UK Puts "AI Sovereignty" on the Line: Led by Microsoft and NVIDIA, US Companies Invest Over £31 Billion in the UK
On September 17, 2025, Donald Trump arrived in London to commence his second state visit to the UK during his tenure.
This time, the focus is on the implementation of technology, not a political show.
Among the entourage, apart from high - ranking diplomats, there are also the CEOs of three of Silicon Valley's most core AI companies: Satya Nadella of Microsoft, Jensen Huang of NVIDIA, and Sam Altman of OpenAI.
They are not here for a publicity stunt but to deploy AI infrastructure on the ground.
Microsoft announced an investment of $30 billion (approximately £22 billion) to build AI data centers, cloud computing facilities, and local R & D teams.
NVIDIA announced that it will deploy 120,000 Blackwell GPUs in the UK and invest £500 million in the UK AI infrastructure company Nscale.
OpenAI launched the Stargate UK project, constructing a local large - model platform in North Tyneside, with an initial deployment of 8,000 GPUs.
Together with companies such as Google, CoreWeave, and Salesforce, the total investment exceeds £31 billion (approximately $42 billion), forming a comprehensive technology investment agreement spanning AI, energy, policy, and chips.
This is not simply corporate expansion but a national - level industrial layout.
The three major US AI giants are turning the concept of "sovereign AI" into reality. They are directly deploying their core capabilities into the infrastructure of other countries, and the UK has become a typical example of this model.
Problems have also arisen:
Is the UK building its own AI capabilities or becoming a node in the global layout of US companies?
This round of global AI competition has moved from the laboratories in Silicon Valley to the policy - making levels of various countries.
Section 1
Microsoft CEO Nadella: The $30 - billion investment is not for the market but to build the foundation of computing power
"We've been doing business in the UK since 1982."
When Microsoft CEO Satya Nadella was interviewed by the BBC, he began by mentioning Microsoft's 40 - year cooperation history in the UK. This is not just an introduction to the company's history but an explanation of why they are willing to invest $30 billion in the UK this time.
Our continuous investment in the UK over the decades is not because of short - term popularity but the stability of policies. This allows us to keep betting and lay the foundation we have today.
This is not just a polite remark.
The $30 billion announced by Microsoft this time is composed of three parts:
One part is invested in "hardware": land, data centers, construction equipment, and supercomputing networks;
One part is invested in "software": local sales, R & D, and operation teams;
The rest is for people: construction, recruitment, and training to form its own AI R & D team.
These deployments are not in London or Oxford but in the northeast of England.
Why this location? Because the UK government has designated this area as an "AI growth zone" and offers policy incentives in terms of approvals, energy, and land.
Microsoft chose to cooperate deeply with the UK AI infrastructure company Nscale and became the core customer of its data centers. Simply put, Microsoft is not just selling technology but becoming an important buyer and user of computing power infrastructure.
During the interview, the BBC reporter repeatedly asked him:
"The UK's economy is currently growing slowly. Do you think AI can really bring about a 10% increase in GDP? How long will it take to see the results?"
Nadella's answer was realistic:
"I hope it's not ten years but five, but it really takes time to see the results."
AI capabilities are indeed powerful, but for it to generate economic value, organizations have to go through a series of changes, redesign processes, and reshape collaboration methods... These stages cannot be skipped.
This statement actually emphasizes:
No matter how powerful the model is, if it's not deployed, it won't generate value;
If AI is not implemented, it's just an algorithm, not a capability;
Ultimately, it depends on who can integrate AI into business, connect hardware to applications, and transform technology into "tools that can be truly used."
So, Microsoft's $30 - billion investment is neither for product creation nor for market competition.
It's to lay the foundation for future AI factories.
AI competition is not in the cloud but on the ground.
Whoever builds the computing power infrastructure first will be able to embrace the next round of the intelligent revolution.
Section 2
NVIDIA CEO Jensen Huang: GPUs are a form of diplomatic language, and sovereign AI is a battle for deployment rights
"Don't forget, this is the hometown of Turing."
