T Mobile Us Inc TMUS

Thesis

T-Mobile US is the #2 or #3 US wireless carrier (depending on how you count) operating inside a stable three-firm oligopoly (VZ, T, TMUS) with cable MVNOs as marginal overhang. Post-Sprint-merger, TMUS sits on the deepest mid-band 5G spectrum portfolio in the country, which is the single most important asset in this business for the next decade. Compounding case: 5G fixed wireless access (FWA) extends the addressable market beyond mobile into ~30M underserved broadband homes; postpaid phone net adds remain industry-leading; merger synergies (now largely captured) drove ROIIC of 47.1% over 5 years, well above peers. The scorecard shows composite 76, ROIC 10y average 6.87% (telco-typical, capital-heavy), net debt/EBITDA of -0.04 (effectively zero leverage at EBITDA level — surprising for a US telco), and PE TTM 18.83 versus 10-year average 37.81. The reverse DCF implies -3.4% growth, meaning Mr. Market is pricing modest decline. The IV base of $782 against a $196 price gives a px/IV of 0.2507 — on its face a 4x. But the scorer warns: "Maintenance capex uncertain (>50% spread)" and "base CAGR clamped from 30.9% to 14.0%." Both flags compress the real margin of safety. At $196, you are paying ~14.5x EV/FCF for a scaled, low-leverage wireless oligopolist. Owning makes sense if (a) FWA economics hold and (b) maintenance capex truly normalizes near reported D&A. If maintenance capex is actually closer to gross capex, the FCF-based IV collapses by half. Buy below $200 with a bull-case trim above $450; recognize the IV range is wider than the scorer's headline suggests.

Moat

T-Mobile's moat is built on three of the five classic moat types: cost advantages, intangible assets (spectrum and brand), and modest switching costs. Pricing power is weak; network effects are essentially absent.

Cost advantage (primary moat). US wireless is a scale game. The fixed-cost base — towers, backhaul, spectrum amortization, retail distribution, billing systems — is enormous, and marginal cost per subscriber is near zero once the network is built. Buffett's framing of GEICO applies almost verbatim: "when a company is selling a product with commodity-like economic characteristics, being the low-cost producer is all-important" [2]. Wireless service is exactly such a commodity. TMUS, post-Sprint, has the highest spectrum-per-subscriber ratio of the Big Three, which translates directly into lower cost per gigabyte delivered. The Sprint merger added scale that had been spread across two sub-scale carriers; the ROIIC of 47.1% over 5 years is the fingerprint of synergy capture. Stress test: a $10B competitor entering today would need to build a fourth national network and acquire mid-band spectrum that simply isn't available in usable quantities — DISH's struggle to do exactly this is the live experiment, and the answer is clear. Erosion risk: cable MVNOs (Comcast Xfinity Mobile, Charter Spectrum Mobile) ride on Verizon's network at wholesale, and they compete on price for ~10–15% of net adds.

Intangibles — spectrum licenses. This is the closest analog to GEICO's regulatory cost advantage [1]. Spectrum is granted by the FCC, finite, and increasingly expensive — the C-band auction cleared at $80B+. TMUS's 2.5 GHz holdings (acquired via Sprint) are uniquely suited for 5G mid-band coverage. Buffett's MidAmerican framing — regulated returns where regulators "will treat capital providers in a manner that will ensure the continued flow of funds" [3] — partially applies, though spectrum is property, not a regulated rate base. The intangible has 10–15-year staying power.

Switching costs (modest). Number portability since 2003 reduced switching costs deliberately. What remains: device installment plans (24–36 month financing), family plans, autopay discounts, bundled streaming (Apple TV+, Netflix). Churn is ~0.9% monthly for postpaid phone — meaningful but not a fortress. Buffett's GEICO at 9.3% market share grew because the cost advantage compounded year over year [4]; TMUS shows the same fingerprint, with postpaid net adds leading the industry for ~25 consecutive quarters.

Pricing power (weak). The Un-carrier playbook was built on UNDERCUTTING the incumbents. TMUS cannot now flip to ATT/VZ-style pricing without inviting MVNO and cable retaliation. Recent price hikes on legacy plans triggered customer backlash and FCC scrutiny.

Network effects (essentially none). Telephone networks have weak interconnection effects; there is no winner-take-all dynamic.

Competitor stress test: $10B + 5 years of effort would not produce a fourth national carrier. The DISH/EchoStar saga is the live counterexample — billions spent, no scaled subscriber base, now in financial distress. The three-player structure is durable for a decade.

