BCE Sells the Towers and Pledges the Data Centers — But Wireless ARPU Still Won't Show Up

The Tower Trade Finally Lands

BCE's third quarter was not a blowout. Revenue rose 2.2%, adjusted EBITDA 0.9%, and the margin slipped 60 basis points to 45.1%. But the call was less about the quarter than about two capital-allocation events that reframe the story: the long-awaited TowerCo monetization and a big scale-up in the AI Fabric plan. La Caisse will take a 49.9% stake in the new tower company for $2.45 billion, implying an enterprise value of about $4.91 billion — in line with recent Canadian tower precedents, CFO Curtis Millen said. BCE keeps control and consolidates the results; the deal closes in Q1 2027.

This is the answer to a question analysts have asked for at least a year. On the November 2025 call, Mirko Bibic said: “we don't have any updates to share at this time with respect to something like a potential tower sale or monetization of infrastructure... clearly, we view our infrastructure as potentially a valuable source of capital.” — Mirko Bibic, President and Chief Executive Officer · 2025-08-07 Now it's done. The proceeds are not earmarked, but the priorities are unchanged: hit 3.5x net debt leverage by end-2027 and fund growth. “$2.45 billion is providing us financial flexibility... We'll hit 3.5x net leverage, and we have financial flexibility to fund growth initiatives, including AI Fabric.” — Curtis Millen, Chief Financial Officer (CFO) · 2026-11-05 Leverage was about 3.7x at quarter-end, with nominal net debt down $680 million year-to-date to $39.6 billion. The other divestiture, Northwestel, is no longer a clean sale — Bibic said it is unlikely to be a full sale of the asset, though a transaction is still expected with the same deleveraging impact. That's a quiet change from the prior script, and the keyword record shows it: Northwestel was a named theme on the November call and has since slipped away.

AI Fabric: Bigger, but the Tape Is Cold

AI Fabric is the growth leg management wants investors to underwrite. The plan announced at the 2025 Investor Day was to monetize 73 megawatts. That has been raised to 373 megawatts, of which 335 are contracted, plus a nonbinding MOU with Saskatchewan for up to 900 megawatts of additional capacity — a potential path to 1.2 gigawatts in the province. The first 300-megawatt phase is under construction; the first data hall is targeted for H1 2027, with all four halls by end-2027. Bibic claims the business is on track to be an over $500 million EBITDA business.

With the contracts we have in hand, Bell AI Fabric is on track to be an over $500 million EBITDA business, and that value is definitely not reflected in BCE's share price today.

But the market's AI data-center complex has been de-rating. The global tape shows AI data centers down roughly 30% over both the 30-day and 90-day windows, with dozens of decliners. Analysts on the call pressed exactly that nerve: Drew McReynolds raised the neocloud debate and CoreWeave; Vince Valentini asked about tenant credit quality over 20-year contracts. Bibic's answer was discipline — a diversified customer base, credit protections, letters of credit, prepayments, and “we're not building on spec. We get a contract, then we build.” — Mirko Bibic, President and CEO · 2026-11-05 That is the right operational answer, but it also means the revenue ramp is back-end loaded into 2027-2028 — precisely when the market may be less willing to pay for distant capacity. Note also that the global keyword set for this quarter puts IPhone Duo and core deposits at the top of the market's agenda; AI infrastructure is no longer the crowd's favourite trade.

The Wireless Goalpost Moves Again

The core wireless business is where the quarter's real tension sits. Postpaid churn fell 7 basis points to 1.06%, the lowest third-quarter level since 2022. That is genuine progress. But ARPU fell 1.7%, hit by the CRTC-mandated elimination of certain connection fees and by lower-ARPU government customers. Excluding the fee change, service revenue was down about 0.5%. The company now expects ARPU growth by mid-2027 — a timeline that has slipped repeatedly. In February 2026, Bibic said: “when we came out of Black Friday in November of 2025, we thought it would be possible to show moderate ARPU growth by Q4 of 2026. And with the pricing we saw in December of 2025 and what we've seen some of our peers do in January, it might be more difficult to get there.” — Mirko Bibic, President and CEO · 2026-02-05 By August 2026, the target was three to five quarters out. Now it is mid-2027. The goalpost has moved from Q4 2026 to mid-2027 in three calls.

Management's framing has shifted too. The new phrase is profitable transactions: quality over volume, more than 100% of net adds on the Bell brand, higher billed rates on new activations (up 11% year-over-year), lower handset subsidies. The 24,000 postpaid net adds include another government sector contract; excluding that, consumer net adds were in line with peers. That is a recurring boost — the same government contract helped Q2 — and it flatters the headline. Then there is Black Friday. Bibic went out of his way to question the industry's promotional calendar: “the U.S. Thanksgiving holiday is an amazing holiday, but it's not a Canadian holiday... always wondered why we engaged so vigorously in deeply discounted Black Friday sales.” — Mirko Bibic, President and CEO · 2026-11-05 It is a signal that BCE plans to sit out the most aggressive Q4 discounting. If peers follow, it could support ARPU. If they don't, BCE risks ceding share.

What's Left

Ziply is improving at the margin. Residential fiber net adds were about 8,600; permit submissions rose 60% from Q2 and nearly 8x Q1; Fiber revenue at Ziply grew 10% year-over-year. The 3 million footprint target by end-2028 is intact, but Curtis conceded he will always want more fiber and that the build takes longer than anyone would like — “in my role, I'm never going to be happy with pace of build.” — Curtis Millen, Chief Financial Officer (CFO) · 2026-11-05 Bell Media was the cleanest beat: revenue +9.7%, EBITDA +8%, Crave subscribers +21% to 5.15 million, digital now roughly half of media revenue, a genuine transformation from the 19% digital share of 2020.

Put it together and BCE is doing what it promised on capital recycling: TowerCo is sold down, Northwestel is being reworked, leverage is heading toward 3.5x. The AI Fabric option is bigger than it was a year ago, but it is still an option — contracted megawatts, not yet contracted EBITDA — and the market's AI infrastructure tape is cold. The core wireless business is improving on churn but still searching for ARPU growth, now promised for mid-2027, and analysts like Vince Valentini are openly asking whether BCE should simply let Quebecor take gross-add share rather than chase it with discounts. The stock's re-rating likely hinges on two proofs: AI Fabric converting steel and power into cash flow, and wireless service revenue actually inflecting. Until then, the significant value Bibic keeps citing will remain a narrative, not a number.