The wire, auto-curated daily.

Anthropic reaches $965 billion valuation amid infrastructure and regulatory shifts

The company secures a massive Series H raise and navigates partial lifting of US government access restrictions for its models.

Anthropic has secured a Series H funding round valuing the company at $965 billion, surpassing OpenAI in private market valuation and solidifying its position as a dominant infrastructure player. This financial milestone coincides with developments in US government relations, where the Commerce Department partially restored access to Claude Mythos 5 for critical infrastructure defenders, though the Department of Defense lawsuit regarding usage in autonomous weapons systems remains active. The company is also deepening its compute strategy, with reports of significant multi-year compute capacity agreements and potential acquisitions like Cursor for $60 billion.

The valuation reflects a market shift where capital is flowing toward companies controlling both the models and the underlying compute infrastructure. As the EU AI Act enforcement approaches on August 2, 2026, Anthropic's scale places it at the center of global governance discussions, balancing innovation speed with regulatory compliance requirements.

South Korea commits $880 billion to AI chips, robotics, and data centers over a decade

The massive investment plan aims to secure the nation's position in the global AI supply chain.

South Korea has confirmed an investment plan of over $880 billion dedicated to AI chips, robotics, and data centers over the next ten years. The initiative includes a substantial portion allocated to four memory chipmaking plants, with major contributions from industry leaders Samsung and SK Hynix. This strategic push aims to secure the country's role in the global AI infrastructure ecosystem, countering supply chain risks and expanding domestic capacity for the hardware powering the next generation of models.

The announcement highlights a growing trend of national-level strategic investments in AI infrastructure, similar to the "Pax Silica" coalition expansion to 35 nations. With India recently seeking assurances against unilateral access cuts, South Korea's commitment reinforces the geopolitical competition for control over the physical and digital layers of AI technology.

Enterprises pivot from "tokenmaxxing" to model routing efficiency as costs rise

Companies are shifting strategies to manage the high expense of frontier models by routing tasks to cheaper alternatives.

Enterprises are increasingly moving away from "tokenmaxxing"—the strategy of using the most capable model for every task—due to unsustainable costs, with some companies burning their entire annual AI budgets in just four months. The new approach focuses on model routing efficiency, where bulk and routine tasks are handled by smaller, cheaper models, while only complex, high-value problems are escalated to frontier models like GPT-5.6 Sol or Claude Mythos 5. Case studies show companies migrating significant traffic to models like DeepSeek V4-Pro to achieve cost reductions of millions of dollars annually.

This shift reflects a maturation in the market where the focus moves from pure capability to practical deployment economics. As OpenAI introduces lower-priced tiers like Terra and Luna, the industry is seeing a more nuanced adoption pattern where cost-efficiency and performance are balanced more carefully in production workflows.

ServiceNow Knowledge 2026 marks the shift from AI experimentation to governed autonomous execution

The Las Vegas event brought ~25,000 attendees and CEO Bill McDermott's message that enterprises are moving beyond pilots to reality checks focused on orchestration and governance at scale.

The event introduced major capabilities including the upgraded AI Control Tower as a full Governance Engine with five dimensions (Discover, Govern, Secure, Observe, Measure), Action Fabric enabling cross-platform AI agent orchestration, and Autonomous CRM where AI agents run processes end-to-end. New AI Specialists now cover every major enterprise function, while ServiceNow Otto provides a unified natural language interface. Partnerships expanded to 30 new integrations including NVIDIA's Arc Framework and Anthropic's Claude.

Action Fabric enables headless AI workflows across the enterprise, connecting ServiceNow to Microsoft Copilot, Zoom AI Companion, and beyond

Employees can now prompt external AI assistants to complete tasks in ServiceNow without leaving their current application, enabling truly cross-system processes under unified governance.

This capability follows the Salesforce Headless 360 pattern announced at TDX 2026 and represents a strategic shift toward ServiceNow as the governed execution layer for enterprise AI. The Model Context Protocol (MCP) server exposes the workflow engine to external agents like Anthropic Claude, allowing them to trigger ServiceNow workflows under the same governance controls. This addresses the "patchwork enterprise" problem where AI layered onto disconnected systems creates governance gaps.

April 2026 brought major ITOM licensing changes, collapsing the tier structure to Advanced and Prime with a new credit-based meter

ITOM now omits the Foundation tier entirely, with Advanced focused on visibility and CMDB integrity, while Prime adds event correlation, log analytics, and full AIOps capabilities.

The new unified credit meter allows fluid scaling between ITSM and ITOM (1 Credit = 1 ITOM SU, 10 Credits = 1 ITSM Fulfiller User), and combined Service Ops SKUs are available for fused IT Operations and SRE teams. Additionally, the broader April 9 2026 pricing changes replaced five legacy tiers with three AI-native tiers (Foundation, Advanced, Prime), bundling Now Assist, Moveworks, Workflow Data Fabric, and AI Control Tower into every tier rather than as separate add-ons.