Build 2026 Unveils Agentic Future, SoftBank Fuels Compute Race, and Copilot's New Billing Sparks Dev Outcry

Signals from the Latent Space

Today, the AI landscape is a dynamic mix of ambitious product rollouts, massive infrastructure plays, and evolving developer economics. Microsoft’s annual Build conference sets the stage for an agentic future, while a colossal investment from SoftBank underscores the relentless demand for compute. Developers, however, are grappling with a significant shift in how they pay for AI-powered coding assistance, and enterprises gain new tools for building trustworthy agents. Simultaneously, the patchwork of US AI regulation continues to take shape, presenting a complex environment for deployment.

Microsoft Build 2026: The Agentic Future Takes Center Stage

Microsoft’s Build 2026 developer conference is underway, with CEO Satya Nadella highlighting the company’s vision for an agent-centric AI ecosystem. Key announcements are expected to revolve around a ‘Copilot super app,’ a new reasoning AI model from Microsoft AI, and a developer-optimized Windows 11 experience. This push signifies Microsoft’s intent to evolve Copilot beyond a mere chatbot into an ‘async coworker’ capable of executing complex, long-running tasks across various domains. The conference agenda heavily features topics like shipping smarter AI systems, agent capabilities, and streamlined developer workflows, indicating a deep integration of AI into the core developer experience.

Why it matters: For developers, this means new SDKs, model access patterns, and Windows developer ergonomics are on the horizon. The focus on agent capabilities suggests a future where AI assistants are not just generating code but actively participating in development cycles, potentially automating more complex tasks and improving overall productivity. This also signals a significant investment in making Windows a more potent platform for local AI development and deployment.

SoftBank Pledges €75 Billion to Build Europe’s Largest AI Data Centers

In a monumental move, SoftBank Group has announced plans to invest up to €75 billion (approximately $87 billion USD) to develop and operate 5 gigawatts of AI data center capacity across France. This represents the largest single AI infrastructure investment in European history and SoftBank’s most substantial infrastructure commitment outside the United States. The initial phase, valued at €45 billion, aims to deliver 3.1 GW of capacity across three sites in the Hauts-de-France region by 2031. This announcement, made at the Choose France summit, reflects a strategic effort to establish Europe as a major AI hub, with France positioning itself as the continent’s compute capital.

Why it matters: This massive investment underscores the insatiable demand for compute power driven by the AI boom. For developers and enterprises, this means a significant boost in available, localized AI infrastructure in Europe, potentially leading to reduced latency, improved data sovereignty, and a more robust ecosystem for deploying large-scale AI models and applications. It also highlights the ongoing global race to build the foundational infrastructure necessary for the next generation of AI.

GitHub Copilot’s Token-Based Billing Sparks Developer Backlash

Effective June 1, 2026, GitHub Copilot has transitioned to a usage-based, token-centric billing model across all its plans. This change replaces the previous premium request allowance with a monthly pool of GitHub AI Credits, consumed by actual token usage (input, output, and cached tokens) at published API rates. While plan prices remain unchanged, the shift has been met with immediate and vocal criticism from the developer community, with phrases like “What a joke” trending on platforms like Reddit and X. Developers are expressing concerns over the unpredictability of costs, especially for longer, agentic coding sessions which now consume significantly more credits.

Why it matters: This change directly impacts the economics of using AI-powered coding assistants for many developers. The move from a predictable, fixed allowance to a variable, token-based system introduces uncertainty and could lead to unexpected costs for heavy users or those engaging in more complex, iterative AI-assisted development workflows. It forces developers to be more mindful of token consumption and could influence the adoption and integration of Copilot into daily coding practices.

Workday Unveils New Tools for Building and Verifying Enterprise AI Agents

At Workday DevCon, Workday, Inc. introduced new agentic capabilities within Workday Build, its platform for creating custom AI apps and agents. These include Developer Agent, allowing developers to build AI apps and agents using natural language within existing agentic tools like Claude Code and Google Antigravity; Agent-Ready Tools, providing guardrails for agents to securely access HR and finance data via the Model Context Protocol (MCP); and Agent Passport, which verifies agent safety and compliance against standards like OWASP LLM Top 10 and NIST AI RMF. These tools aim to address the critical need for secure, accurate, and governed AI applications in sensitive enterprise environments.

Why it matters: For enterprise developers, these new tools are crucial for accelerating the safe and compliant deployment of AI agents in high-stakes areas like HR and finance. The emphasis on guardrails, security verification, and integration with existing agentic development tools helps mitigate risks associated with AI errors in critical business functions, such as payroll or employee data management. This marks a significant step towards making enterprise-grade AI agent development both faster and more trustworthy.

US States Diverge on AI Regulation: Connecticut Enacts, Colorado Scales Back

The landscape of AI regulation in the United States continues its fragmented evolution, with recent developments showcasing divergent state-level approaches. Connecticut’s governor signed into law the Connecticut Artificial Intelligence Responsibility and Transparency Act on May 29, 2026. This multi-part law imposes targeted requirements across several high-profile AI use cases, including chatbots, synthetic media, and automated decision-making, taking a modular approach rather than a single unified framework. Conversely, on May 14, 2026, Colorado’s Senate Bill 26-189 (Revised Colorado AI Act) was signed into law, repealing and reenacting the 2024 Colorado Artificial Intelligence Act with significant amendments. The revised law pares down several of the original’s more burdensome obligations, such as mandatory risk management programs and broad “duty of care” obligations, favoring transparency and disclosure over prescriptive governance requirements.

Why it matters: This divergence highlights the ongoing challenge for developers and companies operating across state lines. While Connecticut opts for a targeted, modular framework, Colorado has chosen to ease its initial, more stringent requirements. This creates a complex compliance environment where developers must navigate a patchwork of regulations, making it essential to understand the specific requirements of each jurisdiction where their AI systems are deployed. It also signals that the debate over the optimal approach to AI governance is far from settled.

The Bottom Line

The past 24 hours reveal an AI ecosystem in rapid maturation, characterized by strategic investments in compute, a push towards more capable and secure AI agents, and a complex regulatory environment. Developer tools are evolving to support this agentic shift, but not without friction, as seen in the Copilot billing changes. Ultimately, the focus is increasingly on practical, governable, and scalable AI, driving both innovation and the need for careful consideration of deployment and compliance.