Stargate capacity, 2M+ chips)

Joint infrastructure development under the Stargate JV (OpenAI, Oracle, SoftBank, MGX). Oracle builds and delivers 4.5 GW of incremental US data center capacity dedicated to OpenAI workloads, on top of the Abilene TX Stargate I site, advancing the $500B/10 GW commitment announced January 2025.

xAI Grok models available on Oracle Cloud Infrastructure Generative AI

Oracle hosts xAI Grok 3 models in its OCI Generative AI service for enterprise customers under zero-data-retention terms, while xAI uses OCI infrastructure to train and run inference for next-generation Grok models. Bidirectional model-hosting plus compute supply arrangement.

Meta $14.3B strategic investment in Scale AI with Alexandr Wang joining Meta

Meta invested $14.3B for a 49% non-voting stake in Scale AI (~$29B valuation), pulling founder Alexandr Wang into Meta superintelligence org while keeping Scale operationally independent. Functions as an acqui-hire plus a deepened data-labeling supply relationship.

Five-year deal: Claude natively available on Databricks Data Intelligence Platform

Five-year model-hosting arrangement under which Anthropic Claude family is served natively inside Databricks Data Intelligence Platform on AWS, Azure, and GCP, with Mosaic AI tooling and Unity Catalog governance, so customers can build agents on enterprise data without replicating it.

SoftBank becomes authorized reseller of Perplexity Enterprise Pro in Japan

SoftBank, already a 2024 investor in Perplexity at a $3B valuation, became the first authorized reseller of Perplexity Enterprise Pro and committed 7,000 sales reps to drive Japanese enterprise adoption. Investment relationship extended into a distribution channel partnership.

Anthropic-AWS expanded partnership ($8B total investment, AWS as primary cloud and training partner)

Strategic investment plus deep infrastructure tie-up: AWS becomes Anthropic primary cloud and training partner, with Anthropic co-developing future Trainium accelerators with Annapurna Labs and distributing Claude through Amazon Bedrock. Amazon remains a minority investor at $8B total.

Snowflake and Anthropic strategic partnership bringing Claude to Cortex AI

Anthropic Claude models hosted natively inside Snowflake Cortex AI service, giving Snowflake customers governed access to Claude against their warehoused data. Expanded in Dec 2025 to a $200M multi-year agreement with joint GTM for agentic AI deployment.

Accenture NVIDIA Business Group launches AI Refinery for enterprise agentic AI

Accenture stands up a dedicated NVIDIA Business Group with 30,000 trained professionals and an AI Refinery platform built on the full NVIDIA AI stack (Foundry, Enterprise, Omniverse). Structure is a joint go-to-market vehicle inside Accenture consulting org rather than a reseller arrangement, with new engineering hubs in Singapore, Tokyo, Malaga, and London.

Palantir to run Foundry, Gotham, Apollo, AIP on Azure Government Secret/Top Secret clouds with Azure OpenAI

Defense/intel sovereign-cloud integration: Palantir full product suite is accredited to run inside Azure Government, IL6 Secret, and Top Secret clouds, with Azure OpenAI (GPT-4 and successors) callable from Palantir AIP. Targets DoD and IC missions; availability gated on government accreditation.

Apple Intelligence integrates ChatGPT into Siri and system writing tools on iOS 18, iPadOS 18, macOS Sequoia

OpenAI ChatGPT (GPT-4o) is embedded as an optional, user-permissioned extension to Siri and system writing tools on Apple devices. Apple obscures IP addresses and OpenAI agrees not to store requests for non-account users. No disclosed equity or cash exchange; ChatGPT is one of several models Apple plans to plug in.

IBM hosts Mistral models on watsonx and adds on-prem Mistral AI with IBM deployment

Multi-stage model-hosting partnership. Started Feb 2024 with open-source Mistral on watsonx, expanded through Mistral Large 2, Medium 3, and Large 3 in watsonx.ai catalog, and culminated in Mistral AI with IBM on-prem add-on for regulated EU customers. Extends IBM open multi-model strategy and gives Mistral a sovereignty-friendly enterprise distribution channel.

Microsoft takes EUR 15M stake in Mistral and hosts Mistral Large on Azure

Multi-year arrangement bundling three pieces: (1) Azure supercompute capacity for Mistral training, (2) Mistral Large distributed as Models-as-a-Service in Azure AI Studio under MACC, and (3) a small EUR 15M equity stake plus joint R&D for European public-sector workloads.

McKinsey and Cohere strategic alliance for enterprise generative AI deployment

McKinsey QuantumBlack arm partners with Cohere to deploy Cohere enterprise LLMs (including on-prem) inside McKinsey client environments, combining Cohere data-privacy-first model deployment with McKinsey implementation and change-management muscle. SI-led joint delivery model.

Oracle and Cohere partner to deliver generative AI services on OCI and embedded across Oracle apps

Combined infrastructure-and-embedding deal: Cohere trains and runs its foundation models on OCI Supercluster GPUs while Oracle embeds those models across Fusion, NetSuite, and industry SaaS. Oracle also participated in Cohere funding round, layering strategic investment on top of commercial integration.

ServiceNow and NVIDIA partner to build custom enterprise LLMs on the Now Platform

Joint development arrangement in which ServiceNow uses NVIDIA NeMo framework, DGX Cloud, and DGX SuperPOD to train ServiceNow-specific LLMs powering Now Assist. ServiceNow is also a reference customer for the same stack internally, and the pact has since expanded into agentic AI co-development.

Zoom invests in Anthropic and integrates Claude across Zoom platform starting with Contact Center

Combined commercial-plus-equity deal: Zoom Ventures took a stake in Anthropic (~$51M, later marked to ~$1.27B per Zoom filings) and Zoom adopted Claude across its federated AI stack, first shipping in Zoom Contact Center. Pattern is strategic_invest paired with ISV-embed in a customer-facing product line.