Erosion risks ranked: (1) cable MVNO share gain at the low end, (2) regulatory action on consolidation or spectrum, (3) FWA cannibalization of legacy mobile ARPU, (4) satellite-direct-to-device commoditizing the basic-coverage layer (Starlink/T-Mobile partnership cuts both ways).

Moat verdict: NARROW.

Management & Capital Allocation

Mike Sievert (CEO since April 2020) inherited the Sprint merger from John Legere and has executed the integration. Srini Gopalan took over as CEO in November 2024. Capital allocation across the five Buffett choices:

Reinvest in operations. TMUS spends $9–12B/year in capex, primarily on 5G network buildout and tower densification. The ROIIC of 47.1% over 5 years is the single strongest data point in the entire scorecard — it is well above the 6.87% 10-year average ROIC, indicating that incremental capital deployed during the Sprint integration earned far above cost of capital. Caveat: this number is heavily flattered by synergy capture that will not repeat. Going forward, incremental capex should earn closer to long-run ROIC of 7–10%, which is fine but not extraordinary.

Acquisitions. Sprint merger (closed April 2020) was the company-defining capital allocation decision and worked. More recent: announced acquisition of US Cellular's wireless operations and selected spectrum (~$4.4B), and the Lumos fiber JV with EQT. The US Cellular deal adds rural subscribers and 600 MHz/2.5 GHz spectrum at reasonable multiples. The Lumos fiber move is a strategic hedge but a departure from circle of competence — fiber has different economics than wireless and a much weaker moat.

Debt. Net debt/EBITDA of -0.04 is essentially zero at EBITDA level, which is genuinely unusual for a US telco (VZ runs ~2.5x, T runs ~3x). This includes spectrum-financing SPEs which obscure some leverage; underlying gross debt is real. Still, the balance sheet is the strongest of the Big Three.

Buybacks. TMUS authorized $14B in 2023 and added another large authorization in 2024. The critical Buffett question: at what P/IV are they buying? Stock averaged $145–$210 over the buyback period; if base IV is even $400 (half the scorer's $782), buybacks were at ~40% of IV — accretive. If true IV is closer to the price (i.e., the scorer's IV is wrong because maintenance capex is understated), then buybacks were at fair value, neither accretive nor dilutive. Share count change of +3.88% over 10 years is mediocre — they have offset rather than reduced the count.

Dividends. Initiated dividends in late 2023 ($0.65/share quarterly), now ~$3.40/share annualized. Modest payout ratio. Reasonable signal of capital confidence; not a key value driver.

Communication. Sievert's investor letters were clear and operational. The Un-carrier branding under Legere/Sievert has been more disciplined and less promotional than ATT or VZ communications. Recent communications about FWA TAM have been more aggressive than I would prefer — they extrapolate net adds rather than discuss long-term ARPU sustainability.

Red flags: (1) share count up 3.88% over 10y means net dilution despite buybacks; (2) executive comp loaded toward stock-based comp which inflates GAAP earnings vs cash; (3) management bonus on "adjusted" metrics that exclude integration costs has had ~10 years to do so. Buffett's GEICO comp plan — based on only "growth in voluntary auto policies and underwriting profitability" [in canon excerpt 2] — is a higher standard than TMUS uses.

Capital allocator: B.

Industry Structure

US wireless is a textbook stable oligopoly with three national carriers (Verizon, AT&T, T-Mobile) plus structurally smaller players (cable MVNOs, US Cellular pre-acquisition, DISH/EchoStar in distress). Porter's Five Forces:

Threat of new entrants — LOW. The capital cost of building a national wireless network is north of $50B and takes a decade. Spectrum is finite and held by incumbents; the FCC controls new licenses. DISH's failed attempt at a fourth national network is the live evidence. Foreign entry (Vodafone, Deutsche Telekom which already owns ~50% of TMUS) is structurally unlikely. New entrants come as MVNOs (Mint, Visible, Xfinity Mobile) — but these are price-takers riding incumbent infrastructure, not real entrants.

Bargaining power of buyers — MEDIUM. Consumers can switch with number portability and have visible pricing. However, switching is annoying (device transfers, family plans, autopay discounts). Enterprise buyers have more leverage. Cable MVNOs have shifted some pricing power to the consumer at the low end.

Bargaining power of suppliers — LOW-MEDIUM. Apple and Samsung (devices) have meaningful power — they extract subsidies. Tower companies (American Tower, Crown Castle, SBA) have modest leverage but TMUS has been actively reducing tower lease cost via the MetroPCS legacy assets and Sprint towers. Spectrum suppliers are governments, which is a different game (regulatory rather than commercial).

