Nvidia affects Canadians through three different channels: the U.S. equities inside their portfolios, the exchange rate translating those investments into Canadian dollars, and the businesses building or buying AI infrastructure. Confusing those channels can make both the opportunity and the risk look larger—or smaller—than they are.
On November 12, Nvidia announced that Jensen Huang had immediately left his president, CEO and board roles. Colette Kress becomes interim CEO while remaining CFO; Harvey C. Jones becomes board chair. The November 17 results schedule remains unchanged. Nvidia’s leadership announcement.
For Canadians, this adds leadership uncertainty to an exposure they may already own through funds. It does not tell us the eventual earnings impact, establish a particular market-price reaction, or turn an older fund-weight snapshot into a current one.
The distinction matters now. Nvidia reported US$96.2 billion of revenue for the quarter ended July 26, 2026, including US$89.0 billion from Data Center. Its next-quarter revenue outlook was US$108 billion, plus or minus 2%, excluding Data Center compute revenue from China. Those are reported results and management guidance, respectively, not interchangeable evidence. Nvidia’s August 26 results.
The company has scheduled its next results discussion for November 17, covering the quarter ended October 25. As of this article’s date, that quarter’s results have not been released. Canadian investors are therefore evaluating an unusually consequential earnings cycle with the last completed disclosure still providing the financial baseline. Nvidia’s results announcement.
The question is not simply whether AI keeps growing. It is where Canadian investors already participate, which businesses capture the spending, and how much future success their investments require.
A Canadian account is not the Canadian stock market
There are three separate labels on an investment: where the security trades, what it owns, and the currency exposure it carries. A Toronto-listed fund bought with Canadian dollars can own U.S. businesses. Its returns can consequently depend on Nvidia even when its owner never enters an order for Nvidia shares.
Vanguard’s VFV illustrates the point. Its September 30, 2026 fact sheet identifies a Canadian-dollar trading currency, an S&P 500 benchmark, and an 8.4% underlying Nvidia weight. The document describes a portfolio of U.S. companies. Canadian packaging does not turn those businesses into Canadian equities. That 8.4% is a September snapshot, not a live November weight; market movements and portfolio changes can alter it. Vanguard’s VFV fact sheet.
The S&P/TSX Composite has its own eligibility and constituent rules. Nvidia’s U.S. shares should not be assigned a direct Composite contribution just because Canadians own them. Nor does a depositary receipt’s Toronto listing establish membership. S&P’s November methodology already incorporates changes taking effect on December 18; an announced eligibility change is also different from an actual constituent addition. S&P/TSX methodology, including its effective-date appendix.
This explains how two apparently conflicting statements can coexist: a Canadian investor can be heavily exposed to Nvidia, while the Canadian benchmark responds mainly through other companies and broader market conditions.
It also changes performance comparisons. If a portfolio contains substantial U.S. equity exposure, comparing its entire return with the TSX alone mixes differences in geography, company concentration and currency. Outperformance in that comparison does not identify which decision created it. A weaker result does not establish that the Canadian stock selection was poor.
The exposure hidden inside overlapping funds
Fund count is a weak measure of diversification. The useful calculation follows ownership through every layer:
Nvidia exposure = direct holdings + each fund’s value × its underlying Nvidia weight.
For a fund that owns other funds, multiply through the layers. An allocation of 45% to a U.S. sleeve containing 8% Nvidia produces a 3.6% Nvidia weight in the overall fund. Do not add the U.S. sleeve a second time as though it were another asset in the account.
Consider this hypothetical C$100,000 portfolio. The VFV weight uses the dated observation above; all allocations and the global fund’s weights are invented for illustration. This is not a reconstruction of an actual historical portfolio.
| Holding | Account value | Assumed Nvidia exposure within holding | Nvidia amount |
|---|---|---|---|
| Hypothetical global equity fund | C$60,000 | 45% U.S. sleeve × 8% Nvidia | C$2,160 |
| VFV | C$25,000 | 8.4%, September 30 snapshot | C$2,100 |
| Direct Nvidia shares, valued in CAD | C$5,000 | 100% | C$5,000 |
| Cash | C$10,000 | 0% | C$0 |
| Total | C$100,000 | 9.26% of account | C$9,260 |
A 30% Nvidia decline, with everything else—including currency—held unchanged, subtracts C$2,778, or 2.78% of that account. That is an isolated sensitivity calculation, not a portfolio stress forecast.
