When people imagine the climate cost of AI, they picture electricity: humming racks, spinning meters, power plants, and lakes getting drained to cool it down. Although this is true, it is only half of the reality.
A meaningful share of server’s lifetime emissions is before the machine is even started. Mining, refining, fabrication, assembly and shipping make up a significant share of emissions from servers. Semiconductor fabrication is extraordinarily resource-intensive.
Why this changes the second-life math
This is where most people are wrong about extending the life of an existing server. By delaying the purchase of a new server, you avoid re-paying the embodied carbon-bill that is generated from the manufacturing process. A rescued A100 arrives with its footprint already sunk. Every additional hour it delivers decreases the footprint even further, whereas a new card starts from zero.
That’s why ‘reuse before recycle’ is not a slogan for us. It is how we operate. We recover materials through recycling, reuse recovers the whole object, footprint and all. The greenest GPU is, in fact, one that already existed.