Contents
DFINITY believes the future of cloud computing lies in sovereign infrastructure, decentralised architecture, and AI-powered software creation rather than dependence on a handful of cloud providers.
As governments race to define digital sovereignty and enterprises reconsider their dependence on hyperscale cloud providers, most of the debate remains theoretical.
Dominique Williams, founder and Chief Scientist of DFINITY, proposes a way forward.
DFINITY has spent more than a decade building an alternative to the traditional cloud: a decentralised computing platform designed to let governments and enterprises retain control over where their applications run, who operates the infrastructure and how data is governed.
The concept moved beyond theory earlier this year. At the World Economic Forum in Davos, DFINITY launched the Swiss Subnet, a sovereign cloud deployment running entirely on 13 independent node providers located within Switzerland. Rather than relying on contractual assurances from a hyperscaler, the infrastructure is designed to provide cryptographically verifiable data sovereignty through an open protocol. The approach is already attracting international interest.
Pakistan is now building a national subnet based on the same architecture, becoming the first country to deploy a decentralised sovereign cloud platform at nation-state scale.
Building a different kind of cloud
DFINITY is attempting to build a decentralised alternative to traditional cloud infrastructure. Rather than using blockchains to record transactions, the Internet Computer is designed to run complete software applications —including backend logic, data storage, APIs and, in some cases, frontend assets — across a distributed network.
Williams says the idea for the Internet Computer came from viewing blockchain differently.
“If you treat a blockchain as a system that can host tokens, then in principle it can also host code.” Traditional web applications typically have three layers:
- Frontend: the website or mobile interface.
- Backend: application logic running on cloud servers.
- Database: where data is stored.
With DFINITY, much of the backend and data storage can run on-chain inside “canisters”—software containers that combine code and state. Instead of deploying backend services to AWS or Google Cloud, developers deploy them to the Internet Computer network, where they run across independently operated data centres.
The network’s native token, ICP, is converted into “cycles,” which pay for computation and storage. Developers can fund these cycles themselves, meaning end users don’t necessarily need to own cryptocurrency to use an application.
For Williams, sovereign cloud is not simply about where servers are located. It is about eliminating dependence on any single infrastructure provider while giving organisations the ability to move workloads, configure trusted infrastructure and maintain control over their software without vendor lock-in.
“We’re trying to create something that functions much more like the internet itself,” he said.
Cloud Engine: separating applications from infrastructure
One of DFINITY’s biggest recent developments is Cloud Engine, which allows organisations to create their own Internet Computer subnet while deciding exactly where it runs, who operates it and the security trade-offs they want to make.
“You can decide exactly which node providers you trust, where those nodes are located and what trade-offs you want to make,” Williams said.
“For example, you might choose providers located entirely within Europe to maximise resilience while remaining GDPR compliant.”
Companies can configure the geographic distribution, resilience and security characteristics of their infrastructure according to their own requirements. Applications inherit those properties automatically because they run on the configured subnet.
Unlike today’s Internet Computer deployments, Cloud Engine is not limited to dedicated hardware. Organisations will be able to run nodes across hyperscale cloud providers including Amazon Web Services, Google Cloud and Microsoft Azure, alongside traditional infrastructure or sovereign hardware. That flexibility also makes Cloud Engine far easier to adopt.
“Most organisations don’t need enormous amounts of compute power,” he said. “They could potentially run a Cloud Engine for only a few hundred dollars a month.”
Cloud Engine’s biggest architectural innovation is that it separates applications from the underlying infrastructure. Rather than tying workloads to a specific cloud provider, the infrastructure itself can change while applications continue running. “Cloud Engine is essentially just a protocol configuration,” Williams said.
“That means you can change the underlying infrastructure without interrupting the applications running on it.”
For example, an organisation could initially deploy its Cloud Engine across Amazon Web Services before later migrating entirely to Google Cloud without taking applications offline. New nodes are added, synchronised and brought into service before the original nodes are removed.
“It’s a bit like a spider walking from one branch to another,” Williams said.
“The infrastructure moves beneath the applications without them ever noticing.”
The same approach allows organisations to move from inexpensive public cloud infrastructure to dedicated sovereign hardware as their deployments grow, or to replace smaller compute nodes with more powerful ones without disrupting running services.
“Your applications inherit different security properties depending on how you configure that private subnet, or Cloud Engine. It gives organisations a level of control that hasn’t previously existed.”
The result is that sovereignty becomes something enforced by the technology itself rather than by agreements with cloud providers.
