NVIDIA Slips to $227 as SMH Surges 1.15% on AI Supercycle Bets

NVIDIA (NVDA) dipped 0.72% to $227.21 on September 30, 2026 — roughly C$322.18 at the current USD/CAD rate of 1.4180 — but the retreat looked more like profit-taking than a loss of conviction. The VanEck Semiconductor ETF (SMH) climbed 1.15% to $606.90 on the same session, signalling that institutional money is rotating within the AI trade, not out of it. AMD also held nearly flat at $607.57, underscoring the breadth of the compute rally.

The Hyperscaler Capex Machine Keeps Running

The core AI investment thesis rests on a simple but powerful feedback loop: the larger the language models, the more compute is required, and the more compute is required, the more the hyperscalers spend. Microsoft, Alphabet, Amazon, and Meta collectively committed over US$250 billion in capital expenditure for fiscal 2026, the vast majority earmarked for data centre build-outs, custom silicon, and AI inference infrastructure. Microsoft (MSFT) closed at $508.96 on Wednesday, barely changed on the day, reflecting the market’s view that its Azure AI revenue ramp is on track and priced in — not a reason to sell. Today’s SMH surge confirms that investors still believe the silicon supply chain has years of pricing power ahead.

AI Adoption Metrics: Past the Hype Inflection Point

Enterprise AI adoption has crossed what analysts are calling the “workflow integration threshold” — the point at which AI tools shift from pilot projects to mission-critical infrastructure. Goldman Sachs estimated in its September 2026 AI Productivity Monitor that generative AI is now embedded in over 40% of Fortune 500 knowledge-worker workflows, up from just 11% at the start of 2025. That uptake is driving inference demand — the compute needed to run AI models at scale — which is increasingly becoming a larger revenue contributor than training workloads for NVIDIA. The NVDA dip today may reflect short-term uncertainty around its next earnings print, expected in mid-November, but guidance commentary from hyperscaler partners remains resolutely bullish.

The Canadian Angle: Cohere, Vector, and TSX Beneficiaries

Canada’s AI ecosystem is emerging as a credible second-tier player in the global buildout. Toronto-based Cohere, founded by former Google Brain researchers and backed by OMERS Ventures and Salesforce Ventures, has positioned its enterprise-focused large language models as the compliance-friendly alternative to OpenAI for regulated industries — a significant differentiator for Canadian banks and insurers. The Vector Institute, now in its ninth year, continues to funnel world-class AI talent into both startups and established Canadian firms. On the TSX, the spillover is visible: Constellation Software (CSU) gained 1.31% to C$2,896.47, reflecting investor confidence that its portfolio of vertical-market software businesses will be net beneficiaries of AI-driven productivity gains. Shopify (SHOP) led the Canadian tech board, jumping 2.95% to $148.26, as its AI-native commerce tools — including Sidekick, its merchant assistant — continue to deepen platform stickiness.

The Risk Worth Flagging: Valuation Has Run Ahead of the Earnings Curve

Even the most committed AI bulls should acknowledge one structural concern: the gap between AI capex spending and measurable AI-driven revenue is widening, not narrowing. NVIDIA trades at approximately 35x forward earnings — a premium that demands near-flawless execution on data centre demand through at least 2028. Any signal of hyperscaler capex moderation, a faster-than-expected commoditization of inference chips (including competition from AMD, Broadcom’s custom ASICs, and Amazon’s Trainium), or a macro-driven enterprise IT spending freeze could compress multiples sharply. The SMH’s 1.15% gain today is encouraging, but retail investors should remember that the ETF is heavily concentrated — its top five holdings represent over 50% of the fund. Diversification within the AI theme, including exposure to Canadian energy infrastructure powering new data centres, remains a prudent hedge.