The Most Anticipated Earnings Call of the Year

Recently, NVIDIA's earnings calls have been hotly anticipated. Tonight, the company is reporting its 2Q25 earnings after the market closes, tying up to be the single most hotly anticipated earnings call of the year . There's even a watch party organized in New York for people to follow the call together.

The Call History

NVIDIA's history of delivering strong results has made its earnings calls a must-watch event. The company has beaten earnings estimates for the last 20 quarters, and the stock has risen after the earnings call in 18 of those quarters. The company's stock has returned 176% over the past year.

As we at MarvinLabs focus on the words rather than only the numbers, we looked into the questions asked by analysts on NVIDIA's earnings calls over the last five quarters.

The Analysts

On average, 8-11 analysts ask questions on any of NVIDIA's earnings calls.

There is a core of six analysts who ask questions on almost every earnings call:

• Stacy Rasgon of AllianceBernstein
• Toshiya Hari of Goldman Sachs
• Joe Moore of Morgan Stanley
• Tim Arcuri of UBS
• Matt Ramsay of TD Cowen
• Vivek Arya of BAML

Other analysts ask questions from time to time. Over the last five earnings calls, ten analysts have asked questions occasionally.

The Questions

"Congrats on the great quarter" - That phrase is a meme for analysts listening to earnings calls. Given the nature of the call, analysts often don't ask questions in the most concise way. Using LLMs to eliminate such fluff, we estimate that the average question asked on an NVIDIA earnings call contains ~63% fluff and ~37% substance.

In terms of topics, our Gen AI has identified the following as the most frequently asked categories of questions on NVIDIA's earnings calls over the last five quarters :

Supply Constraints and Allocation

• [1Q25] Management of supply constraints for the new H20 product in China.
• [4Q24] Reasons for expected supply constraints for the next generation of products, presumably Blackwell, despite easing constraints for Hopper.
• [4Q24] Duration of anticipated supply constraints, potentially extending through calendar 2025.
• [4Q24] Managing product allocation amidst high demand and improving supply to ensure customer readiness and avoid product build-up.
• [3Q24] Fairly allocating products across competing customers, industries, startups, healthcare, and government .
• [3Q24] Impact of China restrictions on Q4 guidance and whether supply constraints are causing shipment redirection.
• [2Q24] Current lead times in the data center and whether redirecting shipments to other regions is reducing these lead times.

Revenue and Financial Projections

• [1Q25] Impact of demand on sales and gross margin in the second half.
• [4Q24] Details on the 40% of revenues from inference, including its growth over the past year.
• [4Q24] Clarification on how reduced lead times for products are affecting the conversion of backlog into revenue.
• [4Q24] Explanation of gross margins returning to the mid-70s.
• [4Q24] Drivers behind the expectation of improved gross margins.
• [3Q24] Expectations for China contributions in Q4.
• [3Q24] Visibility on revenue given the increase in purchase commitments and the impact of China bans.
• [3Q24] Potential for Data Center revenue growth in 2025.
• [2Q24] Breakdown of Data Center revenue between systems and GPUs.
• [2Q24] Factors driving growth: pricing/content versus units.
• [1Q24] Clarification on whether data center sales will continue to grow sequentially in Q3 and Q4 or remain at Q2 levels.

AI and Data Center Business

• [4Q24] Expectations for the data center business in 2024 and 2025 over the past 90 days.
• [4Q24] Newer areas within data center, including software and Sovereign AI.
• [3Q24] How CUDA is enabling acceleration of data processing tasks in AI workloads, including data manipulation before training, between training and inference, and after inference.
• [2Q24] Current number of AI-accelerated servers and future trends in AI server acceleration, in light of the Q3 Data Center outlook of $12 billion to $13 billion.
• [2Q24] Confidence level in the availability of applications or use cases that will allow customers to achieve a reasonable return on their AI investments.
• [2Q24] Assessment of the breadth and depth of AI applications to support a sustained increase in customer investments.
• [2Q24] Reception and momentum of DGX Cloud.
• [1Q24] Key drivers for the sequential increase in data center revenue from April to July.
• [1Q24] Current status of driving acceleration into servers to support AI, given the doubling of data center revenue quarter-on-quarter.
• [1Q24] Role and revenue potential of the AI enterprise software suite in the data center segment.
• [1Q24] The contribution of data center growth from systems supporting DGX Cloud offerings.
• [1Q24] Insights gained about the potential of the DGX Cloud business since its launch.

