NVIDIA Reports Record Earnings, but Faces Margin Pressures

Q4 Revenue Surpasses Expectations; Gross Margins Decline Slightly

NVIDIA (NASDAQ: NVDA) delivered another strong quarter, with Q4 revenue reaching $39.3 billion, up 78% YoY and 12% QoQ, exceeding consensus estimates by 2.8%.

• Gross margins came in at 73% (GAAP), below the 73.4% expectation, reflecting increasing cost pressures.
• Net profit surged to $22.1 billion, marking a 79.8% YoY and 14% QoQ increase, beating forecasts by 5.9%.
• The company projects Q1 FY2026 revenue of $43 billion, surpassing buy-side expectations of $42 billion.

---

Data Center Growth Leads Earnings as Blackwell Chips Launch

NVIDIA’s data center revenue reached $35.58 billion, up 93% YoY, reinforcing its dominance in AI infrastructure.

• The new Blackwell architecture debuted, contributing $11 billion in sales within its first quarter—accounting for 50% of total data center revenue.
• CEO Jensen Huang emphasized a 10-25x performance boost for AI inference workloads, with demand far exceeding expectations.

---

Annualized Net Profit Set to Cross $100 Billion; Forward P/E at 29x

• NVIDIA’s FY2025 net income reached $72.9 billion (GAAP) and $74.2 billion (Non-GAAP), with a static P/E of 43x.
• Bloomberg estimates FY2026 Non-GAAP net income at $111.4 billion, translating to a forward P/E of 28.8x—relatively low compared to its 5-year historical average.
• Compared to MAG7 peers (Microsoft, Apple, Google, Amazon, Meta, Tesla, NVIDIA), NVIDIA’s growth rate remains superior, though its B2B chip business faces long-term cyclicality risks.

---

DeepSeek’s Impact: A Double-Edged Sword for NVIDIA

• Concerns over DeepSeek-R1’s AI inference efficiency possibly reducing demand for NVIDIA GPUs were countered by Jensen Huang’s reassurance.
• NVIDIA sees AI inference as an additive segment, requiring significant GPU resources beyond initial training.
• The Blackwell-optimized R1 version reduced per-token cost by 1/20 and increased inference throughput by 25x, reinforcing NVIDIA’s hardware-software integration strategy.

However, competition is intensifying:

• Huawei’s Ascend 910C GPU reportedly delivers 60% of NVIDIA’s H100 performance, reflecting China’s rapid progress in AI hardware.
• U.S. export controls could further limit NVIDIA’s future market share in China, threatening long-term dominance.

---

Key Takeaways from NVIDIA’s Conference Call

• AI Compute Demand: Next-gen models could require “millions of times” current compute power, indicating long-term demand for GPUs.
• Enterprise AI Spending Growth: NVIDIA expects enterprise AI infrastructure spending to surpass cloud service providers (CSPs), driven by AI automation, digital transformation, and “Agentic AI” applications.
• Blackwell Ultra Launch: Planned for H2 2025, improving transition efficiency from the Hopper to Blackwellarchitecture.
• GTC 2025 Highlights: NVIDIA will unveil Blackwell Ultra, Rubin architecture, next-gen networking, AI inference models, and physical AI systems.

---

Outlook: Growth Momentum Continues, but Competition Looms

NVIDIA maintains strong growth momentum, but increasing AI hardware competition, cost pressures, and regulatory risks could pose challenges ahead.

Investors will be closely watching:

✅ Blackwell Ultra adoption

✅ AI inference market share

✅ Competition from Huawei and other challengers

✅ Impact of U.S.-China trade policies

Conclusion: While NVIDIA continues its AI dominance, long-term competition and margin pressures warrant caution despite its record-breaking earnings and revenue forecasts.