Cantor Fitzgerald reiterates Nvidia stock rating on AI demand growth

Cantor Fitzgerald reiterated an Overweight rating and $350 price target on Nvidia, implying roughly 53% upside from its cited $228.38 share price, after meetings with management highlighted broadening AI demand. The firm sees upside to Nvidia's guidance for more than 70% calendar-2027 revenue growth and ties its target to 13x updated calendar-2028 EPS of $26. Management cited expanding hyperscaler and enterprise demand for agentic AI, while 35 analysts have recently raised earnings estimates.

Analysis

The investable implication is not another near-term GPU demand upgrade; it is a potential extension of AI spending from accelerator procurement into full-rack infrastructure. If hyperscaler budgets remain capacity-constrained rather than ROI-constrained, NVDA’s platform integration raises switching costs while shifting incremental spend toward HBM suppliers (MU), foundry capacity (TSM), networking, power and thermal names (VRT). The key 6-18 month question is whether AI capex broadens to enterprise and sovereign buyers fast enough to absorb capacity after the initial hyperscaler buildout.

SMCI has high revenue sensitivity to Rubin rack deployment but materially less pricing power than NVDA: its upside depends on component availability and execution, while liquid-cooling competition from Dell and HPE can pressure gross margin even in a strong demand tape. That makes SMCI a higher-beta tactical beneficiary over the next 1-3 months, not necessarily the best structural expression. PINS’ infrastructure deployment is strategically validating but too small to alter NVDA earnings; investors should avoid extrapolating individual customer announcements into incremental demand.

Consensus risk is that increasingly aggressive long-range compute-spend narratives are being capitalized before purchase commitments are visible. A pause in hyperscaler capex, a lengthening GPU deployment cycle, or an inability to monetize agentic-AI workloads would hit NVDA’s multiple before it materially hits reported revenue. Conversely, evidence that non-hyperscaler demand is becoming a meaningful mix contributor would justify further multiple durability and favor the infrastructure basket over a single-name position.

Key Decisions for Investors

• Maintain/establish a 6-12 month core long in NVDA, preferably accumulated on semiconductor-sector pullbacks rather than chased on analyst-target headlines. Upside requires sustained order visibility into the next platform cycle; exit or trim if a major hyperscaler signals AI-capex deceleration or NVDA guides to a material gross-margin step-down without offsetting revenue acceleration.
• Use a 1-3 month pair: long NVDA versus short SMCI in equal dollar beta-adjusted exposure. The trade isolates platform economics from lower-margin rack integration; invalidate if SMCI delivers sustained gross-margin expansion and backlog conversion above expectations, which would indicate its systems position is more durable than assumed.
• Add MU and VRT as smaller 6-18 month satellite longs only after confirming HBM pricing/volume commentary and data-center power backlog conversion in upcoming earnings. These are cleaner second-order beneficiaries of system-level deployment, but both are vulnerable if AI infrastructure lead times normalize faster than demand.
• Do not treat the cited long-dated earnings multiple or price target as an entry signal without independently reconciling the underlying EPS estimate, share count, and supply assumptions. Set an alert around hyperscaler quarterly capex guidance: a second consecutive aggregate upward revision is the catalyst to increase exposure; a broad cut is the risk-off trigger for the entire AI hardware complex.