NVIDIA (NVDA) FY25 Q2 Earnings Review

AI and GPU giant NVIDIA (NVDA) released its FY2025 Q2 earnings (May to July 2024) after the bell on 8/28. While both revenue and earnings beat analyst expectations, the market is anticipating even more.

NVIDIA (NVDA) FY2025 Q2 Financial Results

• Revenues: $30.04B (YoY +122%)
  • Data Center: $26.3B (YoY +154%)
  • Gaming and AI PC: $2.9B (YoY +16%)
  • Professional Visualization: $454M (YoY +20%)
  • Automotive and Robotics: $346M (YoY +37%)
• EPS: $0.68 (YoY +152%)
• Gross Margin: 75.7% (YoY +4.5pts)

Data Center

Riding the wave of exploding demand for generative AI and large language models, NVIDIA’s data center division hit another all-time high, surging 154% YoY and 16% QoQ. This reflects broad-based growth across customer types, with significant cloud service providers’ aggressive AI infrastructure deployments contributing about 45% of revenue. Enterprise and consumer internet applications accounted for over 50%, driving the robust QoQ growth in the data center segment.

Both sub-segments within the data center division saw strong growth. The core compute business rocketed 162% YoY and 17% QoQ. NVIDIA emphasized that while customers continue to accelerate purchases of Hopper series chips, they’re also gearing up to adopt the Blackwell series. Meanwhile, the networking business, which provides high-speed, low-latency solutions, saw revenue jump 114% YoY and 16% QoQ, primarily driven by solid demand for InfiniBand technology and Ethernet solutions.

NVIDIA stated that Blackwell series chip samples were shipped to customers in Q2, with some design (mask) adjustments made to improve production efficiency (yield). Blackwell series chips are on track to start mass production and shipment from Q4, potentially contributing billions in revenue. Simultaneously, Hopper demand remains robust, with shipments expected to increase in the second half of 2024.

Despite export ban challenges, China remains a crucial driver of data center segment growth. Q2 data center revenue from China saw considerable QoQ growth, though its share of overall data center revenue was lower than historical levels. According to the Financial Times, despite tightening U.S. chip export bans, Chinese tech giants’ AI-related capital expenditures have doubled compared to last year.

Market rumors suggested delays in Blackwell series chip shipments due to design flaws, but NVIDIA subsequently clarified that production plans are proceeding on schedule. Morgan Stanley’s latest report indicates that NVIDIA’s Blackwell series chip shipment timeline remains unchanged, with deliveries expected in Q4 2024 primarily concentrated in November and December.

The market consensus is that even if the Blackwell series were slightly delayed, demand would be deferred, not lost. The current AI industry demand is robust, but we still face severe supply shortages. It’s expected that NVIDIA can offset any minor Blackwell delays by extending the lifecycle of the previous generation Hopper chips.

Notably, since the Blackwell chip announcement, most customers, including major cloud service providers, have universally modified specifications, requiring data center suppliers to prepare for liquid cooling systems. The cooling market is expected to shift focus towards Direct Liquid Cooling (DLC) starting from 2026.

Gaming and AI PC

NVIDIA’s gaming segment grew 16% YoY and 9% QoQ. While not yet returning to historical highs, it marks the fifth consecutive quarter of YoY growth. Sales of RTX 40 series GPUs and gaming console chips (Nintendo Switch) both increased in Q2. The latest IDC survey shows global PC shipments reached 64.9 million units in Q2 2024, up 3% YoY.

Professional Visualization

The professional visualization segment grew 20% YoY and 6% QoQ, primarily driven by robust sales of Ada architecture workstations. NVIDIA believes the next wave of growth in this segment will be fueled by generative AI and industrial process digitalization trends.

Automotive and Robotics

Automotive segment revenue increased 37% YoY and 5% QoQ, mainly driven by rising demand for autonomous driving platforms and AI cockpits. Total automotive revenue hit a record high of $346 million this quarter. While currently only accounting for about 1% of company revenue, future growth potential remains promising, considering global automakers’ collaborations with NVIDIA on AI development.

