Nvidia in 2026: Why the next leg higher could be driven by AI infrastructure spend
Nvidia has been a long-term winner, rising roughly 1,200% over five years. In 2025, the ride was bumpier. Despite a near 34% year-to-date gain, headlines tied to President Donald Trump's import tariffs, export restrictions on China, and fresh chatter about an AI bubble pressured the stock at times.
Under the surface, the business didn't flinch. Nvidia still designs the fastest GPUs used to train and deploy AI models, and its early lead plus steady iteration has delivered record revenue and profit. Demand comments around Blackwell and its update, Blackwell Ultra, point to a healthy product cycle. Even a strong recent print didn't spark a sustainable rally as macro worries and valuation debates took the spotlight.
The real 2026 driver: an AI infrastructure build-out
Here's the thesis with teeth: AI infrastructure spending could reach $3 trillion to $4 trillion by the end of the decade, according to CEO Jensen Huang. Major cloud customers, including Amazon and Microsoft, have flagged surging demand and plans to expand capacity. That spending wave needs top-performing AI chips-Nvidia's core strength.
As this capex cycle translates into orders, shipments, and deployments, it should flow through Nvidia's earnings. The Blackwell generation is positioned for this phase, with performance and efficiency gains built for scaled AI workloads. For a quick look at the platform, see Nvidia's overview of Blackwell architecture: NVIDIA Blackwell.
If the infrastructure thesis plays out, 2026 becomes the year when procurement moves from trials to full production, backlog converts, and operating leverage shows up in EPS. With shares around 38x forward earnings (reasonable for this growth profile), renewed earnings momentum could be enough to pull the stock back onto buy lists.
What finance teams should watch
- Hyperscaler capex guides: Track Amazon, Microsoft, and other cloud providers' AI-related spend and data center buildouts.
- Data center revenue run-rate: Quarter-over-quarter growth, backlog trends, and visibility into Blackwell and Blackwell Ultra ramps.
- Pricing and mix: Average selling prices, bundle adoption, and networking attach rates that support gross margin.
- Lead times and supply: Signals on capacity, cycle times, and any constraint easing or new bottlenecks.
- Customer concentration: The split across top buyers and any signs of in-house custom silicon displacing orders.
Risks that can swing the setup
- Policy shocks: New tariffs or tighter export rules that disrupt shipments or demand planning.
- AI demand normalization: If workloads or ROI fall short, deployment schedules could slip.
- Competition: Alternatives from rivals or custom chips from large customers compressing pricing or share.
- Valuation sensitivity: Multiple compression if growth slows or if macro tightens.
Positioning ideas for 2025-2026 (for consideration, not advice)
- Treat 2025 as base-building: scale on weakness if the earnings path stays intact.
- Define thesis checkpoints: hyperscaler capex updates, Blackwell shipment cadence, margins, and backlog conversion.
- Think in scenarios: 2026 beat/meet/miss cases with clear triggers to add, hold, or trim.
- Broaden exposure thoughtfully: consider ecosystem names tied to AI data centers, while monitoring correlation risk.
The short version: volatility masked strength in 2025. If AI infrastructure spend accelerates as expected, 2026 could be the year that thesis shows up cleanly in Nvidia's numbers.
If you're upgrading your research stack, here's a practical resource: AI tools for finance.
Your membership also unlocks:
