Amazon's cloud arm is buying billions in Nvidia AI equipment, and some sits in warehouses because data centers lack the power to run it, Business Insider reports.
Amazon Web Services (AWS), Amazon's cloud business, is buying billions of dollars of Nvidia AI equipment faster than some of its data centers can switch it on. Business Insider reports that some of the finished gear is waiting in warehouses. The main problem is electricity.
Under a new process called Early Rack Receipt, AWS expected to receive roughly $5 billion in Nvidia GB300 server racks at US facilities before final deployment. The figure comes from August internal emails reviewed by Business Insider. A rack is a cabinet of computer servers. One email projected roughly $3 billion more in racks each month "for the foreseeable future."
Some AWS sites had room for more racks but not enough available power to run them. Finished equipment was staged at intermediary warehouses until the facilities were ready. One email said at least one AWS team lacked "a clear forecasted path" for getting incoming racks into data centers at the time.
The industry has a name for this: dark GPUs. A GPU is the kind of chip that powers AI. Dark GPUs are chips and servers that are bought but sit unused because of power shortages or other issues. The term can also cover design flaws that stop gear from being fully used once it is plugged in.
The cost is cash. Buyers must pay upfront to secure scarce Nvidia parts. If a data center is not ready, that money is tied up and earns nothing. Borrowing can fill the gap, but the report says lenders have become less willing to back some of these projects.
An Amazon spokesperson said the company routinely stages finished hardware near deployment sites so it can install it once capacity opens up. The spokesperson called that "best business practice." Amazon also disputed that the broader company lacked a path for getting racks into data centers.
CEO Andy Jassy has said Amazon has a "clear line of sight" to strong returns from AI and called the technology a "once in a lifetime opportunity."
Amazon expects roughly $220 billion in capital spending this year, more than any Big Tech peer. Its free cash flow, the cash left after spending on the business, turned negative last quarter for the first time since 2023. It could stay negative for a while.
The new process also changes timing. Before, AWS took ownership of a finished rack only when it arrived at a data center. Now it can pay for and record certain racks earlier in the supply chain. The trade-off is getting scarce chips before rivals do, at the risk of owning idle equipment.
The report does not say prices or services will change for AWS customers. But if costly hardware sits idle, the pressure could eventually show up in the cloud and AI tools that businesses pay for. Three things are worth following:
Buying AI chips is only half the job. Without enough power to run them, even Amazon can end up with expensive hardware that earns nothing.
OpenAI has about $11 billion in hand at a $1.25 trillion value, below its first ask. Microsoft and NVIDIA shareholders have reason to watch what comes next.
Read more: OpenAI raises $11 billion, but at a lower price than it wantedVisa tested 22 AI models on realistic shopping tasks. They passed just over half, and often bought the wrong thing anyway.
Read more: Visa tested 22 AI shoppers, and most still make costly mistakesDARPA wants AI to hunt for design flaws in stock, bond, energy and cloud markets, so enemies cannot use them to cause economic damage that reaches households.
Read more: Pentagon wants AI to find weak spots in stock and energy markets
