Built by Grok 4.6 High

Crown Castle Inc. reported third-quarter 2026 results on Wednesday, October 21, 2026. This analysis was prepared on Monday, October 26, 2026. The quarter ended September 30, 2026. The fiscal year ends December 31, so this print is 2026Q3. The next print is Q4 2026, expected February 3, 2027.

CCI | Earnings Review — Q3 2026

Verdict: decelerating. Reported net revenue growth left a four-quarter band near −4% to −5% and printed −7.2% (−228 bps versus the prior year's third quarter, compared with 2026Q2's −4.9%). That break is services, not the tower book. Site rental was −4.5%, only 48 bps worse than 2026Q2's −4.1%. Sprint-adjusted organic growth was +3.5%, −170 bps versus +5.2% in 2025Q3. AFFO dollars slowed from +9.9% to +8.2%. AFFO per share stayed at +10.7% (−7 bps versus Q2's +10.8%). The lease book is not re-accelerating. The cash line is still growing, and the growth is interest and share count, not leasing.

The print is mixed, and the revenue streak broke. Net revenues of $995 million missed FMP's $1,011.9 million by −$16.9 million / −1.7%, and sat below the 10-analyst low of $1,001.9 million. That ends 11 straight revenue beats. Diluted EPS of $0.67 missed FMP $0.687 by −2.5%, also below the six-analyst low. Site rental of $966 million was about −0.2% versus a Visible Alpha site-rental figure of $968.41 million (carried Oct. 21; not a direct pull). Services of $29 million were −51.7% versus $60 million. That is where the FMP gap sits. Adjusted EBITDA of $673 million was in line with a Visible Alpha $672.2 million figure and −6.3% versus $718 million. AFFO per share of $1.24 beat Zacks $1.20 by +3.3% and the carried Visible Alpha $1.21 by +2.5%. The AFFO beat narrowed from Q2's +13% Zacks beat. Do not score FMP's stored revenue actual of $966 million. That field is site rental, not net revenue.

The guide went up. Demand did not. Versus July 22, both ends of site rental rose $20 million (midpoint $3,875 million), adjusted EBITDA rose $25 million (midpoint $2,715 million), and AFFO rose $5 million (midpoint $1,980 million, $4.60 per share). Sunit Patel split it: the $20 million is Q4 straight-line from a lease-term extension. Excluding that customer transaction, EBITDA and AFFO are each up only $5 million. Organic contribution excluding Sprint and DISH is unchanged at $135 million, or 3.4%. Core leasing is still $65 million in the bridge, but management now expects the bottom of the $60–70 million range. There is no FY2027 outlook. FMP's 2027 total revenue is +1.7%. That does not underwrite a DISH lap of −$220 million.

Tone. October is a spectrum and SpaceX call, not a demand raise. Hillabrant repeated that 2026 is the organic low for Crown Castle and for the industry, and then would not split 2027 into leasing, escalators, and churn. The Q&A tally is 14 well answered / 10 deflected. The answers describe what is already in the supplement. The deflections are SpaceX, the escrow split, the 2027 organic mix, the extension's term, and when upper C-band hits towers.

Three contradictions. February's prepared remarks put DISH churn at $220 million and, one paragraph later, $120 million. Only $220 million ties the billings bridge, the guide is still −$220 million, and nine-month terminations are already about −$159 million. Two hundred basis points of margin was "over the next year" in July and "by 2030" in October, both labeled unchanged. Core leasing is aimed at the bottom of $60–70 million while the bridge still uses the $65 million midpoint inside an unchanged $135 million organic total.

What this is not. Not a price note. No stock price, market cap, or multiple is used. Bloomberg and a direct Visible Alpha feed were not connected. Consensus is FMP, pulled 2026-10-26, plus the labeled Zacks and carried Visible Alpha bars for this print only. SharePoint, OneNote, Outlook, and Excel were unavailable and were skipped. The fiber and small-cell sale closed May 1, 2026. Clean tower comparisons start in 2025Q3 against the restated 2024 base. 2025Q1 and 2025Q2 versus as-reported 2024 are a perimeter cliff, not demand.

