Any time Dish Network comes up in the news with yet another press release about what brand of RAM they're using in their Open RAN edge servers (I'm being facetious...at least so far), you have analysts hopping in to state that there's no way Dish could build the network they say they'll build for the price they say they'll pay. After all, Rakuten is spending nearly 80% of Dish's expected outlay, but for a much denser network on a much smaller landmass, and all else must be equal because Rakuten and Dish are both using O-RAN tech.
Maybe they're right. But there are some salient differences between the two buildouts that muddy the waters enough that "it can't be done" isn't so clear-cut. Same with the assertion that Dish couldn't possibly meet their 70%-of-population buildout requirement with "only" 15,000 cell sites. After all, the US is 25x the size of Japan!
My assertion is that if Dish doesn't build out, it'll be because they decided not to, not because they can't.
First, let's talk about population percentages. Back in 2015, cities, comprising 3.5% of land area, covered 65.7% of the US's population. So, ~133,000 square miles. Let's bump up to 200k sq mi, one-eighth the coverage of T-Mobile's n71 5G network, to make sure we hit 70% of the population. If we're talking about 15k cell sites, that means each site has to have an average of 13.4 square miles of additive coverage. Or a unique coverage radius of 2.1 miles.
Ignoring anecdotal evidence I personally have about T-Mobile n71 coverage reaching 5+ miles from a site...and that's non-standalone limited by a Band 2 anchor...you have T-Mobile's own 1.6 million square mile number, and the 65,000 cell sites that T-Mobile ran pre-merger. I use 65K as a safe upper bound for the number of sites T-Mobile has n71 on because I know for a fact that T-Mobile has n71 on a small fraction of their sites, and to my knowledge haven't thrown n71 radios on sites that previously only had Sprint gear. Anyway, if T-Mobile needed to put n71 on every single pre-merger site to get to their 1.6 million sq mi number (or didn't but added an equivalent number of out-of-footprint sites post-merger, e.g. in North Dakota), you wind up with ~24.6 square miles of unique coverage per cell site, well above the Dish requirement I mentioned earlier.
Yeah, standalone n71 is actually quite good at covering territory. Even better if you can afford to build out a network that provides islands of native coverage in a sea of unthrottled, cheap roaming on the carrier you bought your customer base from. By comparison, Rakuten Mobile's roaming throttles to 1 Mbps after 5 GB.
Remember, Dish can already offer plans competitive with T-Mobile's MetroPCS and AT&T's CricKet right now, through their MVNO agreement with T-Mobile facilitated by last year's merger. That's how much they'll be paying T-Mobile for roaming for the next several years. So Dish can cherry-pick capacity-focused builds on its own gear for just areas that have a large concentration of users or usage, leaving the rest of the network as either a thin layer of n71 and n29 or roaming on T-Mobile. By comparison, T-Mobile has to build out its own coverage in most of the US (save maybe USCellular areas) if they want coverage cheap enough that they're comfortable providing it as native. I'm ignoring n26 at this point as, while existing base stations will support that band, Dish won't own that spectrum for another few years, if they decide to buy it at all.
Now, "only n71+29" doesn't sound like a lot of spectrum. But Dish has 5x5-15x15 of band 71, depending on the market, plus 5 MHz downlink on Band 29. Spread across Dish's comparatively tiny customer base, that's enough to provide comparable speeds to other carriers' sub-2300-MHz NR. This doesn't give Dish the ability to price their network at a premium, but that wasn't going to happen anyway.
The story of a thinly-deployed, value-oriented mobile network isn't a new one, and the previous actors in that story (most notably CricKet and MetroPCS) survived until their slivers of spectrum became too valuable relative to the revenue they were bringing in. The difference with Dish of course is that Dish's spectrum holdings are nationwide, lower-frequency (and thus easier to deploy) for the base layer, and more spacious at the higher end of the spectrum thanks to a 25x15 nationwide n70 band right around PCS. But Dish's position in the market will be similar: good-enough service available for cheaper than the alternatives.
If you're getting the impression that I expect Dish to construct their network on a shoestring budget, pitting vendors against each other in a distinctly Charlie Ergen way, resulting in a network that's rather unremarkable performance-wise but is nonetheless priced such that it attracts 10-15 million customers, maybe more...that's about right. Dish's network won't get an iPhone model that can take advantage of all its bands for years, instead focusing on devices at or below $400, subsidized down to even less, with cheap data plans.
Gloomy, but probably enough to counteract upward pricing pressure as the Big Three attempt to amortize their mid-band 5G build costs, and certainly better than betting the farm on a shrinking satellite TV business.