T-Mobile Is Raising Guidance Into a 46% Drawdown — and Quietly Stopped Counting the Things That Made It Legible
A beat-and-raise quarter, a second consecutive SpaceX question management swatted away, and two KPIs — phone churn and FWA nets — that went missing in action
TMUS · Earnings Call · 2026-10-28
The guide went up; the tape went the other way
By the numbers T-Mobile chose to lead with, this was a clean quarter: 274,000 postpaid net account additions, roughly 4% postpaid ARPU growth, service revenue guided to about $76.8 billion for the year (+8%), core adjusted EBITDA lifted to $37.35–$37.45 billion and adjusted free cash flow to $18.7–$18.8 billion, $100–150 million better at the midpoint. Postpaid ARPA growth guidance moved to ~3.5% from 2.5–3% — the third consecutive upward revision in twelve months, following the February guide and the July nudge to "more towards the 3%." Revenue of $22.8 billion, up 8% year over year, arrived with a 24.1% operating margin, down 0.6pp and a 14.2% net margin, a point lower. The cash engine remains the product: $4.6 billion of free cash flow against $5.5 billion of operating income, a 20.1% FCF margin. And the stock is 45.7% below its March 2025 peak of $272.83, down 19.8% over 90 sessions in one uninterrupted slide. The proximate cause isn't in the P&L: on October 9, SpaceX agreed to buy Grain Management's nationwide 800 MHz low-band portfolio — the block T-Mobile itself divested, with management still discussing "the timing of the close of the 800 megahertz transaction" on the July 2025 call — and the shares fell 13% in a single session. The keyword SpaceX spiked into this quarter's top movers for T-Mobile for the first time in that register.What management stopped counting
Craig Moffett opened the Q&A with the quiet part: "we're all sort of left a little bit in the dark without subscriber metrics." He was asking about fixed wireless and postpaid phone churn. He got neither.That is a reversal, not a refinement. In July, Peter Osvaldik volunteered it: “we had a tremendously successful churn quarter. We had postpaid phone churn of 0.85, so down year-over-year significantly.” — Peter Osvaldik, Chief Financial Officer (CFO) · 2026-07-23 In February, the commitment was explicit: “To give transparency around this metric, we'll absolutely be reporting account-based churn.” — Srini Gopalan, President, T-Mobile US · 2026-02-11 Account churn arrived; phone churn departed. Same story in broadband. André Almeida: “although we don't divulge nets, line nets, as you know, but I think we can state that we had industry-leading FWA nets this quarter.” — Andre Almeida, Chief Commercial Officer · 2026-10-28 Industry-leading, unquantified. The keyword trail agrees: postpaid phone churn sits in the top-30 but only as the thing being declined, and fallow capacity model — the number-two keyword last quarter, the frame management used for six straight quarters — now appears among this quarter's decliners. It went from being T-Mobile's cleverest arbitrage to a term it no longer needs, because the numbers behind it are no longer published. The substitute is defensible and arguably better aligned: accounts are how families buy, and ARPA is a value metric. But it means the two figures that made the growth legible to the street — phone nets and FWA adds — are now both soft. "I can't give you the spreadsheet, as always," Osvaldik quipped. He used to be able to.On, you know, Craig, we didn't give a postpaid phone churn number. We really are focused in on postpaid account net additions and postpaid account churn.
Kinetic Edge, pitched into an AI-infrastructure selloff
The genuinely new material is Kinetic Edge, the AWS-backed distributed intelligence layer for physical AI, plus the world's first AI RAN call made on a live network. Srini Gopalan's answer on how it monetizes is unusually candid and notably un-ARPU-like:A share of the bill of materials, not a share of the monthly bill. He framed it with the incremental TAM test the company applies to MVNOs — new customer base, new channel, or new capability — which is also precisely how he killed the SpaceX question for the second time this year: "I see no room for an MVNO with SpaceX either on the consumer side or on the physical and edge AI side." The tension is timing. The global tape is presently voting against the compute build-out this narrative depends on: over 30 days, AI data center is down 5.4% on 6 advancers against 50 decliners, "data center capacity" −8.0%, "high performance compute" −7.6%. T-Mobile is launching a distributed edge-compute thesis into a market de-rating compute. The one item with near-term operating leverage isn't AI revenue at all — it's spectral efficiency. John Saw admitted the AI-driven radio resource planning beat expectations in New York field tests, with Nokia pointing to 20%–100% spectral efficiency gains as AI RAN matures. That extends the fixed-wireless runway without buying a megahertz.Most of our monetization to date has been through providing that solution as a whole and charging for the solution and taking a share of the reduction of the cost of the device.