Nvidia’s financial results for Q3 2023 caused excitement in the market with revenues of $18.12 billion, a growth of 34% from Q2 and an impressive 206% year-on-year growth. The company appears to offer the right products in a growing market at premium prices. But is this the peak of a rollercoaster journey?
Yole Group has developed in-depth expertise in the computing, and behind that, advanced packaging and memory. The market research & strategy consulting company has released a significant collection of analyses, including technology & market reports, reverse engineering & costing reports, as well as quarterly market monitors. Click HERE to discover this collection.
In its Status of the Processor Industry report, Yole Group estimated the GPU revenue to reach US$22 billion in 2022 and forecast to generate more than US$55 million in 2028. The GPU market experienced a light contraction in 2021, mainly due to a decline in PC GPU shipments and waning enthusiasm for crypto-currency. Projections for 2023 foresee a 3% resurgence to $23.3 billion, propelled by the burgeoning Generative AI trend…
Belinda Dube, Technology & Cost Analyst at Yole SystemPlus and Emilie Jolivet, Director, Semiconductor, Memory & Computing at Yole Intelligence offer you today to take a step back and get a better understanding of Nvidia’s market positioning. Both analysts expose today innovative Nvidia’s technical choices analyzed in the Nvidia H100 Tensor Core GPU reverse engineering & costing report and make the link with its latest financial results.
Over a period of around seven to 10 years, Nvidia has changed. The company has grown its GPU business from being mainly graphics-related and for gaming use, to one that applies to generative AI, cryptocurrency mining and data centers where performance and efficiency are required for training and inference workloads.
The GPUs in the company’s portfolio are convenient and very efficient, although they are expensive. At the moment, the company is able to generate large revenues in the data center market and face little competition from other device providers, such as Intel or AMD.
Despite this period of huge growth, the company is experiencing somewhat of a rollercoaster too, with some restrictions on US companies for the sale of different products worldwide. In particular, restrictions on US goods for sale in China and some parts of the Middle East.
Packaging technology
Nvidia continually adapts processes and packaging technology. One of its latest product releases, the Nvidia H100, uses 2.5D & 3D packaging, CoWoS technology from TSMC, with five HBM on the surface of the component, and using a silicon interposer substrate.
The company is unusual in the market today for continuing to use a full silicon interposer which quickens communication with the GPU for accelerated processing. There is a cost associated with the use of the full silicon interposer but there are benefits in terms of yield because the company has used this process repeatedly and has been able to optimise it.
The HBM in N Nvidia’s H100 product uses TSV which allows for faster communication between the stacked dies. The TSV pitch is smaller compared to the previous generation, this increases TSV density resulting in improved performance and higher bandwidth. However smaller TSV pitch risks reduced manufacturing yields…
Yole Group deeply analysed the latest innovation in the advanced packaging domain. One of its latest technology & market report, High-End Performance Packaging 2023. In this report, the company announces the high-end packaging market to reach $2.2 billion in 2022 and be projected to over $16 billion by 2028. The CAGR is 40%, during this period. “The classical approach combining TSV with µbumps for HBM and 3DS will continue dominating as a solution, while the future will have die-to-interposer HB options…” detail Yole Group’s advanced packaging analysts, in this report.
The AI market
AI and data centers are target markets, and all require high-performance, high-speed operation with a lot of memory.
Nvidia’s ecosystem is select – there are really only three companies selling HBM (Samsung, Micron and SK Hynix). These companies are also benefitting from the growth in the GPU market where the number of HBM per chip is steadily increasing as Nvidia develops its process and packaging technology. Nvidia is able to choose an HBM supplier according to the needs of the client and what memory capacity or performance is needed for a particular GPU.
Nvidia uses TSMC’s 4nm FinFET process technology, combined with advanced packaging technology to increase the number of transistors on a chip to 80billion in a footprint almost identical to the 7nm Ampere A100 with 54 billion transistors. There has been a steady increase from 15 billion transistors in the Tesla P100 and 21 billion transistors in the Tesla V100.
The large amount of memory required adds to the expense of the end product but also supports HBM providers which benefit from the growth fuelled by ChaptGPT and other AI applications.
Financial performance
Nvidia’s revenues have tripled from almost $6 billion last year and the company has reported strong growth of 34% in the last quarter and more than 2005 Q4 2022 – Q3 2023. (which ended October).