Fiber left continuing operations in 2025. The 2024Q4 column below is still the combined company. 2025Q1 and 2025Q2 revenue YoY versus that base is marked n/m. From 2025Q3 the revenue YoY uses the restated towers base. From 2026Q1 both years are continuing operations. Adjusted EBITDA, EPS, and AFFO have no restated 2024 vintage, so those YoY rates stay n/m until 2026. Gross margin here is calculated: net revenue minus site-rental cost of operations minus services cost of operations. Crown Castle does not publish a GAAP gross-profit line.

Consolidated, eight quarters

Clean revenue YoY is −4.3%, −4.2%, −4.8%, −4.9%, −7.2%. Of the $77 million decline versus 2025Q3, site rental is $46 million and services is $31 million. Services went +11.1% → +8.2% → −2.0% → −21.2% → −51.7%. Site-rental YoY on the clean window is −5.1% → −4.8% → −4.9% → −4.1% → −4.5%. The decline narrowed into Q2 and gave 48 bps back. It is not a new site-rental low, and it is not an improvement on the prior print.

EBITDA dollars, where both years are towers, went −6.5%, −4.3%, −6.3%. The margin went the other way: −122 bps, +45 bps, +66 bps, to 67.6%. That wider consolidated margin is mix. Services is the lower-margin line, so a smaller services book cushions the total. Consolidated gross margin is 73.8%, only −11 bps. Site-rental gross margin is 74.3%, flat versus Q2 and −97 bps year over year, with site cost stuck near $248 million versus $250 million. The towers path is 76.3%, 75.1%, 75.3%, 75.3%, 75.0%, 74.3%, 74.3%. EBITDA divided by site rental is a different ratio, about 69.7% versus 70.9% a year ago, roughly −128 bps. Site gross margin compressed. Consolidated EBITDA margin did not.

AFFO dollars: −6.9%, +9.9%, +8.2%. Per share: −7.3%, +10.8%, +10.7%. The level is a new high for the towers window ($446 million → $488 million → $530 million). Do not read the slower percent as a rollover, and do not read the positive percent as leasing. Interest expense was $177 million versus $247 million, −28% YoY. That is fiber-sale deleveraging.

Headline EPS of −9.5% compares a clean 2026Q3 quarter with a 2025Q3 quarter that still included fiber. Continuing-operations diluted EPS was $0.67 versus $0.64, +4.7%, and that rate decelerated 843 bps from Q2's +13.1% ($0.69 versus $0.61). The +5,770 bps "acceleration" in headline EPS versus Q2's −67.2% is the $0.22 discontinued-operations air pocket. It is not an earnings inflection.

Organic bridge — the operating tell

Sprint-adjusted organic held a 4.6–5.2% band for two years and then stepped down. Q3 versus Q3 is the comparison that matters: +3.5% versus +5.2%, −170 bps. The path of that YoY rate in 2026 is +3.1%, +3.9%, +3.5%. Q3 gave back 40 bps of the Q2 bounce in the rate. DISH-out-of-base moved the same way: +3.3%, +4.2%, +3.8%, and the company puts Q3 2025 at 4.1% on that definition (−30 bps).

Escalators do not move. They are $25 million versus $24 million. Non-renewals are −$7 million, the same as −$7 million a year ago. Core leasing on the Daloopa series is $16 million versus $33 million. The quarterly level has been $15 million, $15 million, $16 million. The company's own supplement compares this quarter's $16 million with $20 million in Q3 2025, not with the $33 million series. Use the supplement when quoting management's year-ago core-leasing pair. Either pair is a softer new-leasing year, not a re-acceleration. The 2025 sprint-adjusted dollars ($52 million) are not a clean pair for the 2026 "ex Sprint and DISH" $35 million, because DISH was still billing in 2025. Use the rate.

