The Case Against Breaking Up the Cloud
There is a growing movement to regulate cloud providers by forcing them to separate their platform services from everything else they offer. The idea borrows from railroad regulation, where companies that owned the tracks were barred from also running freight services on them. Apply that logic to the cloud, and providers like AWS, Azure and Google Cloud would have to choose: either offer the underlying infrastructure or offer the services built on top of it, but not both.
It's a tidy theory. It's also based on a faulty analogy, and it would do more harm than good.
The Railroad Analogy Doesn't Hold
Railroads are fixed physical infrastructure. A track is laid, a schedule runs, and the system is fundamentally static. Cloud providers are the opposite. They are defined by the services they expose through APIs, and those services are what make the cloud useful in the first place.
Suggesting that cloud providers should be split into platform and non-platform halves assumes there is a clean dividing line between the two. There isn't. A cloud is a platform all the way down. What one person calls infrastructure is just another service to someone else. The compute, storage and networking primitives are themselves delivered as services, and the higher-level offerings are built on top of them in a continuous stack.
Ask a government committee to draw that line and you get arbitrary regulation that ignores how the cloud actually works. You also risk freezing innovation. The industry's evolution from virtual machines to containers to serverless was only possible because cloud providers could optimize the entire stack. Force them to pick one layer and the incentive to push the whole system forward disappears. We'd still be stacking VMs instead of building the higher-level abstractions that define modern cloud-native development.
Not a Natural Monopoly
Natural monopolies exist when high infrastructure costs make it practical to have only one supplier. Think dams, rail lines or water systems. The cloud doesn't fit that mold.
The global IT market is worth roughly $5 trillion, and cloud computing is only a slice of that. The cloud is not the only option for running workloads, and it isn't even the only option within the cloud market. There are dozens of providers, plus colocation, on-premises infrastructure, bare-metal hosting and hybrid setups. No one is forced into a cloud because of structural barriers; they choose it because the model is attractive.
What is true is that the market has crowded around a few large players, and switching between them is not easy. There is no standard cloud API, so moving from one provider to another means porting more than just application code. That is a real friction point, but it is a technical problem with technical solutions. Container-native platforms and open-source tools are already chipping away at it. Switching costs are not a justification for reregulating an entire industry.
Fixed Costs and the Innovation Problem
The economics of running a cloud are not like running a software company. Cloud providers pour massive fixed costs into datacenters, networking, cooling and hardware that must be maintained and replaced. The services they sell on top of that infrastructure are what make those capital investments viable.
Vertically separate the two, and you need third parties to build and operate managed services across every datacenter and region where the provider operates. That's a tall order even for well-funded companies. Very few organizations have the expertise to build reliable, globally distributed managed services without deep control over the underlying system.
There's also the question of incentives. If a cloud provider can't offer services that make its infrastructure more valuable, why keep pouring money into that infrastructure? With limited ability to monetize the platform through higher-level offerings, prices would climb and investment in the underlying systems would decline. We've seen this pattern play out in regulated utilities, where short-term rent-seeking trumps long-term innovation. There's no reason the cloud would be different.
Vertical Integration Produces Better Technology
Some of the most important advances in cloud computing came from coordination between hardware and software that only a vertically integrated provider could achieve.
A prime example is AWS Nitro, a combination of custom hardware and software designed to improve performance and security. Nitro required AWS to control and optimize every layer of the stack, from the physical chips to the virtualization layer to the services running on top. It's hard to imagine a platform-only provider inventing something like Nitro, because the payoff would not be captured by the provider itself.
Reliability is another area where vertical integration matters. When one company controls the entire stack, it can ensure that each layer works with the others. When responsibility is split between providers and third-party service vendors, outages become finger-pointing exercises. The cloud's reputation for high availability depends on tight, end-to-end control.
The cloud is not a railroad. It's a continuously evolving system engineered for efficiency. Any regulation that breaks up the stack will slow the pace of that engineering, reduce reliability and raise costs for everyone who builds on top of it.
