Does a Server Create a Presence?
A rack of servers a business owns and a block of compute capacity it rents from a cloud provider look similar from the outside. Both put data and processing power inside the United States. They are analysed differently under the rules that determine whether a foreign business has a U.S. tax presence, and the difference is control, not hardware.
The distinction the analysis actually turns on
Equipment can be a fixed place of business for the person who owns and operates it. It generally cannot be one for a customer who merely buys a service measured by usage from someone else who owns and operates it. A foreign business that purchases a subscription to a cloud platform, where the provider owns the physical servers, decides where they are located, controls who has access to them, and can move a customer's workload between data centres without notice or consent, has bought a service. It has not acquired a place of business inside the United States in any sense the permanent establishment concept has historically required, because it has no right to use, occupy, or direct any particular piece of equipment.
A foreign business that owns servers, colocates them in a U.S. facility under its own name, and retains the right to access, configure, and control that specific equipment is in a different position. The hardware belongs to the business, sits at a fixed location, and is used to carry on the business's own operations, whether that means running a web application, processing transactions, or hosting a database that customers query directly. That fact pattern sits much closer to the traditional understanding of a fixed place of business than a cloud subscription does, because the business has something at its disposal in the sense the concept requires, even though no person of the business is physically present at the facility.
What the treaty commentary says about servers specifically
The OECD Model Tax Convention Commentary on Article 5 addresses this fact pattern directly, and most U.S. treaty interpretation draws on it even where the specific treaty in question follows the U.S. Model rather than the OECD Model. The Commentary distinguishes a website, which is a combination of software and electronic data with no physical location and therefore cannot itself constitute a permanent establishment, from the server on which that website is stored and operated, which is tangible equipment with a physical location and can in principle constitute a fixed place of business.
The Commentary then draws the ownership and control line described above. Where an enterprise operates its own server, at a fixed location, for a sufficient period of time, and carries on business through it, the server can constitute a fixed place of business even without personnel present at the location, because the Commentary does not treat human presence as a precondition for a fixed place of business generally. Where the enterprise instead purchases web hosting arrangements from an internet service provider, using server space that belongs to and is operated by that provider, the enterprise ordinarily has no server at its disposal, and the hosting arrangement alone does not create a permanent establishment for the enterprise being hosted.
Whether the equipment's function amounts to a genuine place through which business is carried on, rather than merely supporting it, still has to be assessed separately. The Commentary's general position that certain preparatory or auxiliary functions do not create a permanent establishment applies to a server exactly as it applies to any other fixed facility. A server used only to advertise the enterprise's goods, display a catalogue, or relay data without processing it in any substantive way looks more like an auxiliary function. A server that itself performs the core revenue-generating activity, such as executing the transactions that constitute the enterprise's business, or running the software that is the product being sold, looks more like the place through which business is actually carried on.
Why website hosting alone is generally not enough
The most common fact pattern for a small inbound e-commerce or software business, a website hosted on a third-party platform such as a major cloud provider's managed service, a website builder, or a specialised e-commerce host, generally does not create a U.S. presence under either the domestic trade-or-business test or a treaty permanent establishment analysis, for the same reason a cloud compute subscription does not. The business has bought a service. It does not own, lease in its own name, or exercise operational control over any specific physical server, and the provider can move the underlying infrastructure without the business's knowledge, let alone its direction.
This holds even where the hosting provider's data centre happens to be located in the United States and even where U.S. customers make up the entire customer base, because the location of the server is the provider's business decision, not a fact about where the foreign business itself carries on operations. A business incorporated abroad, run by people abroad, selling to U.S. customers through a website hosted on infrastructure it does not own or control, is not, by that fact alone, engaged in business at the location of the servers.
Where personnel presence changes the analysis
The equipment question and the personnel question are separate, and either one can create exposure independently of the other. A foreign business with no owned or leased equipment in the United States, but with an employee or dependent agent physically present here performing substantive functions, can still be engaged in a U.S. trade or business under IRC §864(b) and, depending on the treaty, can still have a permanent establishment through that person's activity under the dependent agent provisions of the treaty's Article 5, regardless of whether any server is involved at all.
