Moffett’s Remarkable Insight: Low marginal cost means nothing if you can’t sell it

The low cost per bit of 5G means almost nothing if you can’t sell the bandwidth.  I’ve been warning for at least five years that telcos’ biggest problem is overcapacity. In 95% of locations, most of the time, major telcos have excess capacity. The tech is moving incredibly fast, driving cost per bit down at 40%-50% per year.

Hans Vestberg at Verizon discovered this the hard way. He told an interviewer his 5G network would cost 90% less per bit. He would use that cost advantage to create profitable new products. When asked which products Verizon would offer, he said he didn’t want to tip off the competition.

A year and a half later, Verizon has no significant new products or revenue sources. None.

(The network also isn’t performing. Open Signal connected to 5G 4/10ths of 1% of the time.)

Craig Moffett, the best on Wall Street, warns of “The Fallacy of Marginal Cost Advantage.”

All of us are taught from our earliest studies of microeconomics that marginal cost advantage is… well, everything. The cost-advantaged widget maker will always win…

It’s not true with telecom.

Years ago, I was asked by one of the legacy long-distance networks whether the spate of next-generation fiber long-distance networks … would indeed have an “order of magnitude” advantage in marginal costs … We concluded the “next-gen” networks would indeed have marginal costs that were an order of magnitude lower than those of the legacy providers.

But that wasn’t the bad news. The real problem wasn’t the marginal cost disadvantage. It was excess capacity. The capacity of long haul fiber networks had exploded in 1997 and 1998, driven by technology advances.

We are regularly asked whether we buy the thesis that Dish Network can build a next-generation virtualized ORAN network that will have a significant marginal cost advantage,,,

Sure they can. The cost of a network is almost entirely fixed; there’s the cost of network construction, and there’s the ongoing cost of powering, infrastructure rental, repair, and maintenance. Sending more or fewer pulses of light doesn’t change any of those.

Any self-respecting network engineer will tell you that one of the first things they ever learned about network economics is that networks are built to handle the moment of peak demand. It therefore follows that the cost of a unit of capacity at any moment other than the moment of peak demand is zero. It is excess, it is not recoverable, and it is therefore valueless (a unit of capacity one second ago that expired unused isn’t helpful to anyone, and is lost forever).

In the end, there are only three things that matter: what does the network cost to build? What does it cost to operate? And how much revenue can it generate? What Dish is planning in the U.S., is a reminder that building and operating networks is hard. And earning a return on them is even harder.

The key question for Dish, Rakuten, 1+1 Drillisch, and other new networks is not whether their costs are low. It’s whether that can be turned into sales. Jio in India proved it was possible, reaching ~400 million subscribers in 4 years. It won’t be easy.

Early results on Open Ran, SDN, NFV, and the like don’t match the hype.

Alex Choi of Deutsche Telekom is enthusiastic about Open RAN and the other buzzwords as near-future technologies, still with challenges. That’s the near-universal consensus of the top network engineers. That doesn’t mean the new networks shouldn’t pioneer, but I’ve been seeing some unfortunate datapoints”

• Vodafone CTO Scott Petty has been one of the most enthusiastic supporters and is doing some deployments. But Laurie Clarke of New Scientist quotes Petty, “We believe by 2023, we may be able to deploy some scale in the rural parts of our network, but it will take until 2025 to be able to deploy at real scale in our denser urban and suburban areas.”
• Huawei and Ericsson both claim the standard, off the shelf hardware, just isn’t fast enough.  Tareq Amin at Rakuten found out the hard way. “Rakuten pays hundreds of millions of dollars in non-recurring engineering fees to chipmakers like Qualcomm to obtain the components it needs.” Thanks to John Hendel at Politico for the revealing insight.
• AT&T was the first strong backer of SDN/NFV/Ecomp, pouring a fortune into open software. It is quietly cutting back. A friend has been receiving resumes from senior people at AT&T.