Threat of substitutes — MEDIUM and rising. This is the most important force to watch. (a) Cable broadband already substitutes for in-home wireless. (b) WiFi/satellite is encroaching on basic mobile coverage in rural areas (Starlink/SpaceX, Apple emergency SOS). (c) FWA is itself a substitute that TMUS is exploiting offensively, but VZ is doing the same. (d) Long term, satellite-direct-to-device could commoditize coverage entirely; TMUS's Starlink partnership is a hedge but also evidence that the substitute is real.

Rivalry among existing competitors — MEDIUM. Three players, similar capabilities, similar capex burdens, asymmetric strengths (TMUS in 5G mid-band; VZ in postpaid premium; T in fiber+wireless bundling). Pricing has been broadly rational for a decade post-MetroPCS/Sprint consolidation. Risk: a pricing war if subscriber growth saturates. The introduction of Boost (DISH/EchoStar) as a fourth brand briefly threatened pricing discipline; its decline reduces this risk.

Value pool location and trajectory. Most of the industry value sits with the carriers, not towers or device makers (relative to the value flowing through). Apple and tower REITs capture meaningful but bounded slices. Within the carrier value pool, TMUS has been gaining share for a decade. The pool is growing modestly (mid-single-digit revenue) — the real growth is shifting from voice/data to FWA broadband and IoT/connected vehicle. The pool is shifting toward providers with the most mid-band 5G spectrum, which favors TMUS.

Industry Verdict: Good.

Inversion (Bear Case)

Bear case. I am a short seller. Here is why TMUS at $196 is a value trap.

The single event that kills this. A pricing war, triggered by a postpaid net add stall. The bull narrative requires postpaid phone net adds to remain industry-leading at +3M/year. But US wireless penetration is north of 110% of the population (multi-device households) and the smartphone replacement cycle is lengthening. When net adds stall — and they will — TMUS will face the same arithmetic that killed Sprint: high fixed costs, modest pricing power, and a competitor (cable MVNOs at the low end, VZ premium plans at the high end) willing to compete on price. The first real ARPU declines will trigger multiple compression from a PE TTM of 18.83 down to industry-historical 10–12x for mature, low-growth telcos. That alone is 35% downside.

Why the moat is narrower than bulls think. The 47.1% ROIIC of the last five years is one-time merger synergy capture, not a sustainable cost advantage. The 10-year average ROIC is 6.87% — barely above any reasonable cost of capital. Spectrum, the supposed structural intangible, has a clock: 2.5 GHz licenses must be utilized and renewed; the FCC can re-auction; competitors can buy. Buffett's GEICO had a cost advantage that grew with scale year after year for 70 years [1, 2]. TMUS's cost advantage has plateaued — the synergies are largely captured, and the cable MVNOs have built a structural cost advantage of their own (zero customer acquisition cost via existing broadband relationships).

Why management is worse than it appears. Three signals. (1) Share count up 3.88% over 10 years despite massive buybacks — management has been giving away equity faster than they retire it via SBC. (2) The Lumos fiber JV is a circle-of-competence violation; fiber's terminal economics are weaker than wireless and TMUS has no edge there. (3) Aggressive use of "adjusted" metrics that exclude integration costs, restructuring, and SBC. Buffett's GEICO comp plan rewarded only voluntary policy growth and seasoned underwriting profit — clean, non-adjustable [canon 2]. TMUS rewards adjusted EBITDA and TSR, both manipulable. (4) Recent price hikes on grandfathered plans, while accretive in the short run, signal the company is now extracting from the installed base — historically a late-cycle behavior.

What bulls are extrapolating that won't hold. (a) FWA TAM. Bulls model TMUS reaching 12–15M FWA subs at $50 ARPU. The reality: FWA is capacity-constrained (it cannibalizes mobile capacity in dense markets), and as soon as fiber overbuilders and Starlink reach scale, FWA churn will rise. (b) Buyback math. Bulls assume TMUS keeps buying at current levels; if FCF compresses, the buyback shrinks. (c) Synergy run-rate. Sprint synergies are largely done; the next $1B of synergy is much harder to find. (d) The reverse DCF implied growth of -3.4% is NOT the bear case; it is a normalization of a very mature business. The bear case is that growth turns truly negative, not just below-zero on a real basis.

