Nvidia – Silicon Supreme

Nvidia, the graphics processing unit (GPU) designer, had another standout quarter. Since the beginning of the year, Nvidia’s share price has increased by over 120%. Big tech companies such as Amazon, Microsoft, Alphabet, and Meta are buying Nvidia’s chips for their data centers so that they can enhance their artificial intelligence capabilities. Companies and researchers are then able to train and run their large language models from these data centers. To get an idea on how these big some of these AI investments have been, Meta (owners of Facebook, Instagram and WhatsApp) top of acquired 24 thousand H100 GPU chips, which go for between $25000 and $40000 each, to power their Lambda AI service.

Nvidia released their Q1 2025 results last week. Revenue for the quarter was $26 billion, up 262% compared to the same quarter last year. Net income for Nvidia was $14.9 billion, up 628%. Data center revenue, which includes the revenue recognized from the sale of chips for artificial intelligence, was $22.6 billion, up 427% year on year. Nvidia’s gross margin was 78.4% and its net margin 57.1%.

The H100 platform is currently the top-of-the-line chip platform in the market. However, Nvidia has developed a new platform, Blackwell, expected to be released later this year. The Blackwell platform performs 4x faster in training LLM models and 30x faster in inferencing (where the trained LLMs draw conclusions from new inputs) compared to the H100 platform. Nvidia faces the risk of overdeveloping their chips, where customers may wait out certain generations of chips to get the next generation, which is significantly better. Nvidia, however, is confident that current demand far outstrips current supply for both the H100 and Blackwell chips.

Nvidia developed the CUDA software platform to expand the use of its chips in various applications, including AI. However, CUDA is exclusive to Nvidia’s chips. Due to the heavy reliance on Nvidia’s chips, their limited supply, and high costs, other developers have created an open software platform called Triton. This platform allows developers to run AI applications on an array of other chips.

Nvidia also faces significant competition in the chip design field. They are not only competing with established designers like AMD, but also with well-capitalized companies such as Amazon and Microsoft, who are developing their own chips to reduce their reliance on Nvidia. Nvidia has stated that their chip cycle is one year between each generation, so competing companies are aiming at a moving target going at breakneck speed.

Nvidia, Amazon, and Microsoft are held in the Lunar BCI Worldwide Flexible Fund and by Lunar Capital’s Offshore Portfolio Clients.