Nvidia has been able to capitalise on the AI chip boom. This sector has been boosted by generative AI and large language models, all of which requires more computation performance. There is increased computing workloads in data centers which are at the heart of the growth of Nvidia.
The company is announcing new products, and increased performance products in a growing market. It is in the right market, with the right products at high prices and with almost no competition, although some may emerge in coming years.
Some other companies are developing less expensive chips and buyers are eager to buy but they are slow to demonstrate the right business models and how they can monetize using generative AI.
Therefore, whether that growth will be sustainable for many quarters still remains to be seen.
Already, Nvidia has announced that its next quarter’s results will be higher than Q3 2023 at $20B, +/- 2%. However, it may be that the peak is near, with increasing competition and uncertainties around how much product can be shipped to China and other restricted regions in the world.
Regardless, the company has a huge market capitalization of $1.2 trillion. This large capitalization may enable the company to make acquisitions and development investment in the future in the way that its competition cannot do.
Nvidia’s financial results are, without doubts, very impressive. But more than figures including an amazing growth rate, its business model, its offer and behind that, its technical choice are just unique. They ensure Nvidia’s success.
According to Yole Group, Nvidia is might the company to keep in mind in 2023…
Make sure to stay tuned on www.yolegroup.com to follow the computing, memory and advanced packaging industries with deep added-value analysis!
Belinda Dube serves as a Technology & Cost Analyst at Yole SystemPlus, part of Yole Group.
Belinda’s core expertise is memory technology, especially DRAM and 3D NAND flash memory. At the same time, she also investigates IC technologies as well as advanced packaging.
Belinda’s mission is to develop reverse engineering & costing reports. She also works on custom projects, where she works closely with the laboratory team to set up significant physical & chemical analyses of innovative memory chips. Based on the results, Belinda identifies and analyzes the overall manufacturing process and all technical choices made by the memory makers. The objectives of these analyses are to understand the structure of the device, identify all materials used, and point out the link between functionality and technology selected by the memory company.
In addition, a significant portion of her mission is dedicated to a strategic technology watch, where her aim is to identify innovative memory chips and manufacturing processes. Based on her expertise, Belinda updates internal simulation tools and runs custom training sessions and demos with industrials.
Belinda attends many international trade shows & conferences where she collects valuable information and meets leading memory players. She regularly has an opportunity to reveal pertinent results during key onsite presentations and webcasts.
Prior to Yole SystemPlus, Belinda had the opportunity to work on several R&D projects dedicated to MEMS technologies and new substrates at INSA (Lyon, France).
Belinda holds a master’s degree in Instrumentation & Nanotechnology Engineering from INSA (France).
Emilie Jolivet is Director of the Semiconductor, Memory & Computing Division at Yole Intelligence, part of Yole Group, where her specific interests cover package & assembly, semiconductor manufacturing & computing and software fields.
Based on her valuable experience in the semiconductor industry, Emilie manages the expansion of the technical and market expertise of the Semiconductor, Memory and Computing Team.
In addition, Emilie’s mission focusses on the management of business relationships with semiconductor leaders and the development of market research and strategy consulting activities inside Yole group.
Prior to Yole, after an internship in failure analysis at Freescale (France), she was an R&D engineer for seven years in the photovoltaic business where she co-authored several scientific articles. Then, Emilie worked at EV Group (Austria) as a business development manager in 3D & Advanced Packaging.
Emilie Jolivet holds a Master’s degree in Applied Physics specializing in Microelectronics from INSA (Toulouse, France). She also graduated with an MBA from IAE Lyon.
Related article
What is at the heart of Nvidia’s earnings achievement? – Graphics and AI juggernaut Nvidia continues to exceed expectations, while setting loftier guidance for the coming quarter. With the earnings for the 2nd quarter of its 2024 fiscal year, which corresponds to the three-month period ending July 30, 2023, the company has lived up to its own hype and then some. Revenue guidance for the quarter was $11 billion, which Nvidia exceeded by $2.5 billion, giving the company just more than twice the quarterly revenue than the same period last year… Full article
Belinda Dube Technology & Cost Analyst, Yole SystemPlusThe HBM in N Nvidia’s H100 product uses TSV which allows for faster communication between the stacked dies. The TSV pitch is smaller compared to the previous generation, this increases TSV density resulting in improved performance and higher bandwidth. However smaller TSV pitch risks reduced manufacturing yields…