The distinction is substantial. Other holdings may sell AI services, finance infrastructure, or depend on the same corporate spending cycle. They can decline together without owning Nvidia. Conversely, some businesses could benefit if cheaper computing improves their economics. Ownership overlap measures one risk; common economic dependence measures another.
A portfolio review therefore needs two maps. The first lists dollars attributable to Nvidia itself. The second identifies other businesses whose earnings or valuations rely on continued AI investment. Adding the two into one unsupported “AI percentage” would sacrifice clarity: the exposures have different contracts, margins and sensitivities.
The accompanying exposure explorer makes the first map reproducible. It accepts direct and nested holdings, displays the calculation for each, and labels unknown weights rather than silently treating them as zero.
Hypothetical portfolio · one dated weight reference
Find the Nvidia already inside your portfolio
Look through a diversified fund’s US sleeve, a separate US equity fund and a direct holding. Mark missing fund weights as unknown rather than treating them as zero.
Amounts, nested weights and the shock are invented. The 8.4% US fund default is a VFV reference dated September 30, 2026; it is not a current holding guarantee. Do not enter the same asset twice. All other holdings and FX stay fixed.
- Total entered portfolio
- C$100,000
- Nvidia exposure
- C$9,260
- Nvidia share of portfolio
- 9.26%
- Modeled portfolio change
- −C$2,778
- Effect on whole portfolio
- -2.778%
Known-weight portfolio calculation. Other assets and FX are unchanged.
Read the data table and calculation
| Holding | Entered CAD value | Nvidia weight | Nvidia CAD exposure |
|---|---|---|---|
| Diversified fund | C$60,000 | 3.6% | C$2,160 |
| US equity fund | C$25,000 | 8.4% | C$2,100 |
| Direct Nvidia | C$5,000 | 100% | C$5,000 |
| Other assets | C$10,000 | 0% | C$0 |
Nested exposure = holding × US sleeve fraction × Nvidia fraction within that sleeve. Total exposure is the sum of known holding exposures. Shock impact = known exposure × stock-price change; no tax, fees, hedging or market spillovers.
Canadian-dollar trading does not remove currency risk
An unhedged U.S. investment has two moving parts for a Canadian owner. If S is Canadian dollars per U.S. dollar, its simplified Canadian-dollar return is:
(1 + U.S.-dollar return) × (ending S ÷ starting S) − 1.
The multiplication matters. A hypothetical 20% U.S.-dollar gain combined with a 10% fall in the CAD-per-USD rate produces an 8% Canadian-dollar gain, not 10%. A 20% stock decline combined with a 10% rise in that rate produces a 12% loss. Currency can cushion an equity loss or deepen it; neither outcome is assured.
For a dated reference, the Bank of Canada table retrieved for this article reports C$1.4181 per US$1 on November 10, 2026. This is an indicative daily average, not a brokerage conversion quote, an intraday price or a promised execution rate. The calculator keeps that reference separate from its hypothetical scenarios. Bank of Canada daily exchange rates.
Hedging changes this relationship, but it does not remove Nvidia’s business risk. BlackRock’s XQQ targets the NASDAQ-100 Currency Hedged CAD Index, making its stated currency objective different from simply buying a U.S. equity fund in Canadian dollars. It remains an equity portfolio with company and market exposure. BlackRock’s XQQ investment objective.