NVIDIA (NVDA) FY25 Q2 Earnings Call

The NVIDIA board approved an additional $50 billion stock buyback program this quarter.

Data Center and AI

NVIDIA’s data center business performance was outstanding. Compute product revenue grew over 250% YoY, while networking product revenue increased over 200% YoY.

Regarding customer composition, cloud service platforms accounted for 45% of NVIDIA’s data center business revenue, while consumer-facing internet companies and enterprise users made up 50%. This diversified customer base helps maintain stable revenue streams.

Regarding the Chinese market, despite U.S. Department of Commerce export controls, data center business still saw QoQ growth this quarter, becoming one of the primary growth drivers. However, China’s data center business revenue remains below pre-restriction levels.

In the past four quarters, inference demand accounted for about 40% of data center revenue. This indicates that AI models require substantial computing resources not only during training but also in practical application phases.

NVIDIA stated that AI computing power on significant cloud platforms is sold out, indicating robust demand for NVIDIA GPUs.

NVIDIA estimates that Sovereign AI revenue will reach $1-3 billion this year, reflecting increased demand for localized AI solutions from governments and enterprises.

NVIDIA’s NIM (NVIDIA Inference Microservices) technology significantly reduces the difficulty of AI deployment for enterprises, further promoting AI adoption.

NVIDIA predicts that the next generation of AI models will require 10 to 20 times more computing power than current models, implying continued growth in demand for high-performance computing hardware.

Hopper and Blackwell

The H200 product platform accelerated shipments this quarter, becoming the leading revenue contributor. This reflects the market’s enthusiastic demand for NVIDIA’s latest technology.

While the supply-demand situation for the Hopper platform has improved, the Blackwell platform product line is experiencing extreme supply shortages.

NVIDIA expects revenue from new Blackwell architecture products to reach billions in Q4 this year. Meanwhile, H200 shipments will continue to increase in the second half of the year.

The Blackwell platform is expected to start mass production on time in Q4. Jensen Huang believes that due to the AI model arms race, Hopper platform sales will stay the same as the new Blackwell platform launches.

Others

NVIDIA expects its software, SaaS, and support revenue to reach an annual run rate of $2 billion by year-end.

The Spectrum-X high-speed Ethernet product line performed exceptionally well, with quarterly growth exceeding 100%. The product line is widely popular, with xAI using it to build the world’s largest GPU cluster.

All autonomous driving technology vendors use NVIDIA GPUs in their data centers.

The company believes PC market demand is strong and continues to grow, with a healthy channel inventory.

Growth in the professional visualization product line is driven by AI and graphics-related demand.

The Omniverse platform is making excellent progress in the automotive and manufacturing industries.

Robot developers, including BYD Electronics, Siemens, and Teradyne Robotics, are adopting NVIDIA’s ISAAC robotics platform for R&D and production.

Outlook for FY25 Q3

Q3 revenue is estimated at $32.5 billion, up 79% YoY. The YoY revenue growth rate will fall to double digits after four consecutive quarters of over 100% growth. The full-year gross margin remains on track at 75%.

Review and Analysis

According to the latest financial reports from Cloud Service Providers (CSPs), cloud computing infrastructure deployment has not slowed. Large data center operators plan to continue increasing capital expenditure in 2024, strongly focusing on investing in AI.

Investment bank Goldman Sachs estimates that after significantly increasing their capital expenditure in cloud computing this year, the four major North American CSPs – Microsoft, Google, Amazon, and META – are expected to increase their spending further next year.

Financial Results

Following NVIDIA’s Q2 2024 earnings report, the market seemed slightly disappointed with its revenue performance. However, this reaction may reflect unrealistic investor expectations rather than NVIDIA’s performance. Even if AI is the most transformative tech wave of the next decade, its growth rate can only maintain astronomical levels for a while. Those hundreds of percentage points in growth rates are inherently unsustainable.