Valuation trap (multiple compression / regime change). Three layers of trap. (1) The scorer flags maintenance capex uncertainty with >50% spread — this is enormous. If true maintenance capex is closer to gross capex of ~$10B/year rather than the assumed lower number, owner earnings of $17.4B compress to $8–10B and IV halves. (2) The base CAGR was clamped from 30.9% to 14.0% — the unclamped number was clearly absurd, but 14% is still aggressive for a mature wireless oligopoly; sustainable long-run growth is more like 3–5%. At 4% growth and a 12x multiple, IV is closer to $250–300, not $782. (3) Regime change: if a Trump or Vance FCC blocks the US Cellular deal, or if a future administration revisits the Sprint merger remedies, multiple compression accelerates. PE compression from 18.83 to a steady-state 11x telecom multiple gets you to $115. Add a 25% FCF reduction from true maintenance capex and you get to $85.

If I am right, the stock could be worth $90–110 within 3 years.

Lollapalooza Bias Check

Biases active in me right now as the analyst:

Anchoring. The scorer hands me a base IV of $782 against a current price of $196. That ratio of 0.25 is so extreme that my first instinct is to call this a generational bargain. I must consciously discount this anchor, especially because the scorer itself flags two warnings: maintenance capex uncertainty >50% and base CAGR clamped from 30.9% to 14.0%. The unanchored question is: at $196, am I getting a low-leverage scaled wireless oligopolist at a reasonable multiple of normalized FCF? The answer there is much more nuanced — probably yes, but the margin of safety is smaller than the 4x ratio suggests.

Authority bias. TMUS is in the S&P 500, covered by 30+ analysts, owned by every large index fund. The implicit authority signal is "this is a quality compounder." I am letting that override what the underlying numbers say about ROIC at 6.87% — barely above WACC. A no-name microcap with a 6.87% 10-year ROIC would be flagged as low-quality. I should apply the same standard.

Recency. The 47.1% ROIIC of the last five years is recent and salient. It dominates my mental model of management quality. But the 10-year average ROIC of 6.87% and the 10-year share count change of +3.88% are the longer signals. Recency is causing me to overweight the merger synergy window.

Confirmation. I want to find a thesis that justifies a Buy — Buffett-style analyses are more satisfying when they end in a Buy. I should look harder for the disconfirming evidence: maintenance capex, Lumos fiber, share count drift, FWA cannibalization risk, and the cable MVNO take rate.

Incentive bias on management's data. Sievert and Gopalan are paid on adjusted EBITDA and TSR. The numbers they emphasize in earnings calls are exactly the numbers that pay them. I should weight cash flow and share count more heavily than EBITDA and net adds.

Not active here: social proof (I'm not following crowd because the crowd is divided on TMUS), commitment (no prior position to defend), deprival super-reaction (no specific bargain I'd lose by passing).

Net effect. The Lollapalooza tilt is toward Buy. Three biases (anchoring on px/IV, authority of S&P 500 status, recency of merger synergies) push the same way. That is the warning sign Munger flags — when multiple biases compound in the same direction, decision quality collapses. The discipline is to require an extra margin of safety beyond what the numbers headline.

10-Year Outlook

Will TMUS in 2036 look like TMUS in 2026, only larger and more profitable per customer?

Same fundamental business model? Probably yes. US wireless will still be a three-firm oligopoly. Voice/data will still be commoditized. Spectrum will still be the binding constraint. The Un-carrier branding is a tactic, not a model — the underlying business is selling subsidized airtime to a subscriber base.

Customer base larger? Modestly. US population grows ~0.4%/year. Connected devices per capita grow faster — IoT, connected cars, wearables. Realistic 10-year subscriber growth is 1–3%/year, mostly from device proliferation rather than new humans.

Profit per customer higher? Uncertain. Bull case: 5G monetization (FWA, network slicing, IoT premium) drives ARPU up 2–3%/year. Bear case: cable MVNO competition compresses ARPU, especially at the low end. My base case: ARPU roughly flat in real terms, slightly up in nominal terms.

Moat wider or narrower? Slightly narrower. The Sprint merger advantage saturates. The 2.5 GHz mid-band edge attenuates as VZ and T close the gap with their own mid-band buildouts. Cable MVNOs and satellite-direct-to-device commoditize the bottom and the rural edges respectively. The three-firm structure remains, but TMUS's relative position within it stops improving.

Single biggest threat? A 4th competitor enabled by satellite economics — Starlink direct-to-device at scale plus Amazon Kuiper plus a foreign entrant could break the three-firm structure within 10 years. Probability: 20–30%. Impact if it happens: catastrophic for incumbent multiples.

Capital intensity? Roughly the same. 5G to 6G transition will demand another major capex cycle around 2030–2032.

Confidence assessment. I can predict the structure with reasonable confidence, but I cannot predict ARPU evolution, FWA take rates, or the satellite-disruption probability with high confidence. The maintenance capex question — flagged by the scorer with >50% spread — is the deepest source of uncertainty.

CONFIDENCE: medium