Open in interactive viewer → charts, metric popovers & call review

Bell's Quiet Re-Rate: AI Fabric and U.S. Fiber Eclipse the Core

A disciplined Q2 with the substance showing through — Ziply's build reset finally ramps and Canadian sovereign AI capacity starts monetizing.
BCE · Earnings Call · 2026-08-06

Doing What We Said — Literally

BCE's Q2 2026 call read almost like a scorecard recitation. Mirko Bibic opened with “Our Q2 results show continued execution against the strategy we laid out at Investor Day last year.” — Mirko Bibic · 2026-08-06 Stripped of the usual earnings-call gloss, the numbers back it up: consolidated revenue up 1.5%, adjusted EBITDA up 1% at a stable 43.8% margin, more than $1 billion of free cash flow, and net debt leverage down to ~3.7x — helped by $2.5 billion of June debt issuance and buybacks of sub-par securities. The story the company wants told is boring discipline. The interesting part is what the discipline is funding.

Ziply's "Build Reset" Stops Being a Speed Bump

The most operationally concrete news was at Ziply. After telegraphing in February and May that 2026 was a deliberate build reset — step sideways to leap forward — management showed the leap starting. “We're talking about permit submissions being up 4x from April to June.” — Mirko Bibic · 2026-08-06 State-level approvals now cover roughly 75% of the 2027 build funnel; high-level engineering is about 60% complete; contractor capacity and fiber supply are locked in. This is a genuinely company-unique theme. Permit submissions jumped to the top of BCE's own momentum ranking this quarter — it hadn't appeared in any of the prior eleven quarters of curated keywords. The contrast with the previous call is stark. In February, Bibic said “So on Ziply, we're pleased. So 2025 ended exactly as we ... pretty much exactly as we telegraphed it would.” — Mirko Bibic, President and CEO · 2026-02-05 Now the message is that the build — not demand — was the gating factor, and the build is clearing. Ziply delivered 9,600 residential fiber net adds (a post-acquisition record) with penetration tracking the investment case, even as EBITDA margin softened to 40.6% on acquisition costs and one wholesale re-rate. Curtis Millen was unapologetic about the trade: “I'll take that temporary margin pressure for future revenue growth. I think that's a very good trade.” — Curtis Millen · 2026-08-06 Curtis had flagged the shape of this at the start of the year — “we are expecting capital intensity overall to drop, but fixed dollars remaining the same” — Curtis Millen, Chief Financial Officer · 2026-02-05 — and the funded reallocation toward the U.S. is now visible in the CapEx line, up $317 million year-over-year.

AI Fabric: From Paper to Concrete

The strategic center of gravity has shifted unmistakably to AI Fabric. BCE now holds 335 megawatts of contracted capacity — about 90% of the 373 MW embedded in its Investor Day plan — and took its first tenant payment in the quarter, roughly $100 million of a $400 million setup-and-prepayment package, flowing through working capital. Saskatchewan has moved from piling to structural steel, with first-phase operations targeted for the first half of 2027 across four data halls. On the shape of CapEx, Curtis Millen pointed out that the “...vast majority of expected 2026 Saskatchewan AI data center CapEx of approximately $1.3 billion is to be incurred in the second half of the year” — Curtis Millen · 2026-08-06 — a timing shift rather than a stall, with equipment already ordered and delivery schedules on track. This is where the global context sharpens the picture. The top global keyword for the quarter is Batch Zero — the ERCOT interconnection queue that is the heartbeat of U.S. data-center land, cited this cycle by CIFR, Constellation, and Galaxy. BCE is deliberately outside that queue. Its AI capacity is Canadian, sovereign, and wired to its own national network — a different jurisdictional play on the same AI-infrastructure demand. Bibic framed the whole build around structural advantage: power, network, enterprise relationships, artificial intelligence.

We've been very, very transparent that we have line of sight to 800 megawatts, 335 megawatts already contracted. And as we contract more, we'll be funding it.

Mirko Bibic · 2026-08-06
For a company that spent 2021–2025 being defined by regulatory battles and dividend cuts, monetizing sovereign AI capacity is a genuine re-rating signal.

The Core Holds — and Media Inflects

Underneath the pivot, the telecom core is firming. Wireless postpaid churn hit 1.02% — the lowest quarterly print in three years — on 41,594 net adds, with ARPU stable ex-G7 and the monthly recurring component up 0.7%. The company is deliberately trading Net adds for quality: fewer hardware-discounted activations, more trade-ins and device-residual programs, and better product margins. Bell Media posted revenue up 8.9% and EBITDA up 3.8%, lifted by the FIFA World Cup (30.5 million Canadians reached), Crave crossing 5.1 million subscribers (+23% YoY, direct-to-consumer streaming +49%), and digital video ad revenue +39%. Same playbook, different unit: monetize owned premium assets. The narrative is now coherent — arguably for the first time since the Ziply deal. BCE has repositioned from a defensive Canadian dividend name into a dual-engine growth story — U.S. fiber build-out plus Canadian sovereign AI capacity — funded by a disciplined core.

Again, no surprises here. I'm really quite pleased with the momentum that we're starting to build.

For a telecom, "no surprises" and "momentum building" in the same sentence is a signal, not boilerplate. The risk, as always, is that build ramps slip; the evidence this quarter says they're not.