Starlink spectrum deal piles Elon Musk’s plate ever higher

Elon Musk wants to bring Starlink down to earth. His communications company last week agreed to buy a portfolio of “low-band” spectrum ideal for broad mobile coverage, potentially putting it in more direct competition with regular telecoms giants. Even for someone as ambitious as Musk, this is a challenge — and for his investors, funding it will be too.

Starlink, a division of the $2tn SpaceX, already has a constellation of satellites that can beam radio waves from space straight to a customer’s phone — ideal for someone on a remote mountain road, say. But for regular mobile service in a crowded city, it’s no match for waves beamed from a nearby cell tower.

Musk’s new $8bn bundle of spectrum gives him a toehold in such places, where users are both more numerous and, generally speaking, richer. But that will still require towers and antennas. To match the near-total US coverage of companies such as AT&T, Verizon and T-Mobile could cost more than $110bn in capital expenditure and take almost a decade, Bernstein analysts calculate. The largest players have invested a combined $600bn every decade on upgrading their networks since the 1980s, MoffettNathanson analysts reckon.

On top of that, Starlink would need more spectrum. Its latest acquisition brings its holdings to about 80 megahertz. That’s a fifth of what T-Mobile controls. Unlike satellites, spectrum can’t be mass-produced and launched into orbit. It is finite, tightly regulated and supply is released only periodically through regulatory auctions or private sales.

So why bother? Musk likes solving hard problems. Rolling out broad mobile coverage might seem like child’s play compared with setting up colonies on Mars. Buying spectrum gives him option value, and the cost is small compared with the almost $950bn analysts expect the company to invest, much of it on AI data centres, between now and 2030, according to Visible Alpha.

He is also someone who firmly believes in vertical integration: making chips as well as data centres and AI models at SpaceX; batteries as well as cars and robots that run on them at Tesla. Self-driving cars and droids will need to talk to each other, swap data, stream road footage and take instructions. Satellite signals alone, which struggle to penetrate roofs and tree canopies, won’t suffice.

Buying spectrum may therefore be less like a bizarre side quest and more like part of a bigger plan. The problem is that it spells even more capital expenditure at a company already spending beyond its means. Analysts don’t expect SpaceX to be free cash flow positive until at least 2031.

Building out a telecoms network, whatever Musk’s grand vision, would be likely to push that back further. The SpaceX boss is good at doing hard things, but the hardest may be knowing where to draw the line.

gaia.freydefont@ft.com