Alphabet Might Add 9 Gigawatts of Compute Capacity by 2028, Morgan Stanley Estimates

Alphabet’s buildout of data center and power capacity could accelerate further by the end of the decade, according to an estimate cited by Yahoo Finance from Morgan Stanley.

The note, as summarized in the report, projects that Alphabet could add about 9 gigawatts of compute capacity in 2028. In the same assessment, Morgan Stanley also said Alphabet might sell about 4 gigawatts of ten-year compute capacity, a figure described as tied to “ten” and the broader ability to monetize capacity rather than simply consume it internally.

For readers, “compute capacity” refers to the amount of computing that large-scale data center infrastructure can support, typically constrained by power availability and cooling as much as by servers. One gigawatt equals one billion watts, a scale that underscores the magnitude of data center buildouts in the AI era.

The idea of selling compute capacity is part of a broader industry trend where hyperscalers and data center operators can partner with customers or intermediate providers for longer-term access to computing resources. In Alphabet’s case, the reported figures point to a strategy that could balance internal AI training and inference demand with external arrangements that bring revenue or offset costs.

Morgan Stanley’s framing, as reproduced in the Yahoo Finance writeup, does not provide additional particulars in the headline description about where the additional compute would come from, which facilities would be involved, or what exact customer structures sit behind the “sell” figure.

Alphabet did not provide details in the Yahoo Finance item itself about the timing of specific data center projects, the expected capital expenditure, or whether any of the capacity would be supplied through owned infrastructure versus contracted power and hosting arrangements. Those elements are typically the key drivers of how such capacity targets translate into near-term financial impact.

Industry context matters because compute expansion increasingly depends on local power availability, grid interconnection timelines, and permitting. Even when server demand is clear, the bottleneck can shift toward electricity and cooling capacity, making multi-year compute targets more sensitive to execution than to forecasts alone.

What to watch next is whether Alphabet, in future investor communications or disclosures, quantifies the pace of capacity additions, provides updates on energy-related constraints, or clarifies the economics of longer-term compute arrangements. The reported estimates are a starting point, but the most actionable announcement will come from company-provided guidance and any disclosed partner or customer contracts.