A Reset Barely a Year Old
BCE cut its annual dividend by more than 50 percent in 2025, down to 1.75 Canadian dollars, after years of paying out more than its free cash flow could sustainably support amid a regulatory shift that opened Canadian telecom to more competition and triggered aggressive price wars. The company set a new target range: pay out 40 to 55 percent of free cash flow going forward, wide enough, in theory, to absorb a heavy capital year without forcing another cut. On August 6, 2026, less than a year and a half after that reset, BCE reported second-quarter revenue of 6.18 billion Canadian dollars, up 1.5 percent, adjusted earnings per share of 0.65 dollars, and free cash flow of more than 1 billion dollars for the quarter, down year over year because of higher capital spending. The dividend payout for the quarter landed at 55 percent of free cash flow, the very top of the range the 2025 reset was built to keep comfortably inside.
In the same results, BCE raised its full-year 2026 capital expenditure guidance by 1.3 billion Canadian dollars, almost entirely attributable to one project: a 300-megawatt AI data centre campus in Sherwood, Saskatchewan, just outside Regina, where CEO Mirko Bibic told investors piling was complete and structural steel work was underway.
What 400 Million Dollars Buys
CFO Curtis Millen told the August 6 earnings call that BCE received its first tenant payment on the Saskatchewan facility during the quarter, part of approximately 400 million dollars in expected setup fees and prepayments, with the majority of the 1.3 billion dollar capex increase falling in the second half of 2026 and first-phase operations expected in the first half of 2027. Nationally, Bell's broader AI Fabric initiative has around 335 megawatts of contracted capacity with what Bibic described as line of sight to 800 megawatts.
Bell's own figures, released when the Saskatchewan project was first announced in March 2026, project the facility will generate roughly 500 million dollars in annual revenue, 400 million in adjusted EBITDA, and more than 250 million in free cash flow once it reaches full run rate. Those numbers describe the facility after it is complete and fully leased; they say nothing about the cash consumed getting there, which is exactly what the 1.3 billion dollar 2026 capex increase and the 400 million dollars in upfront prepayments represent; money moving now, against revenue that arrives later.
The Two Names Behind the Number
Bell's own announcement in March 2026 named the facility's tenants: Cerebras Systems, the AI chip company, and CoreWeave, the GPU cloud provider, together installing roughly 10 billion dollars of computing hardware inside Bell's 1.7 billion dollar building. A project with two named anchor tenants is not unusual for infrastructure of this scale, but it means the facility's revenue case rests on the financial health of those two companies specifically, not on a diversified customer base.
CoreWeave's own second-quarter 2026 results showed roughly 35 billion dollars in total debt, up from about 25 billion dollars at the end of the first quarter, against total liabilities of 29 billion dollars and equity of only 3.9 billion, with retained earnings of negative 3.38 billion dollars. The company's 2026 capital expenditure guidance of 30 to 35 billion dollars is itself funded substantially through debt, and its balance sheet is among the most leveraged of any major technology company currently expanding AI infrastructure, even as a loan backed by its computing hardware has drawn an investment-grade rating from Moody's. BCE is not exposed to that leverage directly, but its Saskatchewan revenue case, and the prepayments already flowing into this year's free cash flow, run through a tenant carrying that balance sheet.
The Reset Was Supposed to Prevent This
None of this means the Saskatchewan project is a mistake. A 300-megawatt facility with committed tenants and hardware already being installed is a materially stronger starting position than an interconnection-queue speculation. What it means is that BCE's 2025 dividend reset, explicitly designed to build a payout range wide enough to absorb capital-intensity swings without another cut, is being tested inside its first eighteen months by precisely the kind of concentrated, single-project capital commitment that reset was meant to withstand. A 55 percent payout ratio sitting at the top of a 40 to 55 percent target range, in the same quarter that capex rose 1.3 billion dollars for one facility leaning on prepayments from a heavily leveraged counterparty, is not a crisis. It is a live test of whether the range holds.
The transferable instruction for any board or investor watching a capital-intensive infrastructure bet layered onto a recently reset payout policy: track the payout ratio's position within its stated range over the following several quarters, not just whether the dividend itself is maintained, and separate the future run-rate economics a company presents from the current-year cash flow the same project is consuming to get there. A policy reset built to survive exactly this kind of bet only proves durable once it has actually survived one.
Read next: Watch the Credit Markets, Not This Earnings Beat | Mobileye Raised Its Profit 88 Percent On A Tax Law
