Nvidia’s (NVDC34) skyrocketing chip sales, fueling AI models like OpenAI’s GPT-4, will be a major tech story in 2024.
Over the past year, Nvidia’s stock value surged 200%, briefly pushing its market cap above $3 trillion. This growth places Nvidia among the elite companies valued at over $2 trillion.
Many wonder who can challenge Nvidia, which controls 80% of the AI chip market. However, analysts recently told Fortune that this question is misguided.
Daniel Newman, CEO of Futurum Group, asserts there’s currently no “natural competitor for Nvidia in the market.”
Nvidia’s GPUs, originally designed in 1999 for fast 3D graphics in PC games, are perfect for training massive generative AI models.
Companies like OpenAI, Google, Meta, Anthropic, and Cohere need extensive AI chips for model training. Nvidia’s GPUs are the most powerful and have been in high demand for years.
These GPUs come with a hefty price tag. Training top AI models require tens of thousands of high-end GPUs, each costing between $30,000 and $40,000.
Elon Musk indicated that training his company’s X.ai Grok 3 model would need 100,000 of Nvidia’s best GPUs, equating to over $3 billion in revenue for Nvidia.
Nvidia’s success also stems from its software ecosystem. Nvidia‘s CUDA platform has become the standard for AI developers.
It creates a ‘virtuous cycle’ where more users lead to more ecosystem investment, attracting even more users. This entrenched position is difficult for competitors to overcome.
AMD, which controls about 12% of the global GPU market, offers competitive products and improves its software. However, it lacks Nvidia’s established developer base.
Cloud service providers like Amazon’s AWS, Microsoft Azure, and Google Cloud develop proprietary chips but don’t aim to replace Nvidia.
They seek a variety of AI chips to optimize infrastructure and offer cost-effective cloud services.
Navigating Nvidia’s Dominance in AI Chips
Jack Gold of J. Gold Associates notes that Nvidia’s early advantage in a fast-growing market makes it hard for others to catch up.
Similarly, Matt Bryson of Wedbush highlights the challenge of replacing Nvidia’s chips for large-scale AI model training, a dynamic unlikely to change soon.
Startups like Cerebras, SambaNova, Groq, and Etched aim to capture small portions of Nvidia’s market.
They focus on specialized AI needs, particularly in “inference” tasks, where models generate information from trained data.
For example, Etched recently raised $120 million to develop a chip for running Transformer models.
As Nvidia continues to innovate, analysts see potential areas where competitors might thrive, particularly in energy efficiency, a significant cost for AI training and operations.
Lower energy requirements could offer competitive alternatives. Nvidia also faces antitrust scrutiny.
Investigations by French and US regulators could potentially slow its dominance and open opportunities for rivals.
While Nvidia remains the leader, the growing AI chip market may accommodate various players.
If AI becomes as significant as expected, a healthy ecosystem of chip manufacturers and software developers will likely emerge to address specific application needs.
Read More from The Rio Times