Site rental fell −$46 million versus last year. The tagged pieces do not add to that figure. Organic ex Sprint and DISH +$35 million, DISH −$61 million, the walk's Sprint tag of +$5 million, straight-line −$14 million, and prepaid-rent amortization −$3 million sum to −$38 million. The organic-bridge Sprint figure is the other sign, −$5 million, and that list sums to −$48 million. The residual is $2 million to $8 million depending on the Sprint sign. Headline organic contribution to site rental revenue is −$31 million, or −3.1%.

Site-rental billings were $965 million versus $996 million, −3.1%, a wider decline than Q2's −1.7% (about −145 bps on the YoY rate). Cash billings deteriorated faster than the clean organic rate. Towers stayed at 40 thousand. Tenants per tower fell from 2.4 to 2.0. That drop was already in the first quarter. It is the DISH exit, not a new tower build.

Five-year annual history

FY2024 is shown as reported (fiber still in) and restated towers-only. The only clean annual towers comparison is FY2025 versus restated FY2024. AFFO, EPS, and adjusted EBITDA have no restated FY2024, so their FY2025 YoY against the combined company is a perimeter change, not an organic collapse.

Combined-company revenue already rolled over before the sale: +8.6% → +10.2% → −0.1% → −5.9%. The towers year continued it, −4.4%, with site rental −5.1%. Services was still growing on the restated base in FY2025 (+12.0%). The services collapse is a 2026 quarterly event. The new full-year site-rental midpoint of $3,875 million is −4.3% versus FY2025 $4,049 million. It embeds the same mid-single-digit decline this quarter already printed.

One-line verdict: Decelerating. Services caused the reported-revenue break. Site rental stayed a mid-single-digit decline. Organic ex Sprint and DISH slowed 170 bps year over year. AFFO is still growing, more slowly in dollars, and not because leasing re-accelerated.

Pattern: Mixed. Revenue had been a consistent beater and just missed. GAAP EPS is mixed and the clean part of the series turned down in 2026. AFFO per share beat again, by less, and the beat is below EBITDA. Signed revenue-surprise magnitude deteriorated. The miss is not wider than the old beat band. What changed is the sign, after eleven beats.

FMP revenue estimate $1,011.881 million (10 estimates, range $1,001.863–$1,021.900 million). FMP EPS $0.687 (6 estimates, range $0.676–$0.693). Scored revenue is net revenues, not FMP's stored $966 million actual.

FMP's EBITDA average of $697.5 million is not adjusted EBITDA and is not scored. Continuing-operations diluted EPS is also $0.67. Fiber is out, so GAAP and continuing EPS match this quarter. A headline "FFO beat by $0.18" implies a ~$1.06 bar. That is not Zacks and not the $1.21 Visible Alpha figure. It is not used.

Management never said "miss." Patel's bridge explains the year-over-year site-rental decline, not the $17 million gap to FMP net revenue. He called services margin $15 million on the call and said the fourth quarter should pick up, "probably not at the levels we might have guided to, but close enough." Daloopa's adjusted services gross margin is $17 million, −45% versus $31 million. The $15 million is transcript only. Hillabrant tied the shortfall to 600 MHz timing: the fourth-quarter uptick was pushed, and applications are only now flowing. That is a miss they had telegraphed, not evidence that activity re-accelerated. Core leasing halved on the Daloopa series inside an in-line site-rental print. The AFFO beat is interest, cash, and share count. Nobody on the call defended EPS.

Heatmap — last eight quarters

2026Q3 is highlighted. Revenue is as-reported net revenue versus FMP. 2024Q4 uses the original combined-company print, not the later towers restatement. EPS is diluted EPS versus FMP, except 2026Q2, where FMP's stored actual is continuing-operations EPS.

