Bernstein: Why are US carriers refusing to open their networks to Starlink?

What SpaceX moves are putting pressure on carriers?

SpaceX has reached an agreement to acquire Grain’s 800 MHz spectrum and has reportedly discussed using at least 80 thousand US cell sites with tower owners. Low-band spectrum could improve the direct-to-phone experience and reduce the time and cost of building a terrestrial network; the site talks have yet to produce a final agreement or a commitment to build a network.
Although the three major carriers have operated steadily so far this year, shares of AT&T, Verizon and T-Mobile remain 10%–20% below their April highs. Management teams at all three currently rule out giving Starlink mobile virtual network operator access, which would allow it to serve customers using their existing terrestrial mobile networks.
Bernstein believes SpaceX’s existing spectrum remains insufficient to replicate incumbent carriers’ networks. A nationwide competing network would require substantial capital and time, but the company has shown no sign of abandoning a mobile network.

A partner could earn wholesale revenue based on Starlink users’ traffic on its network, while the virtual operator would bear the costs of marketing, customer service, billing and handset subsidies. But if those users were already the partner’s retail customers, wholesale revenue would also need to offset the lost retail profit.
Bernstein estimates that reasonable wholesale revenue per Starlink line is about 5–15 dollars a month. By comparison, a typical postpaid retail customer generates about 55 dollars in monthly revenue and contributes about 22 dollars in EBITDA.
The firm gives an example: 10 Starlink users each contribute 10 dollars in monthly wholesale profit, with 6 coming from competitors and 4 from the partner itself. The 60 dollars in additional profit, less the 12-dollar difference in profit for each of the 4 existing customers, leaves the partner with a total net monthly gain of 12 dollars.

Wholesale fees are usually based on the traffic actually carried. If Starlink shifts more traffic to its own satellites or a future terrestrial network, the partner’s revenue per user will fall; however, seamless switching between satellite and terrestrial networks and satellite capacity remain subject to engineering constraints.
Bernstein previously calculated that, as the share of traffic diverted rises from 88% to 91%, the share remaining on the partner’s network falls from 12% to 9%. The lower the wholesale contribution, the more Starlink needs to attract users from other carriers to offset cannibalization of the partner’s own retail business.
Where customers come from also depends on Starlink’s pricing, brand and bundles. Unlike cable operators with established broadband customer bases, Starlink has yet to build a comparable broadband user base, making the benefits and geographic distribution of cross-selling harder to predict.

Bernstein estimates that, with wholesale access, Starlink would need about 1–1.5 years to launch service. If a partnership agreement is attractive enough, it could reduce the incentive to build its own nationwide network; Starlink could also use it to build a customer base first and then gradually shift traffic to its own facilities.
Without wholesale access, the firm estimates that SpaceX would need about 5–10 years to build its own consumer-grade nationwide network and could begin offering service after 3–5 years. Denying access can therefore delay competition, but may not eliminate it.
Bernstein believes the long-term impact depends more on whether Starlink can own its network and capture the economic benefits of its own facilities. If SpaceX builds a network regardless, a carrier granting access could earn wholesale profits initially while also helping a future competitor enter the market faster. The firm maintains its Market Perform ratings on AT&T, Verizon and T-Mobile and its Outperform rating on SpaceX.