The Gist — Geopolitics & Sovereign Positioning

Top Line

China has moved to restrict state enterprises and government agencies from using OpenClaw AI on office computers, marking Beijing's first policy response to the agentic AI wave and signalling a tightening of tech sovereignty controls amid escalating U.S.-China AI competition.

The Trump administration's legal and regulatory battle with Anthropic is intensifying, with the White House preparing a new executive order targeting the company while Microsoft has intervened to support Anthropic's lawsuit challenging its designation as a supply chain risk — a dispute that could reshape the boundaries of government AI procurement power.

Iran's bombing of Gulf datacenters has made AI infrastructure a direct target in warfare for the first time, forcing U.S. tech firms and regional allies to confront the vulnerability of concentrated cloud computing assets and potentially accelerating sovereign computing strategies in allied states.

Former OpenAI CTO Mira Murati's Thinking Machines has secured a multibillion-dollar, multi-year compute partnership with Nvidia including a strategic investment, marking one of the largest AI infrastructure deals to date and underscoring Nvidia's role as kingmaker in determining which companies can compete at frontier scale.

Google has expanded Gemini AI integration across its Workspace suite in India with support for eight local languages, a significant push into the Global South AI market that positions Google ahead of U.S. rivals in capturing emerging economy AI adoption and enterprise workflows.

Key Developments

China restricts state use of OpenClaw AI in first policy response to agentic AI

Chinese authorities have moved to limit state-run enterprises and government agencies from running OpenClaw AI applications on office computers, according to Bloomberg. The directive comes as companies and consumers across China began experimenting with the agentic AI platform, prompting Beijing to act swiftly to contain potential security risks. The move represents China's first major policy intervention in response to the rapid proliferation of agentic AI systems that can autonomously perform tasks across digital environments.

The restriction targets core state institutions including banks and government offices, indicating Beijing's concern that foreign agentic AI could access sensitive systems, exfiltrate data, or create dependencies on Western AI infrastructure. The speed of the policy response — coming just as OpenClaw usage began spreading in China — suggests authorities view agentic AI as categorically different from previous generative AI tools in terms of security risk, given these systems' ability to navigate networks and execute complex multi-step operations without human supervision.

Why it matters

This is China's first policy drawing a red line around agentic AI in state systems, signalling that Beijing sees autonomous AI agents as a national security threat requiring immediate containment, potentially previewing similar restrictions on other foreign agentic platforms and accelerating China's push for domestically-controlled AI agent infrastructure.

What to watch

Whether China extends these restrictions to private sector companies, particularly those in sensitive industries, and how quickly Chinese firms develop domestic agentic AI alternatives to fill the gap — velocity here will indicate whether this is a temporary security measure or the opening of a new front in tech decoupling.

Anthropic-Pentagon confrontation escalates with Microsoft intervention and threatened executive order

The legal and regulatory clash between Anthropic and the Trump administration is intensifying on multiple fronts. The White House is preparing an executive order targeting Anthropic even as the company's lawsuit challenging its designation as a supply chain risk proceeds in federal court, WIRED reports. Anthropic has told the court it could lose billions of dollars in revenue this year from the supply chain risk designation, which effectively blocks U.S. government agencies from procuring its AI services, according to Bloomberg.

Microsoft has now entered the dispute, backing Anthropic in its legal fight with the Pentagon, Financial Times reports. The software giant's intervention adds significant corporate weight to Anthropic's First Amendment arguments that the government cannot coerce a private company to rewrite its code to serve government surveillance purposes. The confrontation originated when Anthropic refused to allow the Pentagon to use its technology to spy on Americans, prompting the Department of Defense to retaliate with the supply chain risk designation.

Why it matters

This is the first major test of how far the U.S. government can compel AI companies to participate in national security applications over the companies' objections, with implications for whether frontier AI firms can maintain independence from military and intelligence use cases or will be forced to choose between principles and access to the world's largest government procurement market.

What to watch

The federal court's ruling on whether the supply chain risk designation violates Anthropic's First Amendment rights, and whether the threatened executive order materialises — if it does, watch for its specific legal mechanism and whether other AI firms face similar designations for refusing government contracts.

Iran targets Gulf datacenters in first use of AI infrastructure as warfare objective

Iran is bombing datacenters in Gulf states as part of its conflict with the United States, marking the first time AI and cloud computing infrastructure has been deliberately targeted as a military objective, The Guardian reports. The attacks are aimed at destroying symbols of alliance between Gulf states and the U.S., bringing the war directly into the daily lives of millions of people who depend on these facilities for digital services. The targeting of datacenters represents a strategic shift in modern warfare, where control over and denial of AI computing capacity becomes a tangible military goal.