For short-term speculators, obsessing over whether growth was 2% or declined 1% might matter, but this myopic guessing game is meaningless for understanding NVIDIA’s long-term value. From a more macro perspective, data center business growth rates will gradually decline from the current 150%+ to a range of 25%-35% over the next two years. This is not a pessimistic forecast – even a 25%-35% annual growth rate is an impressive achievement over the long term. Beyond revenue growth, NVIDIA’s gross margins will also return to more sustainable levels as supply and demand gradually balance. The 78% gross margin achieved in Q1 may be challenging to replicate, and even this quarter’s 75% is unlikely to be maintained long-term.

NVIDIA’s growth in operating expenses, while potentially putting pressure on operating margins in the short term, is a sage strategy from a long-term perspective. For a company committed to maintaining technological leadership, continued increases in R&D investment are inevitable. Conversely, if NVIDIA chose to cut R&D spending at this stage to pursue higher short-term profit margins, that would be a genuinely alarming signal.

Data Center and AI

Regarding rumors of potential delays in the B200 series new GPU platform, based on information disclosed in the earnings report, even if there are some design adjustments, the overall delay appears quite limited, possibly just a month or two. This level of delay has minimal impact on NVIDIA, mainly because the current flagship H200 GPU platform is fully capable of filling this brief gap. This phenomenon also indirectly reflects the intensity of the current AI arms race – buyers can’t even tolerate waiting a month or two for the next generation of GPUs.

NVIDIA’s move to introduce cabinet-level AI servers, based on discussions in this quarter’s earnings call, aims not to grab market share from system manufacturers but to propose a new architectural solution for next-generation data centers from the perspective of overall computational efficiency. This clarification of strategic positioning makes me more optimistic about NVIDIA’s long-term development prospects.

Given the current industry situation, NVIDIA’s dominant position in AI computing remains unshakable, with both AMD and Intel far behind. Products from some emerging companies like Groq or Cerebras have a long way to go before they’re genuinely commercially competitive and also need more versatility.

However, from a longer-term perspective, NVIDIA will eventually face formidable competitors. This is a fundamental principle of economics – markets with excess profits rarely maintain oligopolistic monopolies in the long run.

Others

Although the overall PC market recovery is slow, NVIDIA’s GPU business growth is significantly faster than the overall PC market. This is likely related to changes in product mix, with some small-scale AI demand providing additional momentum for NVIDIA’s high-end GPU sales. In the long run, NVIDIA’s GPU business growth rate may gradually converge to levels closer to the PC market.

Demand from Omniverse and small-scale AI computing will continue to support revenue growth in the professional visualization product line. I expect this business to maintain an annual growth rate of over 10% for the next two years.

In the automotive business, there was a significant growth rebound this quarter. This is mainly due to fierce competition among Chinese automakers in the autonomous driving systems space. Facing internal competition and pressure from Tesla’s FSD, it’s a foreseeable trend for Chinese automakers to purchase NVIDIA Drive autonomous computing systems in large quantities.

Another trend that is highly favorable to NVIDIA is the rapid development of the robotics industry. I expect robotics to become one of NVIDIA’s key growth drivers in the next two years. Robotics technology forms powerful synergies with NVIDIA’s core strengths in AI computing, autonomous driving, Omniverse, etc., collectively building a robust flywheel that can support NVIDIA’s growth over the next decade.

Conclusion

In just two years, NVIDIA has completed an astonishing transformation, profoundly demonstrating how AI has reshaped the entire tech ecosystem. Today, no one views NVIDIA as merely a gaming GPU company – it has become the absolute protagonist and most prominent winner of the AI era.

Whether NVIDIA’s current dominant position can be sustained is undoubtedly the most critical question to focus on. Given the pace of AI development and the current state of competitors, NVIDIA’s leadership position should be maintained for the next few years. Of course, as the base grows and supply-demand gradually balances, some key indicators will return to more normal “good levels.” From a longer-term perspective, economic principles tell us that excess profits will eventually attract powerful competitors to enter the market. But that might only happen for a few years. And even when that day comes, NVIDIA will still likely hold the largest market share among several strong competitors. In the foreseeable future, NVIDIA will remain a key force in the AI revolution.