- NVIDIA rose 0.57% to $212.17 USD ($295.33 CAD) on September 16, 2026, while AMD outperformed with a 2.19% gain to $504.20.
- Hyperscalers including Microsoft, Alphabet, Amazon, and Meta have committed over USD $300 billion combined in AI infrastructure capex for fiscal 2026.
- Toronto’s Cohere is valued above USD $5 billion and winning enterprise contracts, while Vector Institute talent continues to anchor Canada’s AI pipeline.
- Valuation risk is real: NVIDIA trades at ~35x forward earnings and Microsoft’s Copilot monetization is lagging early analyst revenue-per-seat projections.
NVIDIA (NVDA) edged up 0.57% to $212.17 (USD) on September 16, 2026 — equivalent to roughly $295.33 CAD at the current exchange rate of 1.3920 — as the semiconductor sector held its footing amid a mixed session for tech. The VanEck Semiconductor ETF (SMH) added a modest 0.11% to $542.11, while AMD surged 2.19% to $504.20, signalling that the market’s appetite for AI-adjacent compute hardware remains intact even as software plays like Microsoft slid 1.64% to $497.12.
The Hyperscaler Capex Engine Shows No Signs of Stalling
The four dominant hyperscalers — Microsoft, Alphabet, Amazon, and Meta — have collectively committed over USD $300 billion in capital expenditure for fiscal 2026, with the lion’s share directed at AI data centres, custom silicon, and high-bandwidth networking. Microsoft alone guided for more than USD $80 billion in AI infrastructure spending this fiscal year. That figure, even as MSFT shares softened today, underscores a critical distinction: near-term stock volatility does not equal thesis deterioration. The buildout is structural, not cyclical.
NVIDIA remains the primary toll booth on this highway. Its H200 and Blackwell GPU architectures are sold out well into 2027 according to supply chain disclosures, and data centre revenue now accounts for more than 85% of total company sales. Every dollar a hyperscaler spends on AI training or inference clusters flows, in large part, through Santa Clara. AMD’s outperformance today — up more than 2% — suggests the market is also pricing in a credible second-source dynamic, particularly as its MI300X accelerators gain traction in inference workloads.
AI Adoption Metrics Are Crossing the Inflection Point
Beyond hardware, adoption data is hardening the investment thesis. Enterprise AI deployment rates — measured by active API calls, token consumption, and seats sold for tools like Microsoft Copilot and Google Gemini for Workspace — have accelerated sharply in 2026. Industry surveys now indicate that more than 60% of Fortune 500 companies have moved at least one AI workload from pilot to production, up from roughly 25% at the start of 2025. This is the transition from “spending on AI” to “generating returns from AI,” and it validates the multi-year capex cycle rather than exposing it as hype.
Canada’s Quiet Role in the AI Supercycle
Canada is not merely a spectator. Toronto’s Vector Institute, one of the world’s leading academic AI research hubs, continues to funnel top-tier machine learning talent into both domestic and international companies. Cohere, the Toronto-based enterprise large language model company co-founded by Aidan Gomez, raised at a valuation exceeding USD $5 billion in its most recent round and is actively signing multi-year contracts with financial institutions and government agencies. Alumni of Element AI — acquired by ServiceNow in 2021 — have since seeded more than a dozen Canadian AI startups, several of which are on a trajectory toward TSX or TSX-V listings within the next 18 months. For investors seeking TSX-listed exposure, Constellation Software (CSU) — despite falling 3.03% to $2,870.26 today — operates a portfolio of vertical-market software businesses increasingly integrating AI-driven automation, making it a durable, if indirect, AI compounder.
Valuation Risk: The Gap Between Capex and Monetization
The risk worth flagging is not whether AI is real — it clearly is — but whether the pace of monetization can justify the pace of spending. At $212, NVIDIA trades at approximately 35x forward earnings, a premium that assumes sustained hyperscaler demand and no meaningful supply-chain disruption or geopolitical shock to chip exports. Microsoft’s decline today may reflect investor impatience: Copilot seat growth has been strong, but revenue contribution per seat remains below early analyst models. If hyperscalers begin signalling capex moderation in their Q3 2026 earnings calls — scheduled for October — the multiple compression in AI hardware names could be swift. Position sizing and diversification across the compute stack (hardware, software, infrastructure) remain essential discipline.