The Nvidia CDR is another distinct structure. CIBC identifies it as CAD hedged, with a TSX listing and Nvidia’s U.S. shares as the underlying security. Its notional hedge works through adjustments to the number of underlying shares represented by each receipt. CIBC warns that fees, interest-rate differences, volatility and rebalancing can create tracking differences. A receipt is not one ordinary Nvidia share, and “hedged” does not mean its return will match the U.S. share perfectly. Nvidia CDR details, CIBC’s hedge explanation.
There is a second currency distinction inside the business. Hedging an investment’s USD-to-CAD translation does not neutralize the effect of currencies on Nvidia’s customers, suppliers or global competitiveness. The investment wrapper can alter the shareholder’s translation exposure; it cannot hedge every commercial consequence inside the company.
Hypothetical CAD investor scenario
Separate the stock return from the currency return
The exchange rate is Canadian dollars per one US dollar. A fall in that rate means the Canadian dollar strengthens against the US dollar.
The default rate and returns are scenarios, not conversion quotes or forecasts. The ideal hedge comparison simply removes currency movement; real hedged funds/CDRs have costs, tracking differences and hedge mechanics.
- Ending CAD per US$1
- 1.26
- Unhedged return in CAD
- 8%
- Ending unhedged value
- C$10,800
- Ideal currency-neutral ending value
- C$12,000
- Currency translation effect
- −C$1,200
Read the data table and calculation
| Measure | Scenario result |
|---|---|
| Unhedged CAD return | 8% |
| Ideal currency-neutral return | 20% |
| Ending exchange rate | 1.26 CAD per USD |
| Equity–currency interaction | -2 percentage points |
CAD return = (1 + USD stock return) × (1 + CAD-per-USD rate change) − 1. Ending CAD = starting CAD × (1 + CAD return). USD total return assumes distributions reinvested in USD. Conversion costs, tax, hedge costs and cash-flow timing differences are excluded.
Canada’s infrastructure opportunity has several different profit pools
Canadian participation is not restricted to owning foreign shares. Bell’s June 18 announcement describes an arrangement involving its Merritt, British Columbia infrastructure, Cohere’s models, BUZZ HPC’s cloud layer and Canadian-built Hypertec hardware using Nvidia technology. It establishes a concrete commercial connection among facilities, connectivity, compute and software. It does not disclose enough to calculate the arrangement’s eventual contribution to BCE earnings. Bell’s infrastructure announcement.
TELUS supplies a separate example with a different milestone. In September 2025, it announced that its Rimouski facility was operational and serving customers using Nvidia GPUs and HPE infrastructure. An operating facility provides stronger evidence of deployment than a proposed site, although the announcement alone does not establish its present utilization or investment return. TELUS’s opening announcement.
These activities divide the economics of AI rather than distributing them evenly. The chip supplier sells equipment and systems. A building owner supplies space, cooling and electrical infrastructure. A cloud operator bears equipment and utilization risk. A software provider must convert computing into a service customers will pay for.
The same spending can consequently be revenue for one company and capital expenditure for another. A large hardware purchase can improve the supplier’s results while reducing the buyer’s near-term cash flow. The buyer may recover its investment over time, but recovery depends on service prices, demand, financing and replacement costs.
Data residency also has a narrower economic meaning than complete technological independence. Running sensitive workloads in Canada can create domestic value even when the processors and parts of the software stack come from foreign suppliers. The domestic contribution is the value created here—engineering, operations, construction, connectivity and useful services—not automatically the entire announced project budget.
That distinction matters for public-market analysis. “Sovereign AI” can describe a real customer requirement without revealing the margin available to every participating company. The relevant evidence progresses from announcement to contract, commissioning, paying workloads and cash generation. Each stage answers a different question.
A Canadian supplier can participate without being a Nvidia proxy
Celestica’s October 26 results offer evidence of Canadian-listed participation in the wider infrastructure cycle. It reported quarterly revenue of US$5.83 billion, up 83% year over year, and described demand from hyperscalers and frontier AI labs. Its disclosures discuss networking and compute programs, substantial planned investment, and risks from customer concentration and deployment constraints. Those are company-specific exposures, not a disclosed one-for-one relationship with Nvidia sales. Celestica’s third-quarter results.
