On 4 November, the headline was that SpaceX had picked Qualcomm for its mobile network. Qualcomm’s shares rose, SpaceX’s fell, and the story moved on. I think the more important question is a different one: who sits in the middle? For thirty years, a handful of companies sold mobile operators the radio networks they run on. If SpaceX designs more of its own, the people who should read this announcement most carefully are not the carriers but the vendors.
This article looks at what was announced, what the analysts say, who actually makes what in this industry, and where I think the real risk is.
What was announced
SpaceX and Qualcomm signed a multi-year agreement under which Qualcomm will supply infrastructure chips and software for Starlink Mobile’s terrestrial 5G network. Reports name Qualcomm’s Dragonwing QRU100 platform for radio units and QDU100 for the digital unit, plus commercial-grade physical-layer software, in massive MIMO radios with 64 transmit and 64 receive paths. Deployment starts in 2027 at tens of thousands of cell sites, in waves across several spectrum bands. Financial terms were not disclosed.
Qualcomm CEO Cristiano Amon called it “one of the largest 5G deployments in the United States” and “the beginning of a strategic relationship with tremendous potential,” but added that the revenue is “more of a fiscal 2028 from a revenue standpoint.” Qualcomm shares rose more than 5 percent that day, and SpaceX fell almost 4 percent.
What Qualcomm supplies, and what SpaceX keeps
The detail that matters is the split of responsibilities. Qualcomm provides silicon and the lowest software layer. Michael Nicolls of Starlink said the partnership should “accelerate our timeline to develop our base station end-to-end architecture.” Michael Thelander of SRG Research noted that “SpaceX is handling all the higher layers of the protocol stack.”
Two analysts quoted by RCR Wireless went a step further. Vish Nandlall said: “A key question is whether SpaceX builds the radio. I suspect yes.” Leonard Lee of neXt Curve concluded that “it looks like SpaceX is going to build its own flavor of RAN.”
These are informed readings, not statements by the companies. Neither has said who will manufacture the finished radios or who will build the core network.
Who makes what today
According to Dell’Oro Group, the five largest radio-access vendors in the first half of 2026 were Huawei, Ericsson, Nokia, ZTE and Samsung, together about 96 percent of the global market. The ranking did not change, and revenue is expected to be broadly flat for 2026 after more than two years of contraction. Huawei had a strong second quarter, while Ericsson’s results were softer than expected.
One correction to the way the story is often told. Huawei and ZTE are not part of this competition in the US. Their equipment has been pushed out of American networks under the Secure Equipment Act and the FCC’s rip-and-replace programme. The first-order effect of a US deal falls on Ericsson, Nokia and Samsung, which supply the American carriers. Whatever it means for Huawei and ZTE would come later, and in other markets.
The case that vendors should worry
- The unbundling pattern. Each step in the last decade, from cloud RAN to Open RAN, moved value away from the integrated box and toward silicon and software. A buyer that takes the chips and writes the rest is the logical end of that line. Commentary from 5GWorldPro, written by an industry veteran, makes this point: the traditional vendors “did not win this at the silicon layer,” and their exposure depends on whether SpaceX later buys their radios, software or core.
- A template. If one network proves that chips plus in-house software can run at scale, other large buyers will ask why they pay for a finished system. The same platforms already run in commercial networks in Japan and Vietnam, according to that commentary.
- A weak market. A flat market leaves little room for losing even a mid-sized customer.
The case that they should not
- SpaceX has not replaced anyone yet. It has not named a radio manufacturer or a core supplier. Roth Capital’s Scott Searle listed Ericsson and Nokia among possible beneficiaries of SpaceX’s reported 80,000-site buildout, which Bloomberg reported and which is unconfirmed.
- Scale and spectrum. Counterpoint Research’s Neil Shah estimates that SpaceX holds roughly 80 MHz of spectrum, against 300 to 400 MHz or more for each of the three big carriers, and says that “matching Verizon, AT&T and T-Mobile on density and capacity will take years and serious capital.” The 800 MHz purchase also still needs FCC approval.
- Integration is hard. Open RAN integration has proved slower than planned elsewhere, as it did at Dish, according to the 5GWorldPro commentary. Vendors sell integration, support and a core network along with radios.
- It is one customer. Even a large US network is a small part of a global business.
My reading
I do not expect Ericsson or Nokia to feel this in their results soon. The real signal is quieter. A customer with its own software engineers, its own satellites and a distribution base of millions of Starlink users is buying components, not systems. SpaceX already designs its own satellites and its own user terminals, so building its own base-station architecture would be consistent with how it works. That makes it unlike the carriers that vendors have served for decades.
If this works, vendors will not lose everything. They will lose the right to be the default. They will have to prove, for each customer, what a finished system gives that a chip supplier and a good engineering team do not: a tested core, support at scale and someone to call when the network falls over.
Europe makes this more interesting. The proposed revision of the Cybersecurity Act points toward phasing out high-risk suppliers from European networks, and that is a gift to Ericsson and Nokia. A credible silicon-led option would give European operators a third route. I would not bet on it quickly, since operators in smaller markets have neither the engineers nor the budget to run their own RAN, and they will keep buying systems. But I would watch whether Qualcomm-based radios start to appear in the catalogues of integrators.
What to watch
- Who builds the radios and the core. An announcement naming a manufacturer would settle the main question.
- Ericsson and Nokia. Whether they comment on the deal, and whether they are named as suppliers, on their next results calls in January.
- Tower leases. Whether the reported talks with Crown Castle and others turn into signed agreements.
- FCC approval of the 800 MHz purchase, and any wholesale deal with a US carrier.
- The first sites in 2027. Whether the rollout matches the announcement.
What we do not know
The financial terms of the Qualcomm agreement have not been disclosed, and neither have SpaceX’s per-phase schedule or the number of sites in each wave. It is not clear who will manufacture the finished radios, who will supply the core network, or how much of SpaceX’s design will be proprietary. Analyst readings of the deal differ, and some of the commentary cited here is not independent of the vendors it discusses.
Peter Rizov has more than 25 years of experience in telecommunications, financial institutions and energy.
Disclosure: Nove.news is owned by Quinta JSC, which also operates NoveSIM, a travel eSIM service. The section “My reading” is the author’s opinion. The rest of the article presents reported information and the arguments of both sides and is not a recommendation.

