3
SpaceX has spoken for the primary time about its plans for a terrestrial cell community to work with its satellite tv for pc D2D service to rival AT&T, T-Cellular, and Verizon – together with plans to roll it out on small cells hooked up to its floor infrastructure. It’s betting on “useless zones” within the close to time period, plus a future marketplace for out of doors bodily AI – vehicles and robots – that can dwarf the present smartphone market. Analysts are nonetheless skeptical.
In sum – what to know:
Terrestrial cell – SpaceX/Starlink plans to launch next-generation D2D satellites in 2027 and add terrestrial small cells to undercut normal mobile economics, and construct a hybrid service to rival AT&T, T-Cellular, Verizon.
Machine comms – There’s nonetheless some speak about “useless zones”, however the agency framed the chance round connecting autos and robots, arguing that bandwidth demand will give Starlink a giant share of web site visitors.
Quarterly combine – SpaceX nearly doubled its revenues within the second quarter, but in addition raised capital spending by six instances – happening launch, satellite tv for pc, and AI, largely, compute capability.
SpaceX has confirmed plans to develop its Starlink satellite tv for pc service to the mainstream cell market – on small cells, ostensibly, hooked up to its current floor infrastructure. The plan pitches it squarely towards AT&T, T-Cellular, and Verizon within the US. The agency will launch new “direct-to-cell” (direct to gadgets; D2D) satellites subsequent yr (2027), and a industrial service by the top of it. It expects to take prospects from the big-three US carriers, it mentioned, as a result of its “service will probably be higher”, and to eat into their mixed revenues of about $600 billion per yr.
Nevertheless it maintained its major line about eliminating “dead-zones”, while, in customary trend, additionally speaking up a future world of linked robots and autos that can positively dwarf the prevailing people-led cell market. Analysts greeted its vaguely sketched infrastructure plan with small cells with a level of suspicion. “Elon Musk is nuts if he thinks 5G small cells with Starlink backhaul can supply higher (and cheaper) service than AT&T, Verizon, and T-Cellular,” responded Dimitris Mavrakis, senior analysis analyst at ABI Analysis, writing on social media. (In the meantime, a spent higher stage for a SpaceX rocket has crashed into the Moon after drifting round area for about 18 months.)
Starlink mentioned on a second-quarter earnings name (August 4) {that a} 13-fold improve in spectrum capability (from 5 MHz, through telcos, to 70 MHz, together with its personal), following the acquisition of 65 MHz of AWS 3/4 from EchoStar in late 2025, plus a 10-fold leap within the variety of its D2D satellites, will make its cell service “100 instances higher” than it’s at present – by the top of 2027, the implication goes. It additionally sketched a plan to construct a distributed community of small cells alongside current satellite tv for pc floor terminals to undercut standard mobile build-out prices – and successfully upend analyst forecasts that it’ll take about $100 billion in cap-ex to go face to face with AT&T, T-Cellular, and Verizon.
Besides that isn’t actually what Musk et al are planning – to tackle legacy telcos as a fourth US operator only for smartphone subscriptions. As an alternative, chief government Elon Musk and chief working officer Gwyn Shotwell talked on the decision about connecting autos, robots, and different machines within the bodily AI ecosystem. Musk mentioned: “It’s not out of the query that, in some unspecified time in the future, Starlink will ship a majority of the world’s web – at the very least in international locations the place we’re allowed to function…. And it’s not within the infinity future; it’s [in] lower than 10 years. With the arrival of AI, and humanoid robotics and automobile robotics, the urge for food for bandwidth will probably be a lot larger than prior to now.”
He went on: “You consider a human consuming or producing maybe a couple of hundred bits per second of output. Truly, the common human is outputting lower than one bit per second over the course of a day – which is 86,400 seconds. Computer systems may simply do billions of bits per second constantly over a 24-hour interval. This is the reason I might anticipate the urge for food for bandwidth to develop dramatically with the expansion of AI and robotics. Actually, I believe Starlink is the one factor that may truly service that bandwidth.” Musk recommended a 10-fold improve in V3 satellite tv for pc volumes, plus an additional 10-fold in V3 functionality, will ship a 100-times surge in D2D community capability.
However given the spiralling demand to attach bodily AI, as projected, Starlink reckons it could actually simply take up inevitable value erosion. “Even when our monetization per bit dropped by an element of 10, that will nonetheless imply a 10-times improve within the income of Starlink. I believe individuals are actually underestimating Starlink right here. This can be a huge deal.”
Timeline and structure
In the meantime, Shotwell clarified: “We’ll begin offering service [by the] finish of subsequent yr.” She mentioned the EchoStar deal provides Starlink 65 MHz of its personal spectrum, versus a 5 MHz facility enabled through partnerships with native cell operators, as presently. “It provides an enormous improve in functionality,” she mentioned. “[It] does have terrestrial parts. We positively intend to construct out the terrestrial part.” The agency’s build-out will search to keep away from the usual capital expense incurred by huge telcos, she recommended. As an alternative of deploying standard macro cell towers, Starlink is taking a look at “cap-ex environment friendly” methods to place small cells on its current broadband terminals.
Shotwell remarked: “Let me give somewhat trace: you possibly can put a mobile base station on the gear that holds a Starlink broadband dish. You may have these little femtocells across the nation, and deploy as that you must. You don’t essentially need to spend billions of {dollars} in low-band spectrum all upfront earlier than you deploy your system… Now we have a variety of nice and new concepts for the way we’re going to do it. Will probably be fairly cap-ex environment friendly.” She declined to place an “environment friendly” cap-ex determine on the undertaking.