The vulnerability of concentrated datacenter infrastructure is particularly acute in the Gulf region, where nations including the UAE and Saudi Arabia have invested heavily in becoming regional AI hubs through partnerships with U.S. hyperscalers. These facilities not only serve commercial cloud customers but increasingly host AI training and inference workloads for both civilian and potential military applications. Iran's calculus appears to be that destroying this infrastructure both degrades U.S. allied capabilities and demonstrates the fragility of concentrating AI power in fixed, targetable locations.

Why it matters

The deliberate military targeting of AI datacenters establishes a precedent that will force nations and companies to fundamentally rethink the geopolitical risk of concentrating AI computing in specific locations, likely accelerating sovereign computing strategies, geographic distribution of AI infrastructure, and potentially driving AI capabilities toward more distributed or mobile architectures less vulnerable to conventional strikes.

What to watch

Whether Gulf states and hyperscalers respond by hardening datacenter defenses, distributing AI workloads more widely, or relocating critical infrastructure — and whether other nations adopt Iran's playbook of targeting AI infrastructure to impose costs on adversaries, which would mark a broader shift in how wars are fought in the AI era.

Thinking Machines secures multibillion-dollar Nvidia compute partnership in major infrastructure deal

Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, has secured a multibillion-dollar, multi-year compute partnership with Nvidia that includes at least a gigawatt of computing power and a strategic investment from the chipmaker, TechCrunch and Financial Times report. The deal represents one of the largest AI infrastructure commitments to date and cements Nvidia's position not just as a hardware supplier but as a strategic partner determining which companies can compete at frontier AI scale.

The gigawatt-scale commitment is significant — equivalent to the power consumption of a medium-sized city — and indicates Thinking Machines is positioning to train models at the same computational scale as OpenAI, Anthropic, and Google. Nvidia's willingness to make both a compute commitment and a direct investment suggests the chipmaker views Murati's venture as a potential peer to current frontier labs, effectively using its control over scarce AI infrastructure to shape the competitive landscape by choosing which startups receive the resources needed to build cutting-edge systems.

Why it matters

This deal demonstrates that Nvidia has become a de facto gatekeeper to frontier AI competition through its control of both advanced chips and the infrastructure partnerships required to deploy them at scale, giving the company outsized influence over which nations and companies can develop leading AI systems regardless of their capital or talent advantages.

What to watch

Whether Nvidia announces similar partnerships with other AI startups or national AI initiatives, which would reveal the company's strategy for allocating scarce frontier computing capacity across different geopolitical actors — and whether governments respond by seeking to reduce dependence on Nvidia-controlled infrastructure through domestic chip production or alternative architectures.

Google expands Gemini in India with local language support in Global South AI positioning

Google has expanded its Gemini AI assistant in Chrome to India with support for eight local languages including Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Telugu, and Tamil, TechCrunch reports. The company has also rolled out new Gemini capabilities across Docs, Sheets, Slides, and Drive for Indian users, bringing AI-powered productivity tools to one of the world's largest and fastest-growing digital markets. The expansion represents a significant push by Google to establish AI dominance in the Global South ahead of competitors.

India represents a critical battleground for AI adoption in emerging economies, with over 700 million internet users and rapidly expanding English and local-language digital consumption. By integrating Gemini across its productivity suite with local language support, Google is positioning to capture enterprise and consumer AI workflows in a market where Microsoft, OpenAI, and Chinese competitors are also vying for position. The move is part of a broader pattern of U.S. tech firms racing to secure AI market share in the Global South before local or Chinese alternatives can establish themselves.

Why it matters

Google's multilingual AI expansion into India signals intensifying competition for Global South AI adoption, where first-mover advantage in productivity tools and consumer applications could lock in hundreds of millions of users to specific AI ecosystems, shaping which countries and companies benefit from AI-driven economic growth in emerging markets.

What to watch

How quickly competitors, particularly Microsoft and local Indian AI startups, respond with their own multilingual AI offerings, and whether India's government moves to require data localisation or sovereign AI alternatives, which could force U.S. firms to build India-specific AI infrastructure or cede market share to domestic players.