When NVIDIA CEO Jensen Huang was interviewed, he mentioned the UK's AI foundation: top - notch talents trained by universities, local research institutions, and an early computing tradition.
But what he said next was even more crucial:
"You have your own data, and this data belongs to the UK. You should use it to train your own large models."
This is not just encouragement but a hint: In the era of large models, it's not about who has the most servers but who can keep data and capabilities within their own country.
What NVIDIA is doing is turning this possibility into reality.
The company announced the deployment of up to 120,000 of its latest Blackwell GPU chips in the UK, which is NVIDIA's largest - ever implementation in Europe.
It's not just about selling chips but helping the UK build a complete data center;
It's not just about selling hardware but investing in land, infrastructure, and power supply together with partners;
It even directly invested £500 million in the local UK company Nscale and became an important partner in the Stargate UK project.
GPUs are no longer just technological components but a form of diplomatic language.
Where the chips are, the capabilities are; how much can be deployed means how much can be trained, optimized, and developed. Today's large models are no longer just cloud - based products but on - the - ground resources.
The reporter asked him:
"Do you think the UK really has a chance to be at the center of this round of AI? We here are always a bit hesitant, and people tend to underestimate themselves."
Jensen Huang was confused by this:
I really don't understand. Why be so modest? This is clearly the birthplace of artificial intelligence. Why not be bolder and develop your own AI?
What NVIDIA is bringing this time is not just 120,000 GPUs.
It's telling the UK government that they can have their own AI capabilities.
The prerequisite is that someone is willing to build the underlying computing power infrastructure on your land.
This is not "technology export" but "capability transfer." It's not API access but local computing power implementation.
NVIDIA's view is clear: The first step in sovereign AI is to build a computing power factory capable of training large models in your own country.
Section 3
OpenAI: Stargate UK is a pilot for the "sovereign model" and no longer uses a global - unified API
OpenAI launched the "Stargate UK" project to deploy large - model infrastructure locally in the UK:
Location: Initially, a large - model training platform will be built at Cobalt Park in North Tyneside, UK;
Division of labor: Led by OpenAI, constructed by Nscale, and NVIDIA will provide the first batch of 8,000 GPUs;
Policy: It is included in the UK's "AI growth zone" and will receive government support.
This time, OpenAI is not just announcing a cooperation but truly deploying large models into the UK's infrastructure.
The UK government stated that this project will support the development of sovereign AI, ensuring that the UK can run and train high - quality large models locally.
The key point in this statement is "local operation."
Most of OpenAI's previous models were deployed in the US, and users remotely accessed them via API; now, it is moving some of its training and inference capabilities to the UK.
That is to say:
Data does not leave the country, and local access is available;
The model runs in the UK without relying on overseas scheduling;
The platform is customized for UK developers and enterprises, no longer a standard global package.
Why do this?
The answer lies in the transformation of AI's role:
Large models are no longer just simple cloud services but core infrastructure that needs to be deeply integrated into policies, energy, and regulatory frameworks.
The previous model was global API access, and the current strategy is local construction and customized deployment to adapt to the policy environments of different countries.
The UK's acquisition of the Stargate UK project means that OpenAI is shifting from a "cloud service provider" to a "local deployment partner."
Stargate UK represents a new cooperation model: OpenAI provides technological capabilities, and the UK provides land and policies, with both parties deeply bound.
In this model, the key to AI sovereignty is no longer who develops the model but who controls its operation and updates.
The UK seems to have secured an advantageous position, but a question has emerged: When the technology platform is implemented in the UK, has the real control actually been transferred?
The answer may not be that simple.
Section 4
The UK has gained a lot, but the core control remains elsewhere
When the BBC reporter asked Nadella whether the UK was giving up its digital sovereignty,
He responded directly:
"We are investing locally in the UK. These are the UK's own AI factories. UK companies can use them to create their own technologies."
This statement continues the tone of the previous three sections: cooperation, openness, and win - win.
But behind this statement, a subtle change has occurred in the domestic discussion in the UK.