This is a useful corrective to the habit of calling every related company “the Canadian Nvidia.” A systems supplier, network-equipment business and semiconductor platform owner can participate in the same project while earning different margins and carrying different inventory risks.
They can also fare differently as technology changes. Customers may buy more networking while altering their accelerator mix. Larger contracts may require inventory and receivables before cash arrives. An expanding order book can therefore coexist with greater financing needs. Revenue growth is evidence of activity; it is not by itself evidence that shareholders receive proportionate cash growth.
For Canada’s export exposure, a company’s trading venue is again an incomplete guide. A Canadian-listed business serving global customers can depend more on foreign capital spending than on Canadian household demand. That can diversify its revenue geographically while increasing dependence on a narrow group of large buyers.
The analytical task is to establish the actual customer and product relationship. Without a disclosed contract or revenue breakdown, attributing a supplier’s entire growth to Nvidia would turn an industry association into an invented financial linkage.
Electricity is a constraint, not a guaranteed windfall
The size of an electricity connection request is not the amount of power a completed facility consumes. Alberta’s public dashboard provides a useful example: it records approximately 19,565 MW of requested new data-centre load as of July 30, 2026, explicitly distinguishing requests from approvals. Separately, the AESO says its 1,200 MW first-phase connection allocation has been fully assigned to projects with executed load contracts. Neither number represents currently operating AI load. Alberta’s dated figures, AESO connection program.
For a proposed project, the economic questions begin after the headline. Is power available at the required location and time? Who funds network upgrades? Can the customer curtail demand? Is the price fixed, floating or contractually passed through? What happens if commissioning slips while financing costs continue?
Electricity producers, transmission owners and customers can have different answers. Additional demand can create investment opportunities, but the shareholder benefit depends on contract terms, regulatory treatment and the capital required to serve it. An increase in generation spending is not automatically an increase in returns on that spending.
The power explorer makes the scale tangible without pretending to value a utility. A hypothetical 20 MW IT installation, averaging 65% of its nameplate electrical draw, with a power usage effectiveness of 1.20, consumes approximately 136,656 MWh over 8,760 hours. At an invented C$70 per MWh, its energy bill is about C$9.57 million.
Those inputs describe an illustration, not Bell, TELUS or an Alberta project. Average electrical draw is not the same as billable GPU utilization. The example also excludes demand charges, transmission, taxes, backup generation and equipment replacement. Its purpose is to expose which assumptions convert a megawatt headline into an operating expense.
Hypothetical engineering and energy-cost scenario
Translate a data centre’s power assumptions
Convert IT nameplate capacity and average electrical draw into facility energy. This is a scale calculation, not an investment valuation.
Every input is hypothetical. PUE is total facility electricity divided by IT electricity. Energy-only cost excludes equipment, construction, financing, network/demand charges, tax and other operating costs.
- Average facility draw
- 15.6 MW
- Facility electricity
- 136,656 MWh
- Energy-only cost
- C$9,565,920
- Extra cost for C$10/MWh more
- C$1,366,560
Read the data table and calculation
| Measure | Scenario result |
|---|---|
| Average facility draw | 15.6 MW |
| IT energy | 113,880 MWh |
| Overhead energy | 22,776 MWh |
| Energy-only cost | C$9,565,920 |
Average facility MW = IT MW × electrical-draw fraction × PUE. MWh = average facility MW × hours. Energy cost = MWh × CAD per MWh. A common year has 8,760 hours; a leap year has 8,784.
The hardest risk is the gap between technological success and investment returns
AI adoption can increase while investors earn disappointing returns. That is possible whenever the price paid already assumes even faster growth, stronger margins or a longer period of competitive advantage.
Consider a purely hypothetical business whose earnings rise 25% while its price-to-earnings multiple falls 20%. Multiplying 1.25 by 0.80 leaves its share price unchanged, before dividends. The business improved; its valuation compressed by an offsetting amount. This arithmetic is not an estimate of Nvidia’s current multiple or future earnings.
