Why the Next Data Centre Boom May Not Be About Data Centres at All
For the past fifteen years, technology leaders have been guided by a single strategic principle:
Move to the cloud.
Servers disappeared from office buildings. Data centres were outsourced. Infrastructure became someone else’s problem.
The economics made sense. The scale made sense. The convenience was compelling.
But artificial intelligence is quietly changing the conversation.
Not because the cloud is failing.
Because boards are beginning to ask a fundamentally different question.
Who owns the intelligence of our business?
That question is at the heart of what many now call Sovereign AI, and I believe it will become one of the defining infrastructure themes of the next decade.
Sovereign AI Is Not About Governments
When most people hear the phrase “Sovereign AI”, they think about governments, defence departments, national security agencies and critical infrastructure.
That is certainly part of the story.
Around the world, governments are investing billions in domestic compute capability, sovereign cloud services, AI factories, local model development and national AI infrastructure. Nations increasingly view AI capability as a matter of economic competitiveness and strategic autonomy.
But I believe the far bigger story is what happens when sovereign AI becomes a boardroom issue rather than a government issue.
Consider a company with revenues exceeding $100 million.
Whether it operates in construction, mining, healthcare, manufacturing, logistics, utilities or financial services, it possesses something incredibly valuable:
- Intellectual property
- Operational procedures
- Commercial intelligence
- Historical decisions
- Customer knowledge
- Organisational expertise
Increasingly, these assets exist in digital form.
And increasingly, AI systems are being connected directly to them.
At that point, the question is no longer simply:
“How do we use AI?”
The real question becomes:
“Who controls the intelligence that makes our business valuable?”
The Fundamental Challenge To Enterprise Leaders
Recent commentary from Palantir CEO Alex Karp has brought this issue into sharp focus.
Karp argues that many organisations have become obsessed with consuming AI through external providers while paying little attention to where the value created by AI ultimately accumulates. He has framed the issue as one of enterprise control, ownership of intellectual property, and protection of competitive advantage.
His argument runs to the heart of your firms future value!
If an organisation increasingly relies on external AI providers for its decision-making, workflows, operational intelligence and institutional knowledge, is it merely consuming a service, or is it gradually transferring part of its competitive advantage elsewhere?
Your business’s performance is founded on your competitive advantage (Porter) therefore your “alpha” is what makes you different, Karp argues that enterprises should retain maximum control over their data, workflows and AI architectures.
Therefore AI is no longer just software. It is THE means of production, aided and guided by your talented team!
The New Means of Production
During the Industrial Revolution, ownership of factories determined who captured value.
Today, ownership of:
- Data
- Compute
- Algorithms
- Models
- Operational workflows
may determine who captures value in the AI economy.
This is why Sovereign AI should not be viewed primarily as a cybersecurity issue. Nor is it simply a compliance issue.
It is fundamentally an ownership issue.
The organisations that control their intelligence, data, workflows and AI capabilities will retain greater strategic flexibility than those that become entirely dependent on external providers. After all, if your switch providers 3 years in, where does your ‘context harness’ live, how do you retain the 3 years of Ai optimised/Ai supported value created?
In that context, sovereignty is not necessarily about nationalism.
It is about optionality!
It is about maintaining control over your future!
Why This Creates A Physical Infrastructure Problem
Once boards begin thinking about AI through the lens of ownership, an interesting consequence emerges.
Control requires infrastructure.
Not every organisation will build its own data centre.
Many will not want to.
Most boards do not aspire to become operators of critical facilities.
However, many will increasingly seek environments that provide:
- Greater control
- Greater security
- Greater transparency
- Greater jurisdictional certainty
- Greater ownership of their data and AI assets
This is where the conversation shifts from software to physical infrastructure.
The likely outcome is not a wholesale return to traditional on-premise computing.
Instead, I believe we are entering the age of hybrid sovereign infrastructure.
Some workloads will remain in public cloud.
Many other workloads will move to sovereign cloud.
Some will operate from co-location facilities.
Some will run on dedicated infrastructure controlled directly by enterprises.
The common theme is not location.
The common theme is control.
Why Co-Location May Be The First Major Winner
There is a practical challenge facing organisations pursuing AI sovereignty.
Demand is accelerating far faster than infrastructure can be built.
New hyperscale data centres require:
- Land acquisition
- Development approvals
- Utility upgrades
- Substation connections
- Construction
- Commissioning
These activities typically take years.
Boards, however, are making decisions now.
This reality may create a significant advantage for co-location providers that already possess:
- Power
- Connectivity
- Security
- Operational capability
- Available capacity
Rather than building entirely new facilities, enterprises may initially choose to own equipment while placing it within specialist environments designed to provide sovereign control and operational resilience. Growing demand for AI-ready co-location environments is already being observed across the market.
The Forgotten Opportunity Hiding In Existing Buildings
The most interesting development may not be the construction of new data centres.
It may be the transformation of existing buildings.
AI infrastructure is fundamentally different from traditional enterprise computing.
Higher-density GPU environments require significantly greater power and cooling than previous generations of infrastructure. Industry commentary increasingly points to liquid cooling becoming a critical component of many AI deployments.
Conventional office towers were not designed for this future.
Yet many already possess valuable ingredients:
- Established power connections
- Mechanical services
- Fibre connectivity
- Physical security
- Metropolitan locations
Retrofitting existing facilities may offer a faster route to deployment than waiting for entirely new infrastructure developments. Various industry sources identify retrofit strategies as an increasingly attractive approach for accelerating AI capacity while avoiding some greenfield constraints.
As a result, building owners may begin to see value differently.
Instead of optimising every square metre for office occupation, some may begin optimising for compute occupation.
A dedicated AI floor may become more valuable than another level of conventional office tenancy.
The Coming Shift
My prediction is straightforward.
The next phase of AI adoption will not be defined by who has access to artificial intelligence.
Almost everyone will.
The organisations that win will be those that determine how much of their intelligence they choose to own.
That decision will drive demand for:
- Sovereign infrastructure
- Private AI environments
- Co-location facilities
- Dedicated compute assets
- AI-ready commercial buildings
The conversation will increasingly move beyond model performance and into questions of ownership, governance and economic control.
Sovereign AI is not ultimately a data centre story.
It is an ownership story.
The data centres, co-location facilities, cooling systems and electrical infrastructure are simply the physical consequence of a much larger shift.
A shift toward organisations recognising that intelligence itself has become a strategic asset.
And those who own the factory that produces intelligence may ultimately own the future.
