SpaceX Taps Qualcomm for Nationwide 5G Land Network Spanning Tens of Thousands of Sites

SpaceX will start building a nationwide terrestrial 5G network next year, installing basestations at tens of thousands of sites with silicon and software supplied by Qualcomm. The deal, announced today, is Qualcomm's biggest network contract since the 3G era and could reshape a US equipment market that Ericsson, Nokia and Samsung have long divided among themselves.

It also marks a strategic pivot for Elon Musk's company. Satellites alone cannot deliver a competitive mobile service, and the small cells plan announced earlier had its own shortcomings. A ground macro network at this scale would give SpaceX a terrestrial footprint equal to any one of the big three US operators — AT&T, T-Mobile or Verizon.

What has Qualcomm actually signed?

Under the terms of the agreement, Qualcomm will supply components from its Dragonwing Cellular Infrastructure portfolio for a nationwide radio access network (RAN). These include:

• QDU chips and software for the distributed units (DUs), the servers that perform the heavy computation behind a 5G service.
• QRU chips for radio units (RUs), handling beamforming and part of the digital front end.
• QTR transceivers that manage the remaining radio-frequency signal processing.

Durga Malladi, Qualcomm's general manager of technology planning, edge solutions and data centers, told reporters the deal resembles Qualcomm's work with Vietnam's Viettel — but on a far larger canvas. Viettel's deployment covers only a few thousand sites. "This is huge," Malladi said. "It is 5G, but it is a nationwide network, and when it comes to a nationwide network you have to immediately start thinking of not just a handful of sites but actually tens of thousands of sites."

The scale matches recent reports that SpaceX is negotiating with US tower owners for access to as many as 80,000 sites.

Can SpaceX cope with only 80MHz of spectrum?

Spectrum remains SpaceX's weak point. Musk bought 50MHz of 2GHz spectrum from EchoStar last September, along with 15MHz of uplink spectrum between 1695MHz and 1710MHz. Last month he added 14MHz of 800MHz spectrum from Grain Management, frequencies Grain had previously acquired from T-Mobile.

That leaves roughly 80MHz in total, against T-Mobile's 400MHz — a gap the operator's CTO John Saw was quick to highlight on LinkedIn. Qualcomm's answer is low-band massive MIMO. Where current products typically use 32 transmitters and receivers (32T32R) in midband spectrum near 3.5GHz, Qualcomm has developed 64T64R radios that work in lower bands, down to 2GHz.

"What we have done is brought in 64T64R massive MIMO radios in lower bands all the way down to 2GHz," said Malladi. "It's an innovation from our side and SpaceX actually really liked it."

Gabriel Brown, a senior principal analyst at Omdia, said Verizon has already achieved "amazing results" with 32T32R Ericsson radios in these FDD bands. "If they're spectrum constrained, which they are, then this FDD massive MIMO is the way to get the capacity but especially the coverage benefits," he said. Qualcomm has also shown reporters slides on "giga-MIMO," with 128 or even 256 elements per radio unit — something Brown said could arrive in 5G rather than waiting for 6G.

Who fills the gaps?

Qualcomm is not a systems vendor. Its RAN products cover Layer 1 functions only — the signal-processing algorithms largely run on its X100 accelerator card, which slots into standard Dell or HPE servers. A customer must still source or build Layers 2 and 3 and the core.

Viettel developed its own software stack above Layer 1, and SpaceX appears likely to follow the same path. "They still have to do the Layer 2, the Layer 3, and so on, but the foundational layer comes in from Qualcomm," said Malladi. "SpaceX actually builds additional software on top of it, which is the Layer 2 and the Layer 3 and all the way up to the core."

The deal limits the role Ericsson and Nokia could play, and it lands hard on Nokia in particular. The Finnish vendor had hoped to claw back US market share lost to Ericsson and Samsung since 2020, partly through its Nvidia GPU-based AI-RAN strategy. Qualcomm disputes that approach outright. "AI-based RAN is a lot more than just throw a GPU at it, and you're done," said Malladi. "That makes no sense, actually, from our perspective." He argued RAN workloads run more efficiently on CPUs, with AI inference handled by GPUs or NPUs elsewhere.

Does Musk have a better chance than Dish or Rakuten?

Dish and Japan's Rakuten both failed to crack the mobile market with greenfield networks and radical architectures, and SpaceX is reportedly in talks to acquire Boost Mobile from the bankrupt Dish operation — a move that could hand it wholesale agreements with AT&T and T-Mobile plus Samsung and 1Finity radios.

Neither SpaceX nor Musk has commented publicly. "Our agreement with SpaceX is structured in such a way that as the deployment begins, starting from next year, it's going to ramp up pretty rapidly from there onwards," said Malladi.