Editor’s note: Electronics recycling will be featured in sessions at the 2026 E-Scrap: The Longevity Conference in New Orleans October 26-28.
Alphabet spent $44.9 billion in a single quarter building AI infrastructure, and raised its full-year guidance to $195–$205 billion. For anyone in decommissioning or ITAD, that number should be read as a forward order book for the hardware this sector will eventually be asked to process. Revenue rose 24% to $119.8 billion, Cloud revenue grew 82% to $24.8 billion, and Cloud backlog hit $514 billion. But the capex figure is the one that maps directly onto future asset volume. CEO Sundar Pichai and CFO Anat Ashkenazi both said demand for AI infrastructure continues to outpace supply, which is another way of saying the buildout has not yet peaked.
Set aside the enthusiasm on the call: Pichai’s “exciting momentum,” the 17% growth in Search and Other, the 13% gain in YouTube ads. The operative detail is Ashkenazi’s breakdown of where the $44.9 billion actually went, roughly 60% to servers and 40% to data centers and networking equipment. That split matters more to an ITAD operator than the top-line growth numbers.
It is a direct signal of the physical asset mix now flowing into hyperscale fleets, and by extension, the asset mix that will need refresh, redeployment, or recycling on the other side of its service life. Free cash flow went negative for the quarter as a result of this spending pace, and Ashkenazi said the raised guidance reflects an accelerated delivery timeline, with capex expected to rise again, significantly, in 2027.
The demand figures Pichai cited (over 9 million developers building monthly on Alphabet’s models, roughly 22 billion tokens processed per minute across its APIs) are the demand-side justification for that spending. He framed the company’s ongoing supply constraints as evidence of adoption outrunning capacity, not a sign of weakness. Whatever one makes of that framing on a shareholder call, the practical read for this sector is the same: Alphabet is not slowing down procurement, and the fleet is growing faster than it is being replaced.
Pichai’s commentary on TPUs, Tensor Processing Units, and Google’s custom-designed AI accelerator chips built for machine learning workloads rather than general-purpose computing, deserves particular attention. He described the chips as strategically important both for frontier model training and for serving models through Cloud and enterprise customers, and he said Alphabet is balancing internal TPU demand against external customer demand, including efforts to place TPUs directly inside customer data centers. Ashkenazi confirmed TPU system sales are baked into the $514 billion Cloud backlog, though most of that revenue won’t be recognized until 2027. She also said Alphabet will lean on third-party capacity as a bridge while internal capacity comes online — a detail that points to a more complex, multi-vendor hardware footprint than a single-supplier narrative would suggest, and one this sector should be tracking closely as it plans for future intake diversity.
Alphabet did not discuss decommissioning timing, retirement volumes, or end-of-life planning for any of this hardware on the call. It did confirm aggressive infrastructure expansion, sustained supply constraints, and capex guidance pointing to further growth in 2027. These are the leading indicators the ITAD and recycling sectors have learned to watch to plan their future investments. A buildout this size, weighted this heavily toward servers, does not stay in service indefinitely. The scale of Alphabet’s spending this quarter is a reasonable basis for expecting that when this hardware starts cycling out, the volume will be substantial.