1. It's not about opposing investment but wanting to see more clearly: Who really holds the steering wheel?
UK Prime Minister Keir Starmer called this a "generational - level technology agreement" that will push the UK to the center of European AI. The Treasury issued a statement welcoming the three major technology giants for "bringing technology, capital, and employment."
But Cailin Birch, a global economist at The Economist Intelligence Unit, said something different on a BBC program:
"All the investments sound very attractive, but the benefits still mostly go to the US."
"We are providing land, electricity, and talent, but ultimately, the core capabilities remain in the hands of the other side."
This is not traditional "product export" nor local cooperation. Instead, US technology companies are becoming part of the UK's AI architecture.
2. What exactly have these cooperations brought?
The UK is definitely a beneficiary, at least in the short term:
The investment has arrived;
The number of jobs has increased;
The infrastructure construction has started;
The remote northern regions have been included in the global AI network nodes for the first time.
But the BBC host then analyzed:
"The models running in these data centers are not developed by the UK itself. Most of these jobs are in construction and maintenance, not in model training and instruction architecture. The final interfaces and training scheduling rights of these platforms still lie in Silicon Valley."
The UK has obtained "access rights" but not "control rights."
3. Is this cooperation, dependence, or a combination of both?
On the surface, the UK has obtained a key node in the global AI chain. In essence, it is being incorporated into the global deployment map of US companies.
Cailin Birch believes that this kind of cooperation is more like a structural integration rather than an equal - footed collaboration.
This represents a new type of technology cooperation model:
It's not about buying and selling models;
It's not about importing services;
It's about directly integrating computing power infrastructure into another country's basic system.
The key question is: What has the UK really gained?
The BBC analysis suggests that the UK could either be a beneficiary of AI prosperity or just a host for global AI deployment.
The current jobs and investments are not the key. The key is whether these capabilities will truly "belong to the UK" in ten years.
This section is not simply a question but a calm review.
OpenAI, Microsoft, and NVIDIA have indeed brought resources, platforms, and speed.
But this also raises a question that all countries need to face:
Is this a real cooperation and co - construction, or being incorporated into the global strategic layout of certain countries?
Conclusion
Who can embrace this AI industry
This investment wave is not just simple business expansion but a clear signal:
"The focus of AI competition has shifted from model performance to deployment capabilities."
The UK's gains from this round of bets are: the simultaneous implementation of three major technology giants, the first physical template of "sovereign AI" in the UK, and the northeast becoming an emerging AI industrial belt.
On the surface, it's a win - win cooperation, but in essence, it shows a new division of labor in the AI world: The US exports technical standards and core capabilities, and the UK provides land, electricity, and policy support to jointly build an "localized yet globally - connected" AI node.
For China, it's not about evaluating who has the upper hand in this cooperation but seeing the following points clearly:
True AI strength lies in local deployment capabilities, not model benchmark scores.
The key is a complete closed - loop of computing power, talent, and applications.
In the future, the competition will be about who can build a complete AI industrial system.
The core question in this cooperation model is: Are the participants co - builders or hosts?
National competition in the AI era tests this kind of hosting and transformation ability.
This is what needs to be answered when AI truly becomes a national power.
Reference materials:
https://www.youtube.com/watch?v=vouC1_M1O48
https://www.youtube.com/watch?v=vuNL6iXU2TY&t=4s
https://www.youtube.com/watch?v=S0duds0WbeU&t=56s
https://www.reuters.com/world/uk/uk-us-agree-42-billion-tech-pact-mark-trumps-visit-2025-09-16/?utm_source=chatgpt.com
https://www.bloomberg.com/news/articles/2025-09-16/microsoft-openai-herald-trump-s-uk-visit-with-spending-pledges?utm_source=chatgpt.com
https://openai.com/index/introducing-stargate-uk/?utm_source=chatgpt.com
Source: Official media/Online news
This article is from the WeChat public account "AI Deep Researcher," written by AI Deep Researcher and published by 36Kr with authorization.