Verizon is proving to all of us that 5G millimeter wave can effectively replace landlines in many places. The technology works. However, CEO Lowell McAdam is clear the financial case is still to be proven. Rupert Wood of Analysys Mason is an always interesting analyst. He's speaking at the excellent TNO Ultrabroadband Conference June 27.
In particular, can the cost be brought down enough to make money with a 5G millimeter wave network for fixed wireless? The range today is short, which means you'll need many, many cells with a massive backhaul network. One estimate I've heard is that the U.K. would need a million or so. Building that is so expensive should you go all the way home with the fiber?
Rupert's tentative conclusion: the new 5G entrant would need to "buy/build something like the network you’re wanting to displace." He looks at the question of capital efficiency. (More of his comments below.)
A major overbuild is certainly possible; Verizon is spending $300M on fiber in Boston, where they want to turn off their copper.
Some premises would get fiber direct; others, a millimeter wave network. In other places, the incumbent telco has enough fiber to bring the cost down. Without a doubt, millimeter wave will be the right choice in some places.
Two claimed technical improvements would make a major difference. Ericsson's Karolina Wikander points out that today's wireless backhaul has very low latency and might work for less expensive backhaul. NTT's CTO Onoe believes that beamforming can improve millimeter wave reach; the short range of mmWave may be a "myth." They are both respected engineers worth listening to, but neither technology is established.
Onoe and almost everyone else doubts millimeter wave for mobile will see volume until well into the next decade. Linley Gwinnapp shared with me a private analysis paper that points to the possible high cost of millimeter wave mobile phones.
"Processing 5G’s wider channels and extreme data rates could require 10x more transistors than in today’s phones; these modem chips will be very expensive even if built in 10nm or 7nm technology. Without Moore’s Law, these chips won’t decline in price over time. Furthermore, the higher data rates operate only in new bands at 3GHz and above, adding RF components and antennas to existing designs.
(Linley has a great headline for the story that Intel is exiting 3G and 4G modem chips because their Atom chip didn't make it: Intel Exits Mobile as Atom Bombs. Intel has lost something like $10B over the years trying to find a niche in communications chips. Will Strauss, another analyst I respect, has a colorful look at Intel's struggles.)
Is Moore's Law dying? Marconi Fellow Henry Samueli has contended for several years that a Moore's Law scaling will become challenging as we approach atomic scale. Most chip engineers agree with him. If you listen to Henry closely, you'll realize he is not saying chip progress is over. The historic performance increase driven by smaller feature sizes needs to slow, but shrinks are unlikely to be the driving force after the next few years. I'm pretty confident that advances like faster stepper motors will continue bringing down chip prices, but I suspect heat and size will be harder to advance.
As I wrote this, I came to the tentative conclusion that millimeter wave fixed will be a comparatively modest market until millimeter wave mobile becomes commercial next decade.
Millimeter wave is close to a natural monopoly; two networks cost nearly twice as much to build as one. The four to seven networks normally required for competition to work well seem impossible to me. At the moment, I'm a voice in the wilderness in policy circles saying we need to be ready for competition to become even more limited. Competition as the only solution has become close to religion for most policymakers. Competition is great when it works. Often it doesn't work.
Mike O'Reilly and Jessica Rosenworcel (whom I respected) are burying their heads in the sand because they don't know what to do about weaker competition,
Here's the note from Rupert Wood. The high capital cost of millimeter wave is challenging.
I’m planning a report this summer on the subject.