AI is quietly (for now) breaking the Hyperscaler model.

Following on the theme from my last missive, I believe the real prize in AI infrastructure is not datacentres, it is the orchestration layer. An orchestration layer built on a single foundational principle I have written about extensively – trust.

Which is why Europe has a genuine opportunity to build the trust fabric for Artificial Intelligence (AI) infrastructure and avoid the hyperscaler trap it may currently be sleepwalking into.

When I say Europe, I do not mean just the European Union (EU). The EU has real momentum in this space and will inevitably try to frame it as an EU initiative. But this is a continental opportunity, spanning the Nordics, the UK and the wider European technology ecosystem.

My perspective comes from running my own business, advocating in Brussels for Europe’s small, medium enterprises (SMEs) since the early days of Application Service Providers (ASP) long before ‘cloud’ became the dominant term partnering closely with Microsoft and latterly Amazon Web Service (AWS).

From that vantage point, Europe faces a strategic question … Should it try to replicate the hyperscale cloud giants of Silicon Valley or design a different model of digital infrastructure altogether?

Since it appeared on the scene, cloud computing has been dominated by a small number of vertically integrated platforms such as AWS, Microsoft and Google. These companies built vast datacentre networks, proprietary ecosystems and developer platforms that now underpin much of the global digital economy.

Europe understandably worries about dependence on these platforms but trying to build a direct European hyperscaler risks falling into a hyperscaler trap investing enormous capital in an attempt to recreate an ecosystem that took decades, vast venture funding and uniquely integrated markets to develop. The hyperscaler trap assumes the future belongs to ever-larger centralized platforms.

That is likely a losing game. AWS, Microsoft and Google are the AMG engines of modern tech but Europe does not need to build another AMG subsidiary. It needs to design the road and AI is quietly changing the rules of infrastructure in ways that could favour this approach.

Traditional cloud computing was built for stable enterprise workloads such as on demand compute. Storage,  databases, business applications and web services running continuously inside large centralised datacentres. AI workloads behave differently. Training large language models (LLMs) requires bursts of massive compute in specialised graphics processing unit (GPU) clusters. Inference, the process of running models in production, often needs to happen closer to users, data sources, factories, hospitals or financial systems. Power availability, cooling and regulatory constraints also shape where this compute can run. No sooner did the lightbulb come on for me than I read Nvidia doing a pivot on its GPUs to run them all mantra, to the release of a new generation of ‘inference’ chips. I sense I am on a hot trail…

This is yet another signal that AI infrastructure (and the use case demands from the market) is increasingly resembling a distributed network of compute resources, not a handful of mega-datacentres, calling for a new ecosystem approach.

The AI era is organically evolving infrastructure architecture away from the classic hyperscaler model because it changes the economics of compute. Traditional cloud grew around relatively general purpose, always on enterprise workloads. AI is different, it is far more bursty, hardware specific, data sensitive, power constrained and geographically uneven. That makes a more federated, pooled, policy driven compute fabric look increasingly rational.

Europe already has many of the ingredients for such an ecosystem with regional datacentres, telecom infrastructure, high-performance computing facilities, industrial AI demand and publicly funded research platforms at both national, sector and EU. Initiatives such as EuroHPC Joint Undertaking and GAIA-X are early attempts to organise these assets into federated infrastructure. However they come with a very modern health warning as such systems reperesent a new fronteer in Cyber threat as I wrote about recently in Sovereign AI and the Cyber Risk of the Well-Governed Target.

This points to a different strategic model. Europe’s opportunity is not ownership of hyperscale infrastructure. It is orchestration.

If then compute becomes more distributed, the real question shifts from ‘Who owns the biggest datacentres?‘ to ‘Who controls the orchestration layer (and can be trusted) that coordinates compute across many providers?’

That orchestration fabric could provide an independence trust layer and manage identity, policy enforcement, workload scheduling, cryptographic attestation, regulatory compliance and resource allocation across a federation of infrastructure providers.

In effect, it becomes a continental control plane for sovereign digital infrastructure. By building that layer, there is no need to compete to beat hyperscalers at their own game, simply change the rules of the game.

Crucially, the underlying infrastructure could remain owned by many organisations, datacentre operators, telecom companies, research institutes and sovereign cloud providers.

What unifies them is the trust fabric. A trust fabric I have infered before – Is The AI Boom just the Birth of New Infrastructure? – that could take on bespoke classifications to serve discrete industry use cases and Critical National Infrastructure, imagine that dynamism of infrastructure utilisation …

This federated model plays to Europe’s strengths. Its digital landscape is naturally distributed as aforementioned. Rather than treating that fragmentation as a weakness, Europe could turn it into a structural, global, advantage.

The lesson from Silicon Valley is often misunderstood and the success of the AMGs and others a distraction. Its power does not come solely from large companies but from the ecosystems around them they manifest, the startups, universities, open-source communities, venture capital and shared infrastructure. Having spent over a decade on Microsoft’s Partner Advisory Council, I have had the pleasure of seeing first-hand the traction the Microsoft Partner Network (MPN) ecosystem creates in the market, having worked with the MPN team through three generations of their partner programme.

The real danger is that Europe spends tens of billions expanding AI capacity but does so in a way that simply deepens dependency on foreign orchestration layers, foreign chips, foreign developer ecosystems and foreign control planes. The recent European Parliament study, ‘European Software and Cyber Dependencies,’ is blunt that Europe remains heavily dependent on non-EU software and cloud providers, which creates strategic vulnerabilities.

In summary, I feel Europe’s strategy in this context has a unique opportunity to focus less on building national champions and more on enabling shared platforms and ecosystems that allow innovation to flourish across its vast SME economy that represent an estimated 99.8% of all businesses in Europe generating roughly 58% of EU GDP / value added and accounting for about 67% of private-sector employment or in real numbers over 100 million people. Now that is the real ‘AMG’ engine of Europe.

Europe will invest billions in AI infrastructure over the coming decade, for that you can be certain. The real question is not whether to invest but how that infrastructure is designed. As I wrote, there is a new generation of AI ‘world’ modelling such as V-JEPA appearing that will accelerate this space – Will the Next Cyber War Be Fought Inside Machine Minds?

The last era of digital infrastructure was defined by hyperscalers. The next may be defined by whoever builds the trust fabric that connects the world’s compute. Europe may be far better positioned to win that contest than it realises to leapfrog the hyperscaler paradigm rather than chasing it. Not by building a European AMG but by creating a federated, trusted digital infrastructure orchestration layer that enables thousands of organisations to innovate on top of it, a win for everyone not the AMG few.