Demand

• [1Q25] Anticipation of any pause in demand for Hopper and H100 during the migration to H200 and Blackwell products.
• [1Q25] Sufficiency of demand for H100 to sustain growth during the transition period.
• [1Q25] Competition from cloud customers developing internal programs.

Competition

• [1Q25] Potential for cloud customers to become medium to long-term competitors.
• [1Q25] Reasons for the increased demand for GB200 systems compared to HGX boards and GPUs.
• [1Q25] Comparison with existing x86 solutions from partners like Intel and AMD.
• [1Q25] NVIDIA's strategy for maintaining innovation pace amid rising competition from GPUs and custom ASICs.
• [1Q24] Impact of strong demand on the competitive landscape, specifically regarding custom ASICs, other GPU solutions, and alternative solutions over the next two to three years.

Inference and Training Market

• [4Q24] Impact of large language models (LLMs) on inference revenue.
• [4Q24] Methods used to measure inference revenue, given that the same GPUs are used for both training and inference.
• [4Q24] Potential for today's training clusters to be repurposed as tomorrow's inference clusters.
• [3Q24] The evolution of inference workloads with the shift to large language models and NVIDIA's positioning compared to smaller model inference.
• [2Q24] Segmentation of the inference market between small model inference and large model inference.
• [2Q24] Positioning of NVIDIA's product portfolio, including the Grace Hopper Superchip, to address different segments of the inference market.
• [1Q24] Opportunity size of inference versus training in generative AI.
• [1Q24] Inference scaling with usage while training is more one-time.
• [1Q24] Qualitative comparison of inference as a bigger opportunity than training, and the extent of this difference.

Software Business

• [4Q24] Understanding the different components of NVIDIA's software business.
• [4Q24] Identifying the sources of growth within NVIDIA's software segment.
• [2Q24] Explanation of the evolution and critical elements of NVIDIA's software ecosystem.
• [2Q24] Quantification of NVIDIA's lead in terms of person-years invested in their software ecosystem.
• [2Q24] Current run rate and materiality of the software business.
• [2Q24] Impact of the software business on margins.
• [1Q24] Monetization impact of software in the data center business, particularly with the growth of cloud service agreements.

Others

• [1Q25] Implications of full production of Blackwell on shipment timing to customers.
• [1Q25] Impact of Blackwell's production schedule on delivery timelines.
• [1Q25] Impact of liquid cooling complexities on the transition timeline.
• [1Q25] Anticipation of any pause in demand for Hopper and H100 during the migration to H200 and Blackwell products.
• [1Q25] Sufficiency of demand for H100 to sustain growth during the transition period.

The Hype

It's a rather interesting question why the call is so hyped. NVIDIA is the company of the year, having grown to be a Top 3 company in the world by market capitalization. Before opening on Aug 28, 2024, the company has a market capitalization of over $3tn, trailing only Apple and Microsoft. The company's stock has returned 176% over the past year.

Unlike most other major technology companies, NVIDIA's results offer ample potential for a read across the whole sector. The company's most important customers are the other major technology companies, and its results can provide a good indication of the sector's health.

Another factor helping the hype around the earnings call is the timing of the call. NVIDIA has a non-standard financial year ending in January rather than in December. As a result, it reports earnings about one month after companies have a regular financial year. Having its earnings call in August rather than July (like the other Magnificent Seven) makes NVIDIA's earnings stand out and not overshadowed by the different companies. Compare that with Microsoft, META, Alphabet, and Apple, which all reported earnings in the last week of July, with the previous two reporting earnings the same day.

See You on the Call

We are looking forward to the call and will share our thoughts tomorrow and a list of the most exciting and relevant questions asked on the call. Stay tuned!