Q3 2025 Summary
Published Feb 7, 2025, 7:58 PM UTC- NVIDIA's demand for its new Blackwell GPUs is staggering, with next-generation foundation models starting at 100,000 Blackwells, indicating strong future growth potential.
- The company expects no slowdown until it modernizes $1 trillion worth of data centers, driven by the shift from CPU to GPU computing for AI and machine learning workloads.
- NVIDIA is positioned to capitalize on the rise of AI-native companies and the growing demand for inference computing, as it is the largest inference platform in the world with a large installed base and robust ecosystem.
- Supply constraints in gaming and data center segments may limit revenue growth. NVIDIA's gaming revenue is expected to decline sequentially in Q4 due to supply constraints, despite strong demand. Additionally, the complexity of building Blackwell systems, involving seven different custom chips, may lead to supply chain challenges impacting the ramp-up.
- Near-term gross margin pressure due to ramping new products. As NVIDIA ramps up Blackwell, gross margins are expected to moderate to the low 70s from mid-70s, potentially affecting profitability in the near term.
- Regulatory and geopolitical risks may impact business in China. Changes in U.S. administration and potential new export controls could affect NVIDIA's operations in China. The company anticipates the market in China to remain very competitive and will continue to comply with export controls while serving customers.
| Metric | YoY Change | Reason |
|---|---|---|
Total Revenue | +94% YoY | The jump from previous-year levels was driven by record Data Center performance, fueled by global demand for accelerated computing and generative AI, as well as the broad industry transition to NVIDIA’s GPU-powered solutions. This momentum is expected to continue as more enterprises adopt AI at scale. |
Data Center | +112% YoY | Growth came primarily from high adoption of NVIDIA GPUs, CPUs, and networking for large language models and other AI workloads. External market demand for generative AI and company-specific Hopper/Blackwell platform initiatives helped drive sustained expansion. |
Compute & Networking | +112% YoY | The segment benefited from strong enterprise and cloud service provider investments in AI infrastructure, particularly in Hopper-based GPUs and high-performance networking. Ongoing architectural improvements and new product launches supported further gains. |
Automotive | +72% YoY | The increase is largely attributed to ramping NVIDIA DRIVE platforms in autonomous and electric vehicle programs, which require significantly higher compute. As automakers continue integrating AI capabilities, this segment remains poised for steady, long-term growth. |
Net Income | +109% YoY | Earnings soared on the back of higher revenue, expanded gross margins (driven by Data Center), and operational efficiencies. Forward-looking implications suggest Nvidia will reinvest in R&D and strategic partnerships to maintain momentum in AI and advanced computing. |
Singapore | +185% YoY | While the documents provide limited specifics, the strong rise in Singapore-based revenue may reflect regional expansion and increased operations or supply-chain activities tied to AI and accelerated computing. Sustained demand across Asia-Pacific regions could further boost this figure. |
| Metric | Period | Previous Guidance | Current Guidance | Change |
|---|---|---|---|---|
Revenue | Q3 2025 | $32.5B ± 2% | no current guidance | no current guidance |
GAAP Gross Margin | Q3 2025 | 74.4% ± 50 bps | no current guidance | no current guidance |
Non-GAAP Gross Margin | Q3 2025 | 75% ± 50 bps | no current guidance | no current guidance |
GAAP Operating Expenses | Q3 2025 | $4.3B | no current guidance | no current guidance |
Non-GAAP Operating Expenses | Q3 2025 | $3.0B | no current guidance | no current guidance |
Other Income/Expenses | Q3 2025 | $350M | no current guidance | no current guidance |
GAAP & Non-GAAP Tax Rate | Q3 2025 | 17% ± 1% | no current guidance | no current guidance |
Revenue | Q4 2025 | no prior guidance | $37.5B ± 2% | no prior guidance |
GAAP Gross Margin | Q4 2025 | no prior guidance | 73% ± 50 bps | no prior guidance |
Non-GAAP Gross Margin | Q4 2025 | no prior guidance | 73.5% ± 50 bps | no prior guidance |
GAAP Operating Expenses | Q4 2025 | no prior guidance | $4.8B | no prior guidance |
Non-GAAP Operating Expenses | Q4 2025 | no prior guidance | $3.4B | no prior guidance |
Other Income/Expenses | Q4 2025 | no prior guidance | $400M | no prior guidance |
GAAP & Non-GAAP Tax Rate | Q4 2025 | no prior guidance | 16.5% ± 1% | no prior guidance |
| Metric | Period | Guidance | Actual | Performance |
|---|---|---|---|---|
Revenue | Q3 2025 | $32.5B ± 2% | $35.082B | Beat |
GAAP Gross Margin | Q3 2025 | 74.4% ± 50 bps | ~74.6% = (35,082 − 8,926) / 35,082 | Met |
GAAP Operating Expenses | Q3 2025 | $4.3B | ~$4.288B (SG&A: 898 + R&D: 3,390) | Met |
| Topic | Previous Mentions | Current Period | Trend |
|---|---|---|---|
Structural shift from CPU-based to GPU-accelerated computing in data centers | Repeatedly cited as a fundamental shift, with need for higher performance and reduced costs. | Jensen Huang emphasized modernizing the world’s data centers to handle AI workloads, describing a US$1 trillion opportunity. | Consistently highlighted as essential for AI-driven growth. |
Significant and sustained demand for Blackwell GPUs as a growth driver | Previously described as a major leap in performance, with strong anticipation for launches in H2 2025. No mention in Q4 2024. | Demand exceeds supply; more Blackwells will ship than initially planned, fueling revenue. | Recurring demand outstripping supply. |
