- Lumen is working with Emerald AI to route AI workload data toward available power—not just available GPUs
- The plan pairs Lumen’s programmable network with Emerald AI’s Conductor platform to coordinate data routing and data center power use
- Power-aware AI workload routing could help ease grid strain, but potential cost savings for Lumen customers remain unclear
Lumen doesn’t want to be a passive observer of the AI revolution. It wants its network to be an active participant. To that end, the operator is working with Emerald AI to give customers a new facet of control over how their data is routed.
Today, companies looking to run AI workloads are chasing GPU availability. But Lumen Chief Strategy and Marketing Officer Ryan Asdourian told Fierce that the company believes in the next few years they will need to chase chips and power availability.
So, just as cloud companies like Google allow customers to shift workloads between facilities to chase renewable energy availability, so too does Lumen want to enable customers to send their data to where cheap and plentiful power is available.
The plan hinges on combining the programmable network capabilities Lumen has built with Emerald AI’s Conductor platform, which allows data center operators to flexibly control power use by orchestrating and reallocating workloads and power draw across a fleet of data center assets.
“It’s really a bit of a first of its kind effort where, if the data and the customer is power aware, we will help route all that data to where the power is available,” Asdourian explained. “We believe that especially in the next few years, you’re going to need to construct your plans so that that that data matches where that [power] is available.”
Founded in 2024, Emerald AI has claimed its workload-shifting technology can help free up to 100 gigawatts of capacity on the existing U.S. grid and alleviate some of the strain data centers have put on utilities. The idea has proved alluring to big tech, drawing Google and Nvidia to join Emerald as founding members of the AI Energy Management Alliance (AEMA), which launched in September.
Of course, shifting AI workloads around also requires moving the reams of data they run on. And that’s where Lumen comes in.
“The network has traditionally connected computing resources wherever they happen to be. We’re exploring how to turn that equation around, making the network an active part of how and where computing happens,” Jim Fowler, Lumens’ chief technology and product officer, said in a statement.
It’s not entirely clear whether or how much money Lumen’s customers might be able to save by intelligently routing inference workloads based on power availability.
Asdourian told Fierce turning the network into an intelligent, active participant in the AI era doesn’t stop with new power-aware controls. He also pointed to work Lumen is doing with Fortinet to make the network cyber aware and quantum ready. It is also looking at new controls around sovereignty, he added.
“Programmable gives us the flexibility, but the real mastery in providing a computing architecture is of we parter with the players that are experts in their field that have data that we may not have,” Asdourian concluded.
Read more about Lumen on Fierce Network:
Lumen offers validated dual path wavelengths for redundancy
AI is breaking bandwidth forecasts. Lumen wants enterprises to scale internet on demand
Lumen expands multi-cloud gateway outside its fiber footprint
Lumen just closed its Alkira deal – here’s what happens next