Why cloud sovereignty needs more than geography
However, Dominique Williams remains sceptical of the concept of sovereignty when using software produced by an American company; even if its data is hosted elsewhere, it’s still governed by US law under the US CLOUD Act.
“One argument I often hear is that contracts prevent Google from providing access to customer data. My response is that contracts can’t override legislation.
If a contractual clause conflicts with the CLOUD Act, then the law takes precedence. The clause may exist, but legally it doesn’t change the situation.”
Another argument is that the environment is “air-gapped.” But if government employees can access the system, then by definition it isn’t truly air-gapped.”
He argues the real question is how many people have legitimate access to a system, and what happens if even one endpoint is compromised.
“Whether that’s through malware, credential theft or another attack vector, you’ve potentially opened a path into the environment.
From my perspective, genuine sovereignty requires genuinely sovereign infrastructure and sovereign software.“
Pakistan: sovereign cloud at national scale
Cloud Engine allows countries to build genuinely sovereign cloud infrastructure based on open-source technology, while benefiting from mathematically verifiable security guarantees.
In February, the Pakistan Digital Authority (PDA) and the DFINITY Foundation signed an MoU to create a dedicated Pakistan Subnet on its Internet Computer Platform (ICP). It includes plans for a National Messenger application enabling private, verifiable communications; expanded access to Caffeine, an AI platform incubated by DFINITY; 1,500 licenses of Caffeine to create applications; and capacity-building initiatives across government, education, and entrepreneurship.
Williams said the opportunity in Pakistan arose because DFINITY’s Chief Development Officer, based in the Middle East, had relationships there.
But he believes countries such as Pakistan are also confronting different infrastructure challenges.
“Redundancy across multiple data centres is particularly valuable, especially in regions where infrastructure resilience is a major concern.
For years I’ve argued that data centres will increasingly become targets — whether through terrorism, cyberattacks or armed conflict. Unfortunately, recent events in the Middle East have reinforced that concern. Cloud Engine addresses that problem by allowing applications to continue running even if underlying infrastructure is disrupted.”
He predicts a future with deployments across the Middle East running on multiple cloud providers such as Amazon Web Services, where if one data centre is affected, the applications continue operating because the infrastructure is distributed.
When the user becomes the product manager
Pakistan’s agreement also includes expanded access to Caffeine, DFINITY’s AI application-building platform. However, Williams sees the technology as representing a much broader shift in how companies will create software.
Rather than helping developers write code faster, Caffeine allows users to describe an application in natural language while an entire team of AI agents designs, builds and deploys it.
“When people use tools like Caffeine, they aren’t interacting with a single AI model,” Williams explained.
“Behind the scenes there’s an entire team of specialised AI agents, much like a human software organisation.”
Those agents take on the roles of front-end and back-end engineers, architects, security reviewers, testers and UX specialists, working together to build the application. In this model, the user becomes the product manager rather than the programmer.
“When a non-technical person uses the platform, they’re effectively acting as the product manager. Instead of managing a human engineering team, they’re directing an automated one. That fundamentally changes software development.”
The end of the developer-led platform era?
According to Williams, the company has made a very deliberate strategic shift towards sovereign cloud infrastructure “because we believe the market itself is changing.”
“Historically, software platforms were chosen by developers. If someone spent ten years mastering technologies such as Node.js, Kubernetes, PostgreSQL and Amazon Web Services, they naturally wanted to continue using those tools.
They’d invested years building those skills, and employers were hiring for them. That’s how technology ecosystems develop — through network effects.”
However, he contends that increasingly, AI agents will be responsible for building software without factoring in which frameworks are in fashion and most familiar to developers.
“They simply evaluate which platform best solves the problem.”
As a result, there are instances where product managers and business leaders are becoming more effective at creating applications than traditional software engineers—not because they write better code, but because they understand the customer and can direct AI more effectively.
“That changes the evaluation criteria completely.”
Instead of asking, “Is this the framework developers already know?”, organisations start asking different questions:
- Can AI build effectively on this platform?
- How much does it cost?
- Is it secure?
- Is it resilient?
- Does it avoid vendor lock-in?
Those are very different priorities, and we believe they’ll increasingly shape the next generation of cloud infrastructure,” shared Williams.
As AI changes how software is built, the criteria for choosing cloud infrastructure may change with it. If developers are no longer the primary decision-makers, portability, resilience and digital sovereignty could become as important as the programming frameworks and cloud ecosystems that have dominated the industry for the past decade.