† 2026Q2 FMP actual of $0.69 matches continuing-operations EPS of $0.69, not total diluted EPS of $0.22. Against total GAAP that quarter is a miss of about −44%. The +77% cell is what FMP's own actual produces. It is excluded from the EPS magnitude read. The Q2 AFFO/share +13% uses the Zacks $1.00 bar recorded in the August review versus $1.13. It was not re-fetched.

Last twelve quarters (2023Q4–2026Q3): revenue 11/12 = 91.7%, diluted EPS on the FMP-matched line 8/12 = 66.7%. Last four: revenue 3/4 = 75%, EPS 2/4 = 50%. If 2026Q2 is scored on total GAAP, EPS is 7/12 and 1/4. Either way EPS is not a consistent beater. The 2024Q4 and 2025Q1 EPS misses are fiber-exit losses. The large 2025 EPS beats are an estimate that was still about $0.52–$0.55 against EPS of about $0.67–$0.74. In 2026 the estimate caught up and the company missed twice.

Mean signed revenue surprise: prior eight quarters +1.12%, last twelve +0.94%, last four +0.56%. Absolute surprise did not blow out. Direction did. AFFO-per-share beat magnitude, on the only two clean published bars, narrowed from about +13% to +3.3%.

Crown Castle does not guide a quarter, and it does not guide gross margin, operating margin, or an EBITDA margin. There is no FY2027 number. Hillabrant said he was not prepared to share 2027 guidance. Every "Q4" company figure below is the full-year range minus nine-month actuals. It is a residual, not a quarterly guide.

Where the full-year midpoint has moved

October bounds, linked: site rental $3,853 million to $3,898 million; adjusted EBITDA $2,690 million to $2,740 million; AFFO $1,955 million to $2,005 million; AFFO per share $4.54 to $4.66. Billings stayed $3,805 million to $3,835 million. Site-rental cost of operations fell $5 million at both ends, to $963 million–$1,008 million. Interest expense was unchanged at $787 million–$832 million. Straight-line in the outlook improved from −$70 million to −$50 million. That +$20 million is the entire site-rental revenue raise.

The $275 million customer payment is transcript, not a guidance tag. Patel described a residual economic interest, a term extension, and a churn cap on a subset of towers. Expected Q4 billings lift about $4 million ($16 million annualized). Extra straight-line of $20 million lands in Q4, which is why revenue guidance moved and billings guidance did not. Annualized AFFO accretion about $4 million before the amortization benefit. He would not confirm a five-year term or a 3% escalator. Churn, ex Sprint, stays 50–100 bps. The subset is too small to change that.

Services gross margin stayed $70–100 million in the filing. Patel's "$85 million" is that midpoint. Daloopa's 2026Q3 tags are reversed: the low series prints $100 million and the high series prints $70 million. July was in the right order, $70 million to $100 million. Use the filing. Do not read the tag reversal as a cut or a raise.

What Q4 has to do, at the new midpoint

Nine-month site rental is $2,894 million, adjusted EBITDA $2,023 million, AFFO $1,464 million, organic ex Sprint and DISH $103 million. The "versus prior guide" gap is the October raise landing entirely in Q4, because the same nine-month actuals are subtracted from both ranges.

The high end of the site-rental residual, about $1,004 million, is still −1.5% versus 2025Q4. The entire Q4 range is below last year's quarter. FMP's Q4 total-revenue average of $1,019.7 million is about $38 million above the $981.5 million site-rental residual. That gap is roughly a quarter of services, which the company does not guide as revenue. FMP Q4 EBITDA of $743.8 million is the wrong series: FMP's Q2 EBITDA was $572 million against reported adjusted EBITDA of $675 million. Do not read FMP as the street sitting above the EBITDA guide.

Full-year versus FY2025 at the new midpoint: site rental −4.3% (nine months already −4.5%), adjusted EBITDA −5.2% versus $2,863 million, AFFO +4.0% versus $1,904 million. Diluted-EPS midpoint $2.08 versus FY2025 $1.01 is the fiber transition, not tower acceleration. FMP FY2026 EPS of $1.932 sits inside a $1.76–$2.41 guide that is $0.65 wide. That is not a precise EPS guide.