Conversely, a foreign business that owns and operates servers in the United States, but has no personnel here and limits the server's function to genuinely auxiliary activity such as data backup or content delivery for a business whose substantive operations happen elsewhere, has a materially different case than a business running its actual transaction processing on U.S.-based owned hardware. The two variables, who owns and controls the equipment, and who is physically present and what they do, move independently, and a full analysis has to examine both rather than treating either one as decisive on its own.
| Fact pattern | Who controls the infrastructure | General direction of the analysis |
|---|---|---|
| Website on a shared hosting plan or managed cloud platform | The hosting provider | Ordinarily no fixed place of business for the customer |
| Reserved cloud compute capacity billed by usage, provider-managed hardware | The cloud provider | Ordinarily no fixed place of business for the customer, even at meaningful scale |
| Owned servers colocated in a U.S. data centre under the business's own control | The business itself | Closer to a fixed place of business; function performed by the server then matters |
| Any of the above, combined with a U.S.-based employee or agent with contract authority | Depends on the equipment, but personnel presence is analysed separately | Personnel presence can independently create a U.S. trade or business or a dependent agent permanent establishment |
This is genuinely contested ground
Whether and when a server can constitute a permanent establishment has been debated among tax administrations and practitioners since the OECD Commentary first addressed it, and the debate has not fully settled even as digital business models have made the question more common rather than less. Some tax authorities have taken more expansive positions than the Commentary's general framework would suggest, particularly where a foreign business's revenue-generating function is substantially automated and runs on infrastructure with a fixed physical location, on the theory that automation of a function does not change whether that function constitutes carrying on business through a fixed place. Others hold closely to the ownership-and-control line described above and decline to find a permanent establishment absent either owned equipment or personnel presence.
There is also an active and unresolved international conversation about whether the permanent establishment concept, built for a period when carrying on business required either people or owned physical assets in a location, fits a digital economy well at all, which is part of what has driven proposals for entirely separate taxing frameworks for digital businesses in various jurisdictions outside the United States. None of that broader debate changes the analysis a specific foreign business needs today, but it is a reason to treat any confident, universal answer to "does my server create a U.S. presence" with some skepticism, in either direction.
The domestic question runs alongside the treaty one
Everything above describes the treaty permanent establishment analysis, which only matters to a business resident in a treaty country that claims the treaty's benefits. A business with no treaty available, or one that has not established residence and eligibility under the treaty's limitation on benefits article, is tested instead under the domestic trade-or-business standard of IRC §864(b), and that standard is not identical to the treaty one.
Treas. Reg. §1.864-2 asks whether U.S. activity is considerable, continuous, and regular, based on the facts and circumstances, without the treaty concept's specific focus on a fixed place at the enterprise's disposal. Owned equipment operating continuously inside the United States, performing a function central to the business, such as processing the transactions that generate the business's revenue, is the kind of ongoing, substantial activity the domestic standard was built to capture, and it can support a finding of a U.S. trade or business even in a case where a treaty's more equipment-and-disposal-focused Article 5 language might have produced a different answer. This is the same structural point that runs through every fact pattern in this area: the domestic gate is generally broader than the treaty gate, a business can be inside one and outside the other, and a business without treaty protection does not get the benefit of the narrower analysis at all.
What to document
The facts that matter are ownership or lease of specific equipment in the business's own name, the degree of control the business exercises over that equipment's location and configuration, what function the equipment actually performs in the business's revenue model, and whether any person acts for the business inside the United States independently of the equipment question. A services agreement or terms of service from a cloud or hosting provider, read for what rights it actually grants the customer over specific hardware rather than over a described level of service, is usually the starting document for working out which side of the ownership-and-control line a given arrangement falls on.
This is general information about how the permanent establishment analysis has historically treated server and cloud infrastructure, as of the date written. It is not advice on a particular business's infrastructure, and given how unsettled this specific area remains among tax authorities, a position taken here should be built with a full review of the actual technical and contractual facts rather than assumed from a general description of the categories.