The most revealing part of Google's $17bn commitment to a data centre in south-eastern Wyoming may not be the money. It is what happened before the investment was announced. An ambitious project, conceived by an independent developer with plans to build one of the world's largest AI campuses, has passed into the hands of the technology giant that has the financial resources and computing demand to take it forward. In an industry racing to build at unprecedented speed, the episode exposes an uncomfortable truth, having the ambition to build the infrastructure of artificial intelligence is not the same as having the capacity to deliver it.

Google acquired the unfinished project from Denver-based Crusoe on 6 November, a day after announcing its planned $17bn investment in the Wyoming campus. The original development, launched in partnership with energy infrastructure company Tallgrass in July 2025, was designed initially around 1.8 gigawatts of capacity, with the potential to expand to 10 gigawatts. The value of the deal has not been made public. Reports have pointed to concerns over construction costs and delivery timelines, although Crusoe has described its role as completing the project's pre-development work, including site selection, approvals and power strategy. The distinction matters, the transaction does not, by itself, establish that Crusoe was financially distressed or incapable of building the site. It does, however, illustrate the difficulty of turning a vast infrastructure proposal into an operational asset.

The implications reach beyond the American technology sector. Britain's NHS is becoming increasingly dependent on the same digital infrastructure that Google is spending billions to expand. Artificial intelligence, cloud-hosted systems, electronic patient records, diagnostic tools and data-driven research all require reliable computing capacity. Yet the debate about NHS digital transformation still tends to concentrate on software, procurement and clinical adoption. The physical foundations of that transformation receive far less attention.

That is a strategic oversight. An AI system used to identify abnormalities in medical images, summarise clinical consultations or help clinicians assess risk does not operate in isolation. It depends on servers, electricity, cooling systems, telecommunications, secure networks and specialist hardware. Its performance also depends on the quality, availability and governance of the information it processes. If any part of that chain is unreliable, the promise of faster diagnosis and more efficient care becomes considerably harder to realise.

The NHS has little room for another digital programme that promises productivity but creates additional operational complications. Hospitals are under pressure to improve access, reduce waiting times and manage constrained budgets while dealing with workforce shortages and growing demand. Technology is increasingly seen as a way to address these pressures. Ambient voice tools can reduce the time clinicians spend documenting consultations, AI-supported imaging could help prioritise urgent cases, and better-connected patient records could reduce duplication and improve coordination between services. But these gains depend on systems that work consistently, integrate with existing infrastructure and can be supported when something goes wrong.

The Wyoming deal offers a useful warning about the gap between technological ambition and delivery capacity. Crusoe's original project was developed around an enormous power requirement. Its subsequent acquisition by Google demonstrates the strategic importance of capital, energy partnerships and the ability to manage construction at scale. These are not secondary considerations. They determine whether computing capacity becomes available when customers need it.

Britain is confronting a version of the same problem. In October, the Government published its Data Centre Covenant, recognising that the country cannot rely almost entirely on overseas infrastructure as artificial intelligence becomes more important to essential services, including the NHS. The Government estimates that Britain will need 6 gigawatts of AI data centre capacity by 2030 to protect its national interests and secure access to computing resources. It has also acknowledged that grid connection delays are a major obstacle to expansion. The scale of the challenge is apparent in the electricity connection queue, official figures have identified approximately 140 data centre projects seeking transmission connections, representing around 50 gigawatts of capacity.

The Government has begun reforming the system, with plans to prioritise strategically important developments and align data centre connections with regional infrastructure capacity. But the policy challenge is not simply to build more facilities. It is to decide which developments merit priority, how their electricity requirements will be met, and how the costs will be distributed between operators, businesses and the public. Data centres may support economic growth and public services, but their demands on electricity networks, land and water cannot be treated as inconsequential.

For NHS leaders, this raises questions that extend well beyond the technology department. The health service needs clarity about where its critical data is stored, who controls the infrastructure supporting its digital services, and what happens if a provider suffers a cyberattack, power disruption or commercial failure. Dependence on a small number of global cloud and technology companies can bring scale and technical sophistication, but it also creates questions of concentration, resilience and bargaining power. A hospital cannot simply suspend its digital operations because a supplier is experiencing difficulties.

The issue is particularly important as the NHS expands its use of AI. The Government's ambition to make Britain a leading AI economy and the health service's desire to adopt new technologies are mutually reinforcing. The NHS offers substantial opportunities for research, clinical validation and the development of health technology. Its patient data, subject to appropriate safeguards, can support advances in diagnostics, drug discovery and personalised medicine. But these opportunities will be diminished if the country lacks secure, dependable access to the computing infrastructure needed to process that information.

There is also a workforce dimension. AI is often discussed as a means of easing pressure on overstretched clinical teams. Yet introducing new systems creates work of its own. Staff need training, workflows must be redesigned, information governance requirements must be met and technical support must be available. A poorly implemented system can add administrative burdens rather than remove them. The lesson for NHS management is that digital investment should be judged by its effect on clinical capacity and patient experience, not simply by the number of tools purchased or the speed of deployment.

Crusoe's position makes the Wyoming transaction more interesting than a straightforward corporate retreat. The company recently raised $3.9bn in a funding round that valued it at $30.9bn. Its business now spans data centre development, GPU cloud services and managed AI computing. Founded by Chase Lochmiller and Cully Cavness in 2018, it began by converting otherwise wasted natural gas into electricity for cryptocurrency mining before shifting towards AI infrastructure. Its latest board appointments include Redwood Materials founder JB Straubel. The company has also continued developing other major projects, including a separate Google-linked campus in Texas.

This suggests that the sale should not be interpreted as evidence that independent developers have no place in the AI economy. Rather, the market is separating different forms of expertise. Smaller specialists can identify sites, secure permissions, develop energy strategies and build technical infrastructure. Global technology companies can bring enormous capital commitments and guaranteed demand. The most effective arrangements may combine the two, although the balance of commercial power will not always be equal.

Britain's emerging data centre policy needs to account for this reality. Encouraging private investment is necessary, but public authorities must retain the ability to assess whether proposed developments deliver genuine national value. Faster planning and grid connections should be accompanied by credible commitments on energy efficiency, water consumption, cyber security, local employment and community benefit. The NHS should also have a stronger voice in decisions about the infrastructure on which its future digital services will depend.

The danger is that Britain treats AI as a software revolution while overlooking the industrial and energy systems beneath it. The NHS cannot modernise through algorithms alone. Its ability to use artificial intelligence safely and productively will depend on reliable power, secure computing, capable suppliers, skilled staff and institutions able to manage complex technological change.

Google's Wyoming investment is not an NHS story, and there is no reason to pretend otherwise. But it exposes a question that British policymakers can no longer leave to the technology industry: who will build, power and control the infrastructure on which the country's essential services increasingly depend? For a health service preparing to make AI part of everyday care, that is not a distant question about Silicon Valley. It is a matter of national resilience and the practical ability to deliver better healthcare.