- AMD surged 5.90% to US$505.74 on September 9, 2026, signalling hyperscalers are diversifying AI chip supply away from NVIDIA’s sole dominance.
- Collective hyperscaler AI capex exceeds US$300 billion in fiscal 2026, sustaining the infrastructure supercycle even as individual chip stocks rotate.
- Toronto’s Cohere and Vector Institute alumni represent Canada’s best-positioned plays on the enterprise AI inference layer and rising IP filings.
- NVIDIA’s 35× and AMD’s 40× forward earnings multiples leave both stocks exposed to sharp corrections on any guidance miss or capex slowdown.
AMD shares jumped 5.90% to $505.74 (US$366.35) on September 9, 2026, outpacing every other major semiconductor name and sending a clear message: the AI compute supercycle is broadening. NVIDIA, long the undisputed king of AI accelerators, slid 2.01% to $225.73 — roughly C$311.53 at the current USD/CAD rate of 1.3805 — as traders rotated into AMD on fresh optimism around its MI350 GPU ramp and growing hyperscaler design wins. The Philadelphia Semiconductor ETF (SMH) split the difference, adding 1.19% to $573.73, suggesting the sector thesis remains intact even as leadership shifts.
The Hyperscaler Capex Engine Keeps Running
The structural story underpinning both stocks hasn’t changed. Microsoft, Alphabet, Amazon, and Meta have collectively committed over US$300 billion in AI-related capital expenditure for fiscal 2026 — a figure that represents the largest coordinated infrastructure buildout in the history of enterprise technology. Microsoft closed at $493.95, off 1.15%, a modest pullback that analysts attribute to profit-taking rather than any revision to its Azure AI growth outlook. Every dollar these hyperscalers spend on data centres, custom silicon, and networking gear flows directly into the revenue lines of chipmakers, cooling suppliers, and power infrastructure plays.
Today’s AMD move confirms a key pillar of the AI investment thesis: no single vendor captures the entire supercycle. As hyperscalers pursue supply-chain resilience, AMD’s MI-series GPUs and EPYC CPUs are increasingly specified alongside — or instead of — NVIDIA’s H-series parts. Morgan Stanley estimates AMD’s AI accelerator revenue could hit US$12 billion in calendar 2026, up from near zero three years ago. That kind of trajectory commands attention regardless of where NVIDIA trades on any given day.
Canadian Angle: Cohere and the Inference Economy
Canada’s AI ecosystem is quietly benefiting from the same capex wave. Cohere, the Toronto-based large-language-model company co-founded by Vector Institute alumni, has positioned itself as the enterprise inference layer sitting atop the hardware stack — a software margin story inside a hardware supercycle. The company’s Command R+ model is now embedded in production workloads at several Fortune 500 firms, and its partnership with Oracle Cloud gives it global distribution without the overhead of owning compute. For Canadian retail investors without direct access to private-market Cohere shares, the closest TSX proxy remains Constellation Software (CSU), which closed at $2,853.79, down 5.63% — a sharp single-day drop that appears tied to broader software multiple compression rather than any fundamental AI-thesis deterioration.
The Vector Institute, meanwhile, reported this week that Canadian AI-related patent filings rose 31% year-over-year in the first half of 2026, with Toronto and Montréal accounting for the bulk of new registrations. That intellectual-property momentum is the upstream indicator that commercializable Canadian AI products — from Cohere to a growing cohort of Element AI alumni now leading stealth-stage startups — are building durable competitive moats.
The Valuation Risk You Can’t Ignore
For all the excitement, the valuation math deserves scrutiny. NVIDIA still trades at roughly 35× forward earnings even after today’s pullback, a premium that bakes in near-flawless execution through at least 2028. AMD’s surge pushes its own forward multiple above 40×, pricing in a GPU market-share trajectory that has yet to fully materialize in reported revenue. Any guidance miss — from NVIDIA’s next earnings call or a hyperscaler capex trim — could reprice the entire sector swiftly. Canadian investors holding SMH or U.S.-listed AI names also carry meaningful currency risk: a strengthening loonie against the current 1.3805 USD/CAD rate would erode CAD-denominated returns even if USD prices hold steady. Position sizing and staggered entry remain the most prudent tools in a market this momentum-driven.
