So, an earnings call and an interview, both from last week, and we have a better picture (growing clearer over a number of quarters, already) of Lumen: the US fiber provider said following its Q2 results (where strategic revenue jumped to 53% of its takings) that it is not trying to return its legacy telco business to growth, but to build – and the company is digging with intent, faster than ever (with another 40 million miles of fiber in the US between now and 2031, to take its total footprint to 58 million miles) – around it, and on top of it. This is fairly standard telco-2.0 stuff, but it is rarely so well communicated, and perhaps so well executed. It is certainly backing itself; chief executive Kathleen Johnson told analysts on the call (August 4) that Lumen is combining its network, software, and ecosystem in a way “no other traditional [telco] is”. AI is the justification, of course.

AI makes distributed infrastructure more complicated, it reasons – going in between data-center clouds and AI-factory clusters (‘east-west’; DCI), and up and down to enterprise premises (north-south’) – and creates demand for simpler connectivity and control along the way. The scale-across is where the biggest buzz is, increasingly long-haul on “AI corridors” between scattered clouds. This is why Google Cloud has just appointed Verizon on a £1bn DCI deal (eventually) for dark fiber, to use as it wishes; it is why Zayo is building 8,000 miles of new long-haul fiber with Nvidia, across the “fastest-growing AI corridors” (where “AI demand is emerging, not just adding capacity where networks already exist”). But the economics are complicated; in fact, they’re “terrible”, Lumen said on the call.

“There’s great headline-generating news in those [fiber builds], but the returns are terrible,” said Christopher Stansbury, president and chief financial officer, responding to a question from Bank of America (BofASecurities). “And what I mean by that is that if you look at the cost to build new trenches, deploy new conduits, you’re looking at returns that are at or below cost of capital levels.” But Lumen also doesn’t care. It has this new Alkira business – its “bullseye” purchase, completed a month ago, being integrated into its Lumen Connect – which changes its calculation, and lets customers piggyback on any fiber anywhere, with the same policy and service controls, and a single “pane-of-glass” to manage it. Value will migrate upwards, it says, and it won’t make the mistakes of the past.

Johnson said on the call that telcos lost, last time, because innovation happened around the network, and not in it. Lumen wants to reclaim the software layer telcos ceded to cloud vendors. Alkira is the mechanism to pull control and intelligence back inside the network – while its own fiber routes, being quickly doubled (more than) in the US, and also upgraded to 400G at 90% of hyperscaler on-ramps, gives it something over-the-top software companies don’t have. That is the wager, anyway: own enough of the physical layer to make the network useful, but not so much that every new customer or route requires another pricey trench. Alkira makes the proposition carrier-agnostic; Lumen can sell the control plane over its own network where possible, and over somebody else’s otherwise.

“The carrier should be the right carrier who can provide the best connectivity, the best latency,” said chief technology officer Jim Fowler on a call with RCR last week. Which sounds almost heretical for a telco, until you think what Lumen – like everyone; but more immediately, and possibly more succinctly – is trying to sell. It is not fiber. Or not principally fiber. It is the ability to make a complicated collection of fiber, clouds, clusters, and security functions behave like a single network. Lumen’s NaaS business is growing at three times the market rate, reckons Fowler. More than 20% of NaaS adopters in Q2 were new to Lumen. Johnson said on the call existing customers were adding NaaS circuits rather than simply migrating away from legacy services.

Strategic revenue grew 14% year-on-year and reached 53% of total business revenue, up from 45% a year earlier. The 100G and 400G wave business in North American enterprise grew nearly 11%, with sales up almost 35%. That last number points to the physical demand underneath all the talk about software. AI needs capacity, and increasingly it needs capacity between places that were not previously so heavily interconnected – per Zayo’s strategy, too. Lumen says the market for east-west connectivity – cloud to cloud, data center to cloud – is growing at 10-13%, against only 1-2% for traditional enterprise north-south premise-to-cloud market. Who knows what that means for the edge inference story, supposed to have a whirligig impact on the southern-end of the latter.

But Lumen’s strategy with its legacy business is also interesting. Legacy revenue was down 15% in the quarter. “Extending the life of legacy products… will not be a pillar of our future success,” said Stansbury. Enterprise voice, say, is going end-of-life, contributing a “very-low single digit” percentage of sales. The firm is maximizing cash from the old network to redeploy it into higher-return services. Once customers are through the door, the sums improve sharply. “When you go in with the second, third, fourth service, you get very close to zero marginal cost and little-to-no cap-ex.” Lumen said its internal “modernization/simplification” will deliver $700m of annual savings by the end of this year and $1 billion next year.

The exercise has also taught the company the “economic value of individual circuits and routes to market” – to be “more aggressive on how we ultimately wind down that portfolio”, said Stansbury. AI forecasts are being revised constantly, including down, and software does not just turn low-margin telcos into high-return tech-cos – or whatever. Digital revenue is only $39m for now, so it is hardly a financial transformation yet. But Lumen is at least making a coherent attempt to change the game. Maybe better than most others. More to follow from Lumen.