Amazon Web Services is doubling down on its homegrown artificial intelligence chips.
According to a report from 富途牛牛 (Futu), AWS has raised its third-quarter shipment target for servers built around its in-house ASICs — application-specific integrated circuits designed for AI workloads — by as much as 30%.
ASICs are custom-built chips tailored to a specific job, in this case running AI computations. That focus can make them more efficient for their intended task than general-purpose processors. By designing its own, AWS aims to lean less on outside chip suppliers and to shape the hardware around its own cloud and AI needs.
The upward revision, as described by Futu, signals accelerating momentum behind AWS's internal silicon effort. A higher shipment target suggests the company expects stronger demand for AI-capable servers running its custom parts, and is willing to commit to more units to meet it.
Beyond that top-line figure, the source does not provide additional specifics such as exact unit numbers or naming the individual chip lines involved.
Why it matters: When one of the world's largest cloud providers accelerates a push into its own AI chips, it points to a broader industry shift — big buyers increasingly designing silicon in-house rather than relying solely on third-party suppliers, a move that could reshape who profits from the AI hardware boom.