Wednesday 23 September 2026

The Bank of Canada holds at 2.25% on 28 October. We think its next move is still more likely to be a cut than a hike, but not before CPI ex-gasolineCPI ex-gasoline The Consumer Price Index with gasoline removed. When it sits well below the headline rate, the inflation is coming from fuel rather than from the broad economy. has stopped climbing, so variable-rate relief stays paused through the winter.

Why we think so. Governor Macklem’s speech had two halves, and they pull in opposite directions. The recovery is real: GDP grew at a 3.3% annualised pace in the second quarter, non-energy exports rose 14.5% to their highest level since early 2025, business investment rose at an 8.8% annualised rate and unemployment edged down to 6.4%. The warning is also real: the latest U.S. tariffs cover about 5% of Canada’s goods exports to the United States, and if they stay, growth could be roughly halved in the fourth quarter, to below 1%. A tariff slowdown argues for lower rates eventually, but it arrives slowly. Inflation of 3% is here now. CPI has sat around 3% for four months, mostly because of fuel; every 10% rise in oil adds about 0.2% to CPI, and damaged refining capacity has made pump prices behave as if oil were almost US$40 higher than it is. With ex-gasoline inflation at 2.4% in August and the Bank saying the upside risks to inflation have grown, a cut now would lean against its own forecast. A tariff shock that halves growth also removes any case for following the Fed higher. That leaves a hold.

What the market is pricing. The consensus for 28 October is a hold at 2.25%, and we agree. Where we take a view is on direction afterwards: the slowdown Macklem described tilts the next move toward a cut, but only once the inflation numbers allow it.

What it means. For a household with a variable-rate mortgage, a line of credit or anything priced off prime rate, the relief of the past two years has paused, and the Bank is not promising when it resumes. Budget on today’s rate through the winter rather than on a cut. For a business that sells into the United States, 5% of national exports sounds small, but the tariffs land on a few sectors, and within those sectors the hit is the whole margin, not 5% of it. Check whether your products sit inside the tariffed categories and decide now how much of the cost can be passed on; the customer conversation is easier before the invoice than after. The broader planning question is how much income or revenue depends on a U.S. buyer, and whether the budget balances if rates stay put for another year while fuel stays expensive.

What would change our mind. A September CPI reading in mid-October with ex-gasoline inflation above 3% would put a hike on the table. A Labour Force Survey showing job losses spreading beyond the tariffed sectors, or the 5-year yield falling below 3.3%, would bring a cut forward. We would say so.

What to watch. The September CPI in mid-October, the price of oil as the UN General Assembly week unfolds, and the 28 October decision with its new Monetary Policy Report. There is no major Canadian release today.

A Trump-Pezeshkian handshake would take US$5 to US$10 off the barrel in a day, but the price that matters for a household budget is set by tanker traffic, not photographs. We expect Brent to stay between US$90 and US$105 until traffic through the Strait of Hormuz is verifiably back to normal for a sustained period.

Why we think so. Tuesday’s fall was driven by diplomacy rather than supply. Brent fell 1.9% to US$98.46 a barrel, down from nearly US$110 last week, and the November WTI contract settled down US$1.85 at US$90.52. The trigger was talk of contact at the UN: President Trump used his General Assembly address to defend the war and to accuse Iran of dragging out talks on reopening the strait until after the U.S. midterm elections, while Secretary of State Rubio said Trump remained open to meeting Iranian President Pezeshkian in New York, and Tehran has hinted at talks. Nothing physical has changed. The strait is still not open, a Saudi pipeline is closed and Houthi forces have made gains near shipping lanes. The price is still far above the roughly US$72 it traded at before the war began about seven months ago.

The bigger picture. Oil futures react within minutes to diplomatic headlines, because traders price the chance of a deal. The physical supply that households and businesses actually buy recovers far more slowly: ships must return, insurers must be satisfied and damaged infrastructure repaired. That is why a headline can knock several dollars off the price in a day while the range holds for months. In Canada, refining adds a second layer. As Governor Macklem noted, damaged refining capacity has made pump prices behave as if crude were almost US$40 higher than the quoted barrel, so even a lower crude price may not reach the pump quickly.

What it means. Canadian inflation is at 3% mostly because of fuel, and by the Bank of Canada’s rule of thumb a 10% rise in oil adds about 0.2% to CPI. For households, the fuel line of the budget should be set on today’s pump price rather than a peace-deal price, and a vehicle purchase or longer commute is best tested at today’s price plus 10%. A business that ships or delivers should check whether customer contracts allow fuel surcharges to pass through and whether its suppliers can pass theirs on. Borrowers should read high fuel prices as one reason the Bank is holding. For investors, the offset is narrow but real: Alberta’s revenues, the energy weight in the TSX and the dollar all benefit from a high barrel.

What would change our mind. Two consecutive weeks of near-normal tanker transits, or Brent holding below US$85. Either would mean the market has decided the supply regime has changed.

What to watch. Any U.S.-Iran contact at the UN General Assembly, today’s U.S. EIA weekly crude inventories (consensus a draw of 0.7 million barrels), the Canadian September CPI in mid-October and the Bank of Canada on 28 October.

AI’s first financial effect on Canadian households will be longer job searches in exposed occupations rather than mass layoffs. Over the coming year, the useful response is liquidity, not prediction.

Why we think so. Governor Macklem’s Halifax speech, citing Statistics Canada, gave the evidence in a single line. Almost one in five Canadian businesses used AI to produce goods or deliver services in the previous 12 months, roughly three times the share of two years earlier. Over the same period private-sector hiring has strengthened and unemployment has edged down to 6.4%, so AI is not showing up as job losses across the economy. It is showing up in job-finding rates, which have declined in the occupations most exposed to it. That is the pattern you would expect when firms adopt a new tool: rather than dismissing staff, they hire fewer people for the tasks the tool can now handle, and those already looking for that kind of work wait longer. Business investment rose at an 8.8% annualised rate in the second quarter as firms spent to raise productivity, which points the same way.

The bigger picture. New technologies have usually reached workers first through hiring rather than firing. Adoption spreads gradually, firms let attrition do the work, and the change becomes visible in who gets hired and how long it takes. The consumer side is moving fast too: Meta’s Muse agent, released only in the United States and Canada, recorded 2.8 million installs in 12 days and about 642,000 U.S. daily active users, nearly three times ChatGPT’s at the same point. When adoption runs this quickly, the gap between technology and hiring practices can open quickly as well.

What it means. The effect is not a layoff headline. It is a longer job search for people in administrative, customer-service, junior analytical and some technical roles, which shows up in a household budget as months without income. If AI tools already do part of your job, size the emergency fund at six months of expenses rather than three, and hold it somewhere it earns interest, such as a high-interest savings account or a short GIC ladder. Check how long group benefits or private coverage would replace income if a search ran long. The planning question is how many months the household could run on one income or none, and whether that number was set when jobs were easy to find. For a business owner, AI is both a cost and a competitive pressure; for a parent, it is a question about what a teenager should train for. None of this is a reason for alarm.

What would change our mind. Job-finding rates in AI-exposed occupations recovering over the next two Labour Force Surveys, or unemployment rising across all occupations. Either would make this an ordinary slowdown rather than a structural shift.

What to watch. The next two Labour Force Surveys, and the Bank of Canada’s 28 October decision and Monetary Policy Report for any further comment on AI and the labour market.