The document is called “C-RAN: The Road Towards Green RAN,” and I only found it because I went looking for why my LTE trial coverage keeps getting explained to me in terms of electricity bills. China Mobile’s research institute published it this month, and the core idea is aggressive enough that I read it twice to be sure I wasn’t misreading a translation.

Here is the problem they are staring at. Every cell site is a little building with its own baseband hardware and its own power draw, plus an air conditioner running around the clock to keep that hardware from cooking itself. Multiply that by the scale China Mobile operates at, hundreds of thousands of sites, and the base station stops being a radio and starts being an energy line item. Powering and cooling all that distributed compute is, by their accounting, where most of the network’s energy actually goes. Not the radios talking to your handset. The boxes at the bottom of the tower.

So the proposal is to empty the tower. Take the baseband units out of every site and haul them back into one central room where they can be pooled. What stays at the mast is just the radio itself, the remote radio head, connected back to that central pool over fiber. They call the pool a BBU hotel: one air-conditioned room doing the processing for a whole cluster of sites, instead of a thousand hot little closets each running at a fraction of capacity.

What I find sharp is the pooling logic, not the energy story. If your baseband is centralized, you can share it. A tower covering an office district is slammed at 10am and idle at midnight; a tower over apartments runs the other way. Today each one carries its own dedicated hardware sized for its own peak, which means both sit half-wasted most of the day. Pool the compute and one site’s quiet hour pays for another’s rush. That is the kind of statistical multiplexing data centers have been living on for years, finally pointed at the radio network.

I am not going to pretend this is close to real. The paper is upfront that the whole thing rides on fronthaul, the fiber between the central pool and the radio heads, and that link has to carry raw radio samples at brutal bandwidth with almost no tolerance for delay. Whether operators outside China can get that fiber economically is a question the document mostly waves at. And there is a version of this where centralizing everything just relocates the cost instead of removing it.

Still. I keep coming back to one implication the authors are careful about but clearly mean. Once your baseband is software running in a central pool on general compute, the radio access network starts looking less like telecom equipment and more like a data center that happens to have antennas. Whoever writes that software, and whoever owns the interfaces between the pool and the radios, ends up owning the network.

Nobody in Europe or North America is talking about this paper yet, as far as I can tell. Right now it looks like a Chinese operator solving a Chinese electricity bill. I don’t think it stays that.

Sources