NVIDIA

NVIDIA is a semiconductor company founded in 1993 that originally designed graphics processing units for gaming. Its GPUs turned out to be exceptionally well-suited for the parallel matrix operations that underpin neural network training and inference, and NVIDIA pivoted aggressively into AI hardware starting in the early 2010s.

The company's dominance in AI rests on three pillars: hardware, software, and ecosystem lock-in. Its A100 and H100 GPUs became the standard training accelerators for large language models, and the H200 and B200 followed as inference-optimized successors. CUDA, NVIDIA's proprietary parallel computing platform, is the software layer that most deep learning frameworks — PyTorch, TensorFlow, JAX — are built on top of. Switching away from NVIDIA hardware means rewriting code that depends on CUDA, which keeps customers on the platform even as competitors like AMD, Intel, Google (TPUs), and startups like Groq and Cerebras offer alternatives.

NVIDIA's market capitalization surpassed $3 trillion during the AI boom, briefly making it the most valuable public company in the world. The company's data center revenue — driven almost entirely by AI workloads — grew from $15 billion in 2023 to over $100 billion annually. Whether this concentration of AI infrastructure in a single vendor is sustainable or healthy for the ecosystem is an open question the industry is actively debating.