Over the past 12 months, a sweeping alignment between federal agencies and private AI firms has redrawn the boundaries of public–private collaboration.
With contracts now topping $800 million, government-funded research partnerships expanding rapidly, and major workforce initiatives underway, the message is clear: Washington is betting big on Big Tech to secure its future in the AI age.
What began with voluntary safety pledges and research consortia has rapidly matured into a tightly woven infrastructure of commercial deployments, national security engagements, and workforce training initiatives. It’s a strategy driven by pragmatism and powered by urgency.
In January 2025, President Trump issued a new AI-focused executive order titled “Removing Barriers to American Leadership in Artificial Intelligence”. This formally repealed the previous Biden order: “Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence” which was passed in 2023 and directed agencies to review, suspend, revise, or rescind any actions taken under it that were inconsistent with his administration’s AI priorities.
The new order focused on deregulation, innovation and national AI leadership, while stripping away the previous emphases on safety, risk management and transparent oversight.
The National Institute of Standards and Technology (NIST) now hosts the U.S. AI Safety Institute Consortium (USAISI), a 200-member cross-sector alliance that reportedly includes big industry AI players. This group is setting benchmarks for model evaluation, red-teaming, and risk management, critical for shaping both U.S. policy and global AI norms.
Meanwhile, the National Artificial Intelligence Research Resource Pilot (NAIRR), led by the National Science Foundation, offers public researchers access to powerful models and computing infrastructure for several providers, although the names of those companies are not yet publicly detailed.
The mission is to “democratise” access to advanced AI tools and ensure the U.S. retains global leadership in AI-enabled science and innovation.
But perhaps the most significant pivot came in July 2025, when the U.S. Department of Defence announced $800 million in contracts across four commercial AI firms: OpenAI, Google (via DeepMind and Gemini), Anthropic, and Elon Musk’s xAI.
Each has been tasked with building “agentic” AI tools: systems capable of planning, reasoning, and acting semi-autonomously to be deployed across logistics, intelligence, battlefield coordination, and internal operations.
In a dramatic turn, xAI, whose Grok chatbot recently came under fire for biased and offensive outputs, was awarded a $200 million deal to tailor its models for the Pentagon. Dubbed “Grok for Government”, the suite includes customised LLMs for classified environments and may be rolled out to other agencies via the U.S. General Services Administration.
OpenAI, for its part, has launched OpenAI for Government, a bespoke offering designed for federal use. The contract includes model deployment in secure environments for use cases in cybersecurity, document processing, veteran services, and medical diagnostics.
This blurring of lines between commercial AI labs and federal agencies raises profound questions.
First, it shows that raw AI capability has become a national resource, akin to oil, nuclear power, or semiconductors. The U.S. is formalising access to its own “strategic compute reserve” by building relationships with the companies that control the most powerful models.
Second, it underscores the risk of dependency. A handful of private firms, many with opaque governance structures and profit motives, now have privileged access to public sector priorities. This has already sparked concern among watchdogs and policymakers, especially given AI’s role in surveillance, decision-making, and even automated warfare.
Experts, including members of the UK AI Council, have raised concerns that if governments fail to develop core AI infrastructure independently, they risk becoming reliant on a small number of commercial actors whose incentives may not align with the public good.
Not all partnerships are defence-related. In Virginia, Google has launched a programme to train 10,000 state residents in AI skills, offering free or low-cost certification through community colleges. It’s part of a broader effort to close the digital skills gap, but also an important political move, earning goodwill in a region home to major data centres and intelligence agencies.
Elsewhere, Amazon, IBM, and Nvidia are supporting federal reskilling initiatives across civil service agencies and universities. These efforts reinforce the idea that the U.S. AI strategy is as much about domestic preparedness as it is about frontier technology.
What does this mean for capacity planners, network builders, and infrastructure owners?
First, expect cloud capacity to shift. U.S. government contracts often come with strict compliance standards (e.g., FedRAMP, IL5 security), driving demand for sovereign cloud and hybrid edge-cloud solutions. Providers that can meet those specs, or co-locate near federal installations, may see long-term growth.
Second, compute availability will tighten. As more U.S. supercomputers are reserved for federally backed AI training, private firms and universities may face higher costs or longer lead times. This could ripple into European and Asia-Pacific markets as multinationals look abroad for spare capacity.
Third, AI infrastructure is now national infrastructure. The line between public procurement and private provisioning is disappearing. As the U.S. forges ahead, other nations, including the UK, may have to decide whether to follow suit with similar partnerships or invest more heavily in sovereign models and infrastructure.
The United States is no longer treating AI as a peripheral technology. With military, civilian, and research systems now relying on commercial platforms, we are witnessing the rise of an AI-industrial complex, one that mirrors the dynamics of aerospace or telecommunications in previous eras.
RELATED STORIES