Generative AI adoption fueling data center and enterprise revenue growth | Cited as a major catalyst in prior calls, with frontiers in model building and enterprise adoption. | Data center revenue up triple digits YoY, driven by AI factories, agentic AI, and generative AI workloads. | Consistent engine of revenue growth across calls. |
Supply constraints and complex production challenges for new GPU launches | Recurrent caution about yields, mask changes, and cooling challenges in earlier quarters. | Blackwell ramp described as complex (multiple custom chips, configurations), with supply trailing enormous demand. | Ongoing production bottlenecks amid high demand. |
Gross margin pressures tied to new product ramps and higher operating expenses | Earlier quarters noted inventory provisions and rising engineering costs tied to new product introductions. | Margins initially in the low 70% range due to Blackwell costs, with OPEX up 9% sequentially. | Persistent margin headwinds from ramp-up costs. |
Concerns about possible overinvestment in AI infrastructure affecting long-term demand | Past calls underscored immediate ROI and strong, sustained demand; not raised in Q4 2024. | Management dismissed near-term digestion risk, expecting multi-year modernization and AI-native growth. | Widely discussed, but NVIDIA remains optimistic. |
Competition from custom GPUs and ASICs threatening market share | Addressed only in Q1 2025, highlighting NVIDIA’s flexible architecture vs. specialized ASICs. | No discussion of custom GPU/ASIC competition. | Limited mention, not raised again. |
Export control restrictions reducing data center revenue in China | Persistently noted in earlier calls; shipments of export-compliant products continue. | China data center revenue grew sequentially but remains well below pre-restriction levels. | Continuing impact, though partial rebound observed. |
Transition from air-cooled to liquid-cooled systems creating adoption hurdles | Focus in Q2 and Q1 on engineering complexities and some customers not yet ready for liquid cooling. | No discussion in this period. | Occasional challenge, not revisited recently. |
Software offerings (e.g., NVIDIA AI Enterprise) | Cited across earlier calls as a crucial subscription platform (US$4,500/GPU/yr), targeting multi-billion-dollar run rates. | NVIDIA AI Enterprise shown as a key driver, with nearly 1,000 companies using NIM; revenue expected to more than double. | Consistently mentioned, seeing broad adoption. |
Emergence of sovereign AI infrastructure | Threaded throughout earlier calls with countries investing in localized AI supercomputers. | Noted as an ongoing driver, especially in Asia and Europe building their own AI data centers. | Continuing focus on domestic AI capabilities globally. |
-
Gross Margin Outlook
Q: Can NVIDIA recover gross margins to mid-70s% in 2025?
A: Colette Kress believes reaching mid-70s% gross margins in the second half of next year is a reasonable goal, depending on the product mix and ramp. -
Blackwell Ramp and Supply
Q: How is the Blackwell production and supply ramp progressing?
A: Jensen Huang stated that Blackwell production is in full swing, exceeding previous estimates by delivering more units this quarter. Despite demand exceeding supply, they are increasing production with supply chain partners and are on track with their annual roadmap. -
Hopper and Blackwell Demand
Q: Will Hopper shipments decline as Blackwell ramps up?
A: Colette Kress mentioned that Hopper demand remains strong, and they will continue selling Hopper in Q4 across all configurations, including China. Customers are also building out Blackwell, so both products will be available, and it's possible for Hopper to grow between Q3 and Q4. -
No Expected Digestion Period
Q: Is a digestion period expected in hardware deployments?
A: Jensen Huang believes there will be no digestion period until the world's $1 trillion data centers are modernized for machine learning and generative AI, anticipating continuous growth over the next several years as they build out AI infrastructure. -
Gross Margins During Blackwell Ramp
Q: Will gross margins decline during the Blackwell ramp?
A: Colette Kress expects gross margins to be in the low 70s% during the initial Blackwell ramp due to new configurations and chips, but aims to reach the mid-70s% quite quickly in subsequent quarters as they improve yields and product mix. -
Inference Market Growth
Q: How does NVIDIA view the growth of the inference market?
A: Jensen Huang sees significant growth potential in inference as AI adoption expands. He aims for widespread 24/7 inference usage across industries, including AI-native startups, and notes that applications requiring high throughput and low latency benefit from NVIDIA's architecture. -
Networking Business Outlook
Q: What is happening with the Networking business?
A: Colette Kress explained that despite a slight sequential decline, networking is growing year-over-year and is critical to data center operations. They expect growth to resume as they prepare for Blackwell and more systems incorporating NVIDIA's networking solutions. -
Sovereign AI Demand
Q: What is the status of sovereign AI demand?
A: Colette Kress stated that sovereign AI demand remains strong, with the pipeline intact as countries build foundational models in their own languages and cultures. This leads to growth opportunities in Europe and Asia Pacific. -
Gaming Supply Constraints
Q: Are supply constraints affecting the Gaming segment?
A: Colette Kress noted tightness in gaming supply due to rapid sell-through and efforts to ramp all products. They expect to be back on track with more supply in the new calendar year but will be tight for the current quarter. -
Scaling of Large Language Models
Q: Is scaling of large language models stalling?
A: Jensen Huang stated that foundation model pretraining scaling continues, and they've discovered new ways to scale through post-training and inference-time scaling. This drives greater demand for their infrastructure, including Blackwell.