Cloud Backlog
The cloud backlog is the sum of all amounts that customers have contractually committed to a cloud provider but that have not yet been paid out, because the service will only be delivered in the future. Investors read the figure as a preview of how strongly revenue is likely to grow in the coming quarters.
Large data centers rent out computing power and storage space over the internet. This business is called cloud. Customers rarely book this for just one month; instead, they often sign contracts running three or five years. However, the money only flows in once the service is actually used. Everything a provider has already firmly secured but is not yet allowed to record as revenue is called cloud backlog. It is thus a kind of filled order book, from which the revenue of the coming years is worked off.
What the figure reveals about future revenue
A quarter’s revenue describes the past. It says little about what comes next. The backlog, by contrast, shows how much business is already under contract. That is why analysts covering Amazon, Microsoft, and Google now watch this figure almost as closely as profit.
The amounts are enormous. At the major providers, they run into the mid-hundreds of billions, in some cases higher than the entire group’s annual revenue. If the backlog grows faster than revenue, this is seen as a sign of sustained demand. If it shrinks even though revenue is still rising, that is a warning sign for the coming years.
Since the AI boom, this metric has taken on additional significance. Companies that operate large AI models rent computing power over very long periods. Such deals can cause the backlog to jump by double-digit billions in a single stroke. It is precisely announcements like these that move share prices on reporting days.
From contract signature to revenue recognition
Under accounting rules, a company may only show money as revenue once it has delivered the service. A contract worth 500 million euros over five years therefore does not appear immediately on the balance sheet. It moves into the backlog and is worked off from there quarter by quarter. In annual reports, the usual technical term for this is Remaining Performance Obligations, or RPO for short — that is, the performance obligations still outstanding.
You can picture this like the order book of a carpentry workshop. Twenty ordered kitchens represent work that is certain to come. But each kitchen is only paid for once it is installed. The order book reveals the workload of the coming months, not today’s bank balance.
The time breakdown is important. Companies disclose what share of the backlog will convert to revenue within the next twelve months. A backlog consisting almost entirely of very long-term contracts looks impressively large but does little to help the current year. Moreover, not every commitment is set in stone: customers can renegotiate contracts or fail to fully use up minimum purchase commitments.
Where the metric turns up in the news
The term is most commonly encountered in tech companies' quarterly reports. Phrases like “backlog of 400 billion dollars” or “RPO up 30 percent” come from these releases. The figure also regularly appears in analyst commentary on Oracle, Nvidia customers, or AI startups.
It should not be confused with the term backlog as used in software development. There, it refers to a list of open tasks in a project and has nothing to do with money. In the finance section of a news site, it almost always means the order volume.
A typical misconception is also to treat the backlog as money already earned. It is a promise, not a bank balance. Anyone reading reports on this therefore pays attention to two things: the growth compared to the previous year, and the share that is already due next year.