Canada’s infrastructure buyers face a related distinction. More useful AI can increase demand for computing, but efficiency improvements can reduce the computing needed per task. The eventual spending outcome depends on whether cheaper services generate enough additional usage to outweigh that reduction. Greater usage is not a mathematical guarantee of higher spending on every hardware generation.
Policy creates another connection across borders. Nvidia’s quarterly filing describes export controls as a risk to products, distribution and downstream customers. It also says geographic revenue is assigned by direct-customer headquarters, which may differ from the end customer or delivery location. Consequently, neither a Canadian facility announcement nor a regional revenue table supplies a clean measure of “Nvidia’s Canadian revenue.” Nvidia’s quarterly filing.
A Canadian project can be locally controlled yet remain exposed to foreign technology rules and supply chains. That is a commercial dependency to investigate, not proof that a particular Canadian purchase is prohibited.
The most informative next evidence will be specific: Nvidia’s forthcoming reported results, dated fund holdings, Canadian operators’ commissioned capacity and paid workloads, suppliers’ cash conversion, and power contracts that have progressed beyond requests. Those observations help separate growth already delivered from growth still embedded in expectations.
For a Canadian investor, the starting point is correspondingly concrete: identify the underlying holdings, identify the currency treatment, and identify the business relationship behind each claimed beneficiary. Nvidia can matter greatly to the portfolio without belonging to the Canadian benchmark, and AI can matter greatly to Canada without every AI-related investment producing the same result.
Sources, evidence and downloadable research (15 records)
Research checked November 12, 2026. Dates below describe the evidence, not a publication date. Dynamic source pages may change.
- NVIDIA Announces CEO Transition
nvidianews.nvidia.com · 2026-11-12, effective immediately
- NVIDIA Announces Financial Results for Second Quarter Fiscal 2027
investor.nvidia.com · Quarter ended 2026-07-26
- NVIDIA Sets Conference Call for Third-Quarter Financial Results
investor.nvidia.com · Scheduled 2026-11-17; quarter ended 2026-10-25
- Vanguard S&P 500 Index ETF | VFV — Factsheet
fund-docs.vanguard.com · 2026-09-30 holdings snapshot
- S&P/TSX Canadian Indices Methodology
www.spglobal.com · Current publication includes some changes effective 2026-12-18
- Bank of Canada — Daily exchange rates
www.bankofcanada.ca · Rate date 2026-11-10
- iShares NASDAQ 100 Index ETF (CAD-Hedged) | XQQ
www.blackrock.com · Product page checked 2026-11-12
- Nvidia CDR (CAD Hedged) | CIBC Canadian Depositary Receipt
cdr.cibc.com · Product page checked 2026-11-12
- CIBC CDR FAQ
cdr.cibc.com · Checked 2026-11-12
- Bell AI Fabric, Cohere, Hypertec and BUZZ HPC announce landmark deal to advance sovereign AI in Canada
www.bce.ca · Announced arrangement
- TELUS opens Canada’s first fully Sovereign AI Factory
www.telus.com · Opening announcement
- Celestica Announces Third Quarter 2026 Financial Results and Will Hold 2026 Investor and Analyst Day
corporate.celestica.com · Quarter ended 2026-09-30
- Alberta — AI Data Centres: By the Numbers
www.alberta.ca · Data-centre request snapshot 2026-07-30
- AESO — Large Load Projects
www.aeso.ca · Status checked 2026-11-12; section publication date unavailable
- NVIDIA Form10-Q for quarter ended July26,2026
investor.nvidia.com · Quarter ended 2026-07-26
Download evidence register · Download assumptions and formulas · Download editable article
Prepared for North of Numbers with AI assistance. General financial education; no independent review by a qualified financial professional is claimed. Calculators run on this page without uploading or saving your inputs. Results are illustrations, not recommendations. Published November 12, 2026. Historical observations retain their stated dates.