streugut.ch
Archive

The direction

Where the AI race is heading

AI competition is moving away from a single frontier race and toward control of the full path from model to deployed workflow. That shift is carried by enterprise rollouts organized through implementation partners, by agents being embedded inside governed software environments, and by a widening compute contest from chips to cloud capacity. What is particular now is that distribution, deployment, infrastructure, and compliance are tightening into one operating system for scale.

The move

What actors actually did

Microsoft and EY commit $1 billion for enterprise AI; agent breakthroughs and infrastructure expansions mark the week

On May 21, Microsoft and EY announced a global initiative aimed at helping companies scale AI across their operations. The two firms are jointly investing more than $1 billion, introducing co-developed solutions and deploying Copilot to all EY employees. This marks a significant move in the ongoing push for enterprise-level AI adoption.

Microsoft also expanded Copilot Studio's capabilities on May 26, making computer-using agents generally available to automate actions on websites and desktop applications. Meanwhile, xAI launched the early beta of Grok Build on May 25—a terminal-based coding agent for SuperGrok Heavy subscribers—which focuses on planning and multi-step execution. Alibaba Cloud revealed new agentic AI products and its latest Qwen3.7-Max language model on May 26, alongside homegrown chips and infrastructure upgrades for global clients.

Alongside these advances, the underlying infrastructure saw notable activity. Oracle expanded its AI Supercluster capacity with GPU additions on May 22 to better train large models, and Sakura Internet in Japan announced plans on May 24 to raise capital expenditures up to sevenfold to meet AI data center demand. These moves highlight the industry’s efforts to support expanding AI workloads.

Sources