Musk added: “As an alternative of getting to deploy these very costly and difficult-to-locate giant mobile base stations, we really feel fairly assured we are able to deploy a lot of type of small stations basically. They’re actually simply Starlink dishes that additionally present connectivity within the cell spectrum bands, and have them be everywhere. The Starlink antennas are situated on the roofs of homes and companies, so that they’ve received a type of clear view for offering connectivity on to cell telephones on the bottom. In truth, offering connectivity that’s, we expect, most likely higher and better bandwidth than what’s presently accessible from mobile suppliers. This is a crucial level.”
Shotwell said: “I anticipate us to have the ability to purchase fairly a couple of of their (the massive operators’) prospects as a result of I believe our service will probably be higher. We’ll eradicate useless zones leveraging mainly the satellites in orbit.”
Warning and watch-outs
Telecoms analysts greeted the rather-vague infrastructure plans with some warning. Writing on social media, Dimitris Mavrakis, senior analysis director at ABI Analysis, said: “Elon Musk is nuts if he thinks 5G small cells with Starlink backhaul can supply higher (and cheaper) service than AT&T, Verizon, and T-Cellular. Certain, if [SpaceX/Starlink] deploy thousands and thousands, then maybe. However connecting them just isn’t the largest drawback; getting energy to them and discovering the optimum location are the largest issues – not to mention economics, ROI and many others. Except it’s pondering of current Starlink places, which isn’t correlated with mobile protection necessities.”
He added: “The business has already concluded femtocells should not a viable major deployment mannequin, even for MNOs… Certain, [they] can be utilized for filling in protection blackspots within the dwelling, however not for nationwide cell broadband companies.”
In the meantime, Subhadip Roy, a senior analyst at Counterpoint Analysis, responded: “Starlink is not positioning itself solely as a satellite tv for pc connectivity companion… [It] nonetheless has an extended option to go [but], whereas 65 MHz is barely a fraction of the licensed spectrum held by incumbent US operators, SpaceX seems to be betting on a hybrid satellite-terrestrial structure and its current Starlink infrastructure to deploy cell protection quicker and extra cost-effectively than conventional macro networks.”
He added: “The timing is fascinating. AT&T, Verizon and T-Cellular introduced their Direct-to-Machine (D2D) three way partnership almost three months earlier than SpaceX publicly revealed its terrestrial ambitions. Whereas the JV wasn’t introduced as a response to SpaceX, it now seems to be like a strategically beneficial transfer as satellite tv for pc connectivity turns into more and more aggressive. AST SpaceMobile’s place could turn into much more vital. With AT&T and Verizon already aligned with AST, operators now have an alternate D2D ecosystem as SpaceX strikes nearer to competing for retail subscribers. Present Starlink partnerships are value watching. Starlink presently works with operators together with T-Cellular, Rogers, KDDI, Optus, One NZ, Salt, Entel and plenty of others. If Starlink finally launches its personal retail cell providing, these partnerships may face channel conflicts or evolve into new partnership fashions.
Second-quarter scores
All advised, SpaceX reported larger second-quarter income and improved working earnings, and beat analyst expectations as its rocket, satellite tv for pc, and AI items all posted positive aspects – though shares later fell, erasing most of their early positive aspects as buyers digested the corporate’s capital spending. Income rose 92 p.c, nearly double, versus the identical interval a yr in the past, ending the quarter at $7.8 billion – comfortably forward of forecasts of $6.81 billion. Adjusted EBITDA nearly tripled to $3.5 billion, up 191 p.c from a yr earlier. The corporate’s web loss narrowed by $467 million to $541 million. However its spending sky-rocketed to $18.3bn – greater than six instances its year-ago spend.
Shares initially rallied, closing up 9.4 p.c earlier than falling 8.6 p.c in after-hours buying and selling to $114.60. The corporate spent $18.4 billion in the course of the quarter, together with roughly $15.8 billion on AI compute infrastructure. It ended the quarter with $100 billion in money and marketable securities and a backlog of $47.5 billion. Administration mentioned capital expenditure will stay at present ranges via the second half of the yr.
However the outcomes underline the corporate’s diversified portfolio throughout area transportation, satellite tv for pc broadband, and AI computing. The Starlink enterprise stays its largest, producing $4.3 billion in income, up 66 p.c year-on-year, with 1.7 million net-new subscriber additions within the quarter. The launch enterprise additionally continued to develop, with area section income reaching $962 million. SpaceX mentioned it accomplished 78 launches in the course of the first half of 2026, delivering greater than 1,000 tonnes of payload into orbit. It goals to extend annual payload capability from round 2,500 tonnes at present to multiple million tonnes a yr, with a longer-term ambition of 10 million tonnes.
The AI division was the fastest-growing, with income up by 247 p.c to $2.6 billion. SpaceX will standardize on Nvidia’s Vera Rubin structure and start launching ‘Starmind’ AI satellites subsequent yr. The corporate expects AI compute capability to strategy 10 GW by the top of 2027. Grok 5, skilled on SpaceX’s engineering information collected over the previous 25 years, is anticipated earlier than the top of this yr. The corporate expects to complete 2026 with greater than 2 GW of AI compute capability, up from 1.4 GW on the finish of the second quarter, earlier than increasing to round 15 GW of energy and cooling capability by the top of 2027, with a longer-term goal of roughly 20 GW.
It expects to attain a $100 billion annualised income run fee by the top of 2026. Administration additionally pulled ahead its inside goal for $1 trillion in annual income to 2030, from a earlier projection of 2031 – whereas suggesting there may be additionally a “non-zero likelihood” of getting there as early as 2029.