FY2027. Not issued. The one tagged 2027 figure is contracted site rental billings of $3,937 million, up from $3,913 million at the Q2 print. Against the FY2026 billings midpoint of $3,820 million, billings already under contract are about +3.1%. That is a floor in the lease file, not a revenue guide. FMP FY2027 total revenue of $4.101 billion is +1.7% versus FMP's $4.034 billion for 2026, and only about +5% by 2029. FMP's +48% EPS step into 2027 is a depressed 2026 base, not a leasing forecast. The reiterated "$2.1 billion of AFFO across 2H 2026 and 1H 2027" is a sentence, not a range. Q3 AFFO plus the Q4 residual is about $1,046 million for the second half, which would leave about $1,054 million for the first half of 2027 if the sentence is taken literally. That illustration is not a guide.

Tone versus July. July refused to raise EBITDA because services were weak. October raises EBITDA with a non-cash item and spends the hour on spectrum, SpaceX, and edge. The industry language got longer. The 2026 leasing language got shorter: toward the bottom of a range they did not cut. They raised the accounting guide and talked the demand guide down.

Reported revenue YoY is a cliff, then a tight negative band, then a fresh low. The cliff is fiber leaving the presented P&L. The fresh low is real in the reported rate and is mostly services.

† Fiber cliff. Continuing operations versus as-reported 2024, which still includes fiber. Do not read −35% as tower demand. ‡ First clean towers comparisons, versus restated 2024. The +3,052 bps is the cliff ending.

Inflections worth keeping

1. 2025Q1 — accounting. Presented revenue YoY −35.3%. Not durable. It reverses as soon as the base is restated.
2. 2025Q3 — base reset. First clean towers YoY, −4.3%. Sets the band. Not demand returning.
3. 2025Q4 — least-negative clean print. Revenue YoY −4.2%. Did not hold.
4. 2026Q3 — reported-revenue break. −7.2%, −228 bps. New low of the clean window. Site rental only −48 bps. Services did the rest, and management calls that timing (600 MHz pushed into Q4). The site-rental piece is the run-rate.
5. 2026Q2 — GAAP EPS air pocket. Diluted EPS −67.2% on $0.22. Continuing EPS that quarter was $0.69. Not durable.
6. 2026Q3 — EPS "acceleration." Headline −9.5% and +5,770 bps versus the air pocket. Do not use. Continuing EPS decelerated 843 bps to +4.7%. AFFO per share did not decelerate in any meaningful way (−7 bps).

Over the clean window, site rental YoY improved from −5.1% to −4.5%. The just-reported quarter gave back the best clean site-rental rate (−4.1% in Q2). Services decelerated −1,915 bps and then −3,051 bps. A line that is about 3% of revenue produced about 40% of the dollar decline and almost all of the growth-rate break.

The organic rate that drops DISH from the prior-year base was +3.3%, +4.2%, +3.8%. Underlying tower leasing is still growing, and it is slower than last year. Nothing in the eight-quarter print shows reported revenue or headline EPS re-accelerating. The lines that would have to turn first — site rental YoY, the ex-DISH organic rate, and AFFO per share — are respectively slightly worse, slightly slower, and stable.

The catalysts that can move the next two prints are not in this quarter's AFFO beat. Q4 site rental is still guided down about 4% year over year, and FMP's +1.7% 2027 revenue does not underwrite the DISH lap management calls the trough. Next print: about February 3, 2027.

Contracted remainder-of-year billings are $972 million, essentially the whole Q4 billings midpoint. Almost no uncontracted Q4 billing is left. Discretionary capex this quarter was $58 million versus $36 million, +61%, and it is ground-lease buyouts.