Signals & Trends

Datacenter infrastructure is becoming a contested military asset in interstate conflict

Iran's bombing of Gulf datacenters marks the first instance of AI and cloud computing infrastructure being deliberately targeted as a military objective in interstate warfare. This establishes a precedent that AI computing capacity — previously viewed as civilian digital infrastructure — is now fair game in conflicts between nations. The implications extend beyond the immediate Iran war: any nation hosting significant AI datacenters, whether for domestic use or as part of hyperscaler networks, must now account for these facilities as potential military targets. This will likely drive several strategic responses: hardening of critical AI infrastructure with air defense systems, geographic distribution of AI workloads to reduce single points of failure, development of rapidly deployable or mobile AI computing to avoid fixed targetable locations, and potentially reconsideration of where nations choose to locate AI infrastructure relative to potential adversaries. The vulnerability is particularly acute for countries that have positioned themselves as regional AI hubs through partnerships with U.S. hyperscalers, as these facilities represent both strategic assets and symbolic targets for adversaries seeking to impose costs on U.S. alliances.

Sovereign AI controls are shifting from import restrictions to use restrictions

China's move to restrict state use of OpenClaw AI represents an evolution in how nations enforce AI sovereignty — not by blocking imports or access entirely, but by restricting which actors within the country can use foreign AI systems. This is a more granular approach than blanket bans: it allows civilian and commercial experimentation with foreign AI while creating a hard boundary around state systems that handle sensitive data or critical functions. The policy signals that Beijing views agentic AI as categorically different from previous AI tools due to its autonomous capability to navigate systems and execute complex operations without human oversight. Other nations will likely adopt similar tiered approaches, particularly for agentic AI: allowing general consumer and commercial use while restricting deployment in government, critical infrastructure, defense, and sensitive industries. This creates a more complex compliance landscape for AI companies than simple export controls, as they must navigate varying use restrictions across different customer segments in each market. It also suggests that the next wave of AI sovereignty enforcement will focus less on technology transfer and more on operational control over how AI agents interact with national systems and data.

Nvidia's infrastructure partnerships are becoming as strategically significant as its chip sales

The Thinking Machines deal reveals that Nvidia's role in shaping AI geopolitics extends beyond its dominance in chip manufacturing to include strategic allocation of scarce computing infrastructure through multi-year partnerships. By committing gigawatt-scale compute and making direct investments in specific AI startups, Nvidia effectively decides which companies can compete at frontier scale regardless of their capital resources or talent. This gives Nvidia outsized influence over the AI competitive landscape: a startup with Nvidia backing can leapfrog rivals with more funding but less access to infrastructure, while companies or countries on the wrong side of Nvidia's allocation decisions face structural disadvantages in developing cutting-edge AI. The strategic implications are significant for national AI competitiveness — nations seeking to develop frontier AI capabilities must secure not just Nvidia chips but also infrastructure partnerships, making Nvidia a critical dependency in AI sovereignty strategies. This also creates potential leverage points for U.S. policy: if Nvidia's infrastructure allocation decisions align with U.S. strategic interests, the company becomes a de facto tool of AI statecraft, but if Nvidia pursues purely commercial partnerships, it could enable AI capabilities in nations the U.S. would prefer to constrain. Competitors and governments should watch whether Nvidia develops explicit frameworks for infrastructure allocation or whether decisions remain opaque and commercially driven.

Explore Other Categories

Read detailed analysis in other strategic domains

Capital & Industrial Strategy

Amazon raised $37 billion in US bonds and is planning a euro offering to reach nearly $50 billion total, part of a record $66 billion single-day corporate debt issuance. The move signals how hyperscalers are financing AI data centres amid rising capital costs and tighter credit windows.

Compute & Infrastructure

Amazon raised $50 billion in a record-breaking bond sale explicitly for AI infrastructure, part of $65 billion issued in a single day across corporate debt markets. Separately, Princeton Digital Group plans $5 billion in datacenter debt for Asian expansion, signaling that existing compute capacity falls dramatically short of projected AI demands and traditional capital markets are being reshaped to close the gap.

Frontier Capability Developments

Mira Murati's Thinking Machines Lab secured a multibillion-dollar compute deal with Nvidia just one year after founding, covering at least one gigawatt of capacity—equivalent to 1.5 million H100 GPUs. The partnership includes a strategic investment, suggesting Nvidia is hedging against the OpenAI-Microsoft-Google oligopoly by backing alternative architectures.

Public Policy & Governance

The White House is preparing an executive order against Anthropic while the AI startup's lawsuit challenging its Pentagon supply chain designation advances in federal court. The case tests whether government can compel AI companies to compromise safety protocols for surveillance purposes, with Microsoft arguing such coercion violates the First Amendment.

Safety & Standards

Anthropic is in federal court fighting a DoD supply-chain designation triggered by its refusal to modify AI systems for domestic spying. The company claims billions in potential revenue loss, while Microsoft backs its First Amendment challenge. The case tests whether government can penalize AI labs for honoring their own safety commitments.