As of 2024, NVIDIA has 80% control over the global GPU semiconductor chip market.
At its recent GTC 2024 (March 18-21, San Jose) the company unveiled its Blackwell chip series, the most powerful GPU optimized for AI. The chip houses 208B transistors, and delivers 5x the performance and 4x the memory of its predecessor.
Current competitors include Intel, AMD, and startups on the horizon like Groq, founded by a developer of Google's TPU (tensor processing unit). Groq has developed an LPU (language processing unit) delivering faster, more efficient large language model performance.
OpenAI is also reportedly exploring a massive chip company project, on the scale of trillions of dollars.
Given the global increase in demand for AI chips:
Is it likely a serious competitor will emerge for a slice of NVIDIA's market dominance, or is the company just too far ahead of the competition?
Serious competitor would need to shave off at minimum 15% of NVIDIA's market share in this timeframe.
Resolves Dec 31 2027.
1,000People are also trading
Useful yardstick for the 15% test: AMD's whole data center segment, CPUs included, was $8.5B in the quarter to September 26, against Nvidia's $89.0B in the quarter to July 26, so AMD is under 9% of the two combined even before taking out EPYC. MI450 only began shipping in Q3, and AMD guided Q4 to about $17B of total revenue, so the next two prints matter more than this one. One odd read on the odds from AMD's 10-Q: it valued the OpenAI and Meta warrants at $178.58 a share at September 26, about 28% of the stock price, and those warrants vest only as the customers reach purchase milestones. That's my reading of a blended figure, not a stated probability. Data: AMD's 10-Q and Nvidia's Q2 results, processed by Equibles: https://equibles.com/stocks/amd/documents/d3d5a8f2-b186-473a-a9fc-3ce83cbd32c6?utm_source=manifold