The repetition to weigh is the trough. They have said it the way they said it in July: 2026, then better, without a 2027 organic number. FMP's published revenue path is already +1.7% in 2027, the first positive year, then +2.1% in 2028 and +5.3% in 2029. From the 2026 FMP revenue of $4.034 billion to the 2029 figure of $4.409 billion is about +9%. That path is a slow positive year, in the neighborhood of contracted 2027 billings at about +3.1%. It does not underwrite a $220 million DISH lap, and it does not underwrite SpaceX or upper C-band, because management did not number either. That is the setup into February. It is not a reason to expect Q4 site rental to grow.

Ten analysts. Nobody was cut off. The hour was not a re-litigation of the print. It was an attempt to turn two sentences into units: "2026 is the organic low," and "spectrum transactions are a catalyst." Management answered the accounting. They refused every pin that would make the bullish language investable this year.

Tally: 14 well answered, 10 deflected.

The deflections are one cluster. SpaceX will not be sized: no chipset, no escrow economics, no MLA, no legal path to the radios, no site count, no dollar, and a direct "has SpaceX called?" declined twice. 2027 will not be bridged, which leaves a −170 bps organic deceleration and a core-leasing shortfall unexplained in forward terms. The extension will not be parameterized, so a model cannot tell how many years the $20 million of Q4 straight-line is spreading. Edge will not be classified between GPU and other workloads. At 100 sites against about 40,000 towers, that question is about whether a future line of business exists, not about this quarter's AFFO.

The pattern is a team that is specific when the fact is already in the supplement, and unwilling to be specific when the fact would commit 2027 leasing or a named new tenant. The raised guide is real. Organic dollars were held. The lines that were raised are straight-line revenue, a $5 million cost benefit, and the capex to buy more ground leases.

Three document contradictions. Two are high. One is medium. Labeled guidance updates are not in this list. Two organic percentages on two denominators (3.4% and 3.6%) are the same $135 million and are not a contradiction. Leverage of 6.5x versus 6.3x is the same stack with and without the new straight-line.

This call is spectrum, a possible new terrestrial tenant, and edge. It is not rates, CPI, wages, tariffs, or the consumer. Versus July 22, management dropped the layoff and carrier-reorg explanation for soft services and dropped the Europe-versus-U.S. comparison. The new stack is: 2026 is the industry organic low, even as core leasing sits at the bottom of the range; 600 MHz is slower than hoped; SpaceX is discussed as a potential tenant; edge is a named trial and still immaterial.

Rates. Patel: quarter-end debt is 97% fixed, weighted-average maturity seven years, leverage 6.5x, or 6.3x if the Q4 straight-line is included. The interest-expense range was not changed. No Fed path. Neutral for the demand thesis. Incremental cash use is land under towers — discretionary capex midpoint $240 million, $200 million net of about $40 million of prepaid-rent additions — not a buyback and not a rate-cut story. Those leverage and fixed-rate figures are the call, not a new Daloopa tag.

Inflation. Absent. Do not read a CPI print into this quarter. The margin path is self-help: $65 million of annualized operating-cost savings, about $20 million of EBITDA and roughly 50 basis points since the original guide, and still more than 200 basis points by 2030 versus 2026. The land-cost gap versus competitors was not given a percent and the peers were not named. On July 22 he had named American Tower and SBA and a roughly 11% gap. The catch-up target is unchanged. The names were omitted.

End demand. No handset sell-through. Hillabrant said mobile data shows no signs of stopping, cited an unnamed carrier publication for payload doubling over five years, and relayed three carrier CTOs on a mix shift from about 90% downlink toward something closer to 50/50. Patel agreed AI-agent traffic is a growth driver and that it means fiber to the tower and more edge compute, not only a software upgrade. He did not say the orders are in hand. Same direction as July, weaker attribution (Ericsson's GB-per-month path was not repeated), and more specific on uplink. Positive on a multi-year lag for U.S. towercos and radio OEMs. Not a 2026 leasing beat.

