We buy retired enterprise servers and put them back to work

Most of what a server costs the planet is spent before it answers a single request. Mining, refining, fabricating, shipping. That cost is paid once, when the machine is built, and the only thing that lowers it is keeping the machine running longer. So we buy enterprise hardware on its way to scrap, rebuild it, and write software small enough to run on it.

Embodied carbon
Most of a server's footprint is spent before it is switched on
Second service life
Enterprise machines rebuilt after their depreciation schedule ends
CPU inference
Models small enough to serve without a GPU
Lower prices
The saving from all of it shows up on your invoice

It only works in this order

Each step makes the next one possible. Take away the first and none of the rest hold.

Write lean softwareCompute got cheap, so software got slow. Nothing forced it to stay small any more. Making it small again is work, and every step below this one depends on it.
Modest hardware becomes sufficientOnce the software fits, the machine under it does not have to be current. What kept forcing the upgrades was the code on top of it, not the silicon.
Scrap hardware becomes viableEnterprise servers leave service when a depreciation schedule ends, not when they stop working. A fleet that only has to be sufficient can be built out of what that schedule discards.
Less waste, and a lower billThe environmental result and the price come out of the same decision. We lead with the price, because you can check it.

Where the carbon is spent

New silicon usually wins on efficiency per watt. Each generation does more work per unit of energy than the one it replaced. An argument built on beating new hardware there would last about one product cycle.

Embodied carbon works differently. It covers everything spent mining, refining, fabricating, and shipping a machine before it serves anyone. It is paid once, at manufacture, and the only thing that moves the number is how long the machine stays in service.

The largest operators reached the same place from the accounting side. Alphabet and Microsoft both extended how long they keep servers, because the machines still worked long after the depreciation schedule said to replace them.

AlphabetExtended the estimated useful life of its servers and reported that the change cut depreciation expense by roughly 3 billion dollars in 2023.
MicrosoftReported the same move for server and network equipment, from a four-year depreciation schedule to a six-year one.

No server has a smaller footprint than the one nobody had to build.

e-waste / the stream we buy from
Generated in 2022
62 Mt
Formally recycled
22.3%
Projected for 2030
82 Mt

UN Global E-waste Monitor 2024.

veritate / compute
unit active on this byteINT4 weights

Shrinking the model is the other half

Reusing hardware only helps if the work fits on it. Making the models smaller is the same problem one layer up from the rack, and it is what the research program works on.

Folding quantization into training took a research model from 326 MB to 17 MB on disk. A shipping recipe holds an effective 1.98 bits per weight at a cost of 0.021 in cross-entropy. Activation sparsity trained into the network reaches 87.5 percent zeros at 0.037 cross-entropy, after about three minutes of fine-tuning on a CPU. A KV cache gives a 19x speedup at a 512-byte context with byte-exact output, and a 197 thousand parameter distilled exit head drafts ahead for another 1.74x, again byte-identical.

Each figure carries the quality cost we measured beside it, and the research log says which ones are wall-clock.

Ternary QAT326 MB to 17 MB
Effective precision1.98 bits/weight
Activation sparsity87.5% zeros
KV cache19x, byte-exact

What this looks like as a product

The chat runs on the fleet described here and costs nothing to use. The inference API is priced per token on the same hardware, and the cloud runs on the same fleet at published rates.