- Qualcomm is the big winner here as its Dragonwing will power tens of thousands of Starlink Mobile 5G macro sites
- Analysts say Elon Musk likely skips Nokia, Ericsson and Samsung. SpaceX will build on Qualcomm's Layer 1 and handle the rest itself
- Permitting, fiber, power and backhaul may slow the rollout more than radios
Qualcomm celebrated a significant win last week, saying its Dragonwing platform will support tens of thousands of terrestrial 5G macro sites for Starlink Mobile’s 5G network across the U.S. Deployment will begin in 2027.
The deal certainly raises Qualcomm’s profile in a segment of the industry it hasn’t been super active in for decades: network infrastructure. But it also raises the usual questions about which additional vendors SpaceX might tap and which ones it doesn’t need.
Unlike traditional mobile operators, it’s a good bet that Elon Musk won’t need the likes of Nokia, Ericsson or Samsung to build his network, according to analysts.
“He’s a builder. He’s not a buyer,” said Roger Entner, founder of Recon Analytics. “Look at Tesla. He has his own semiconductors. He has his own batteries. The only thing he does not make himself are the tires. Everything else is made by Tesla.”
Ookla’s lead industry analyst Mike Dano agrees that SpaceX has the manufacturing muscle, saying it’s entirely possible that SpaceX will be able to build the radios it needs for its terrestrial network using Qualcomm’s platform. “SpaceX has extensive, vertically integrated manufacturing capabilities that it uses for its fixed broadband service,” he noted.
The power of towers
Building the gear is one thing. Finding a place to hang it is another. Reports surfaced last month that SpaceX is in talks with cell tower owners like Crown Castle and others to secure at least 80,000 sites for its antennas.
“I expect the company to mostly use traditional, high-elevation cell tower sites for this buildout, in order to leverage the propagation characteristics of the 800 MHz spectrum it’s purchasing from Grain. Like EchoStar/Dish before it, I expect SpaceX to begin its buildout in downtown areas and expand out from there,” Dano said.
One wild card is Dish’s abandoned gear. There’s still some speculation that Musk could buy Dish’s old radios and associated gear that is still sitting on towers and other sites across the country, although that would run counter, at least temporarily, to Musk building his own radios.
It’s possible that new gear made by and specifically for Starlink Mobile will take the place of Dish equipment on cell towers and other structures – but if that is the case, it won’t be in the prime position. Because it was the fourth entrant, “Dish had the lowest rung on the ladder,” meaning not the best location, said Earl Lum, president of EJL Wireless Research.
Unless Musk and his engineers have come up with some other way, “he’s going to be working at a disadvantage from many of the towers because he’s got the lowest real estate on the tower, which means that his reach and the footprint of that cell site and the diameter of that cell site is going to be less than the guy on the top,” Lum said. “He may lose several 100 feet of cell site range because of his positioning.”
64T64R everywhere? Analysts aren’t sold
On the radio side, Qualcomm will be providing Layer 1 processing and it will be up to SpaceX to acquire the remaining infrastructure with Layer 2 and Layer 3 software, possibly using ODMs for equipment boxes. Starlink Mobile will use Massive MIMO radio units with 64 transmit and 64 receive paths (64T64R), an architecture that will improve coverage, capacity and spectral efficiency, enabling the network to serve more traffic efficiently, according to Qualcomm.
Lum said he thinks 64T64R is over-rated and that 32T radios would be adequate. “What you get from 64 and with the penalty you pay in power and cost and size, I think you’re better off with 32,” he said.
Signals Research Group (SRG) President Mike Thelander agreed that “it’s a bit surprising that everything is 64T64R since we are not convinced it is needed everywhere while it may be difficult to deploy everywhere.”
Then again, “we are sure SpaceX is aware of this concern while it needs all the help it can get with coverage and capacity given its existing spectrum holdings,” Thelander said in a Nov. 9 SRG report.
Why Qualcomm is the smart bet
Overall, Lum said he sees Qualcomm as the right call. In a LinkedIn post, he spelled out why Qualcomm as the chip supplier is a smart bet for Starlink Mobile.
“The platform works. This doesn't mean that he won't design and build his own ASICs in the future, but time is money and he doesn't have the time to develop ASICs to build out the Starlink Mobile network by 2027,” Lum said. “Using Qualcomm is the shortest path to deployment. Also Qualcomm already has reference designs for the DUs, macro RUs, and massive MIMO AAUs so Starlink Mobile is not starting from zero.”
As for funding, with a market cap of more than $1.9 trillion, SpaceX certainly seems to have the capital to buy the components and build all new radios.
“All of the suppliers are already there in an existing ecosystem,” Lum told Fierce. “He just has to go to all the ones that are already there that have been supplying Ericsson or whoever and buy from them. Easy peasy from the purchasing part of it.”
Before any of that gear goes live, the radios will need to pass certification at the Federal Communications Commission (FCC) and be assigned an FCC ID number before they’re actually deployed in the field.
“For Starlink, that should not be a problem. He’s got enough engineers,” Lum said. “Let’s just say if he has a hole in his expertise, that will be easily filled.”
Who else gets a piece?
If SpaceX does go shopping, potential radio components include power amplifiers from Skyworks, which merged with Qorvo earlier this year, as well as memory from suppliers like Micron, Samsung or Hynix. Other possible vendors include Marvell or Broadcom for Layer 2 and Layer 3 switching and Dell or HPE for server technology.
Other vendors that could be beneficiaries are Cisco and Ciena in optical switching.
Antennas could be one of SpaceX’s differentiators. Lum said he’d recommend using MatSing lens antennas or “some other equivalent multi-beam solutions” from Amphenol and other antenna suppliers. “This would potentially blow away network performance when compared with AT&T, T-Mobile and Verizon,” he said in the LinkedIn post.
Of course, they’re going to need backhaul. Lum said sites may require 1-2 Gbps of dedicated backhaul, with some needing 5 Gbps.
Lum added that “anyone with fiber, whether it’s regional or national, they’re gonna make out like a bandit.”
All told, the permitting process, plus tower access, fiber leases, power, backup systems and installation logistics may limit the network rollout more than anything that has to do with radio manufacturing.
Big prize in the sky
The terrestrial buildout is only half of the story. Given SpaceX’s propensity to do everything on its own, the satellite component may be the most interesting, according to analysts.
Ian Fogg, research director at CCS Insight, said an interesting question is whether SpaceX will look to use more external suppliers for its second generation direct-to-device (D2D) satellites as it expands its 5G support.
“Winning a deal to supply components for Starlink Mobile satellites would be an even more significant win for a traditional cellular network supplier, because it would place the company at the forefront of innovation in a growth area,” he told Fierce via email. “Combining satellites with terrestrial connectivity is a key part of upcoming 6G standards. It is part of the 6G pillar called ‘ubiquitous connectivity.’”
For now, Qualcomm has the best seat at the table. Whether a traditional RAN vendor and others in the wireless supply chain are invited to SpaceX’s party is very much up in the air.
Read more stories about SpaceX on Fierce Network:
T-Mobile execs stay unfazed by SpaceX threat
SpaceX’s nationwide terrestrial 5G plans get real as Qualcomm signs on
Opinion: The Orbital Gambit — physics is settled; economics isn’t
Do SpaceX’s ‘missing pieces’ carry a Dish logo on them?