Industry. "2026 is the low point for organic growth, in our business and across the industry" is an explicit sector call. It is an assertion. The organic dollar was not raised. If he is right, American Tower and SBA Communications are in the same trough, with the same spectrum and SpaceX optionality. Neither ticker was spoken on October 21. AMT reports October 27. That call is the peer test of the industry-trough sentence, not a CCI revenue line.

Customers, a possible new tenant, and everyone else

AT&T. Spectrum is owned — Hillabrant characterized the 600 MHz and 3.45 GHz outlay as $23 billion, which is his figure, not a Crown Castle fundamental — and deployment is behind CCI's own hope. The customer is still selecting an OEM. 600 MHz applications are the largest piece of the expected services pickup. Mixed near term, better later for AT&T and for towers. Better for whichever radio OEM wins. None was named. Mildly better for handset makers already supporting the band. None were named.

Verizon. Named once, for the AWS-3 licenses, grouped with AT&T and SpaceX as a delayed positive for towers. No Verizon-specific leasing or capex comment. Same lag logic, less detail.

T-Mobile. No new demand signal. Hillabrant's point was operational: many of the people who built DISH came out of T-Mobile and, in his view, can move quickly if a buyer reuses that grid. Sprint remains a $5 million Q3 offset, not a new cancellation wave.

Unnamed largest customer. The $275 million residual deal was not assigned to AT&T or T-Mobile. Patel denied a deferred purchase-price structure. Better for that carrier on contract simplicity, and better for CCI on churn visibility, on a small subset of sites. Straight-line, not cash billings, is what lifted the site-rental revenue outlook.

DISH and EchoStar. Worse for DISH as a tenant: the revenue is already leaving and the gear was abandoned. Better as an asset for whoever can legally reuse 12,000-plus sites, and therefore better for CCI on time-to-market if that reuse happens. The escrow question was answered with a mediation pause, not a split. No dollar update for CCI or for other tower creditors. Do not assume the escrow is near payment.

SpaceX. The best external read-through on the call, and still uncontracted. Better for SpaceX if it colocates: 800 MHz plus a built DISH grid is a shortcut versus a four-year greenfield. Better for CCI, which skews to the top 100 markets, and for other public towercos, if the design is colo rather than "thousands and millions" of owned microcells. Mixed for AT&T, Verizon, and T-Mobile: a new terrestrial competitor is a reason for them to spend (good for towers) and a share threat (bad for them). July's "way too early" moved to "assume we are talking to all of them." That is still not a master lease. Grain was the analyst's word for the 800 MHz seller. Management never said Grain. Do not infer a Grain–CCI contract. Amazon was in the same analyst question and was not confirmed.

Available Infrastructure and power. Better for that edge partner: a public tower landlord is leaning in. Conditionally better for CCI if 100 sites become hundreds or thousands. Not better yet for the P&L. Patel said edge revenue is still small. Large campus builds still face multi-year power delays; CCI is selling 0.2–0.5 MW on power that is already at the site. The 25-site / up to 100 MW remark is optionality, not booked load. Better for whatever utility can deliver incremental power on a 9–11 month clock, if the sample generalizes. CCI's stated preference is to use power already on site.

FCC and the FAA. An auction of about 800 MHz, including about 160 MHz of upper C-band by July 2027, is a dated policy tailwind and not a 2026 organic driver. The binding constraint Hillabrant acknowledged is clearing, including altimeters. He rejected a 2030 start without offering a year. Slightly better than the bear timing case. Not a schedule.

Not named, so not promoted into this print: American Tower, SBA, Ericsson, Nokia, Samsung, Apple, Qualcomm, and the hyperscalers, except for the labeled July contrast on land cost. No Europe. No fixed-wireless sentence. No consumer brand.

What is new versus July 22. SpaceX went from too early to the central third-party debate, still with no lease. Edge went from unnamed trials to a named partner and 100-plus sites. Services softness was re-explained as 600 MHz timing, with applications starting to flow, rather than as carrier layoffs. Organic trajectory did not turn. Rates, CPI, and Europe were not part of the story.