On March 18, 2024, at its GTC conference, Nvidia announced the Blackwell GPU platform, the successor to the Hopper generation whose H100 chips had become the scarce, sought-after engine of the generative-AI boom. The centerpiece was the GB200 Grace Blackwell Superchip, which pairs two Blackwell GPUs with a Grace CPU over a 900 GB/s NVLink connection. Nvidia claimed the rack-scale GB200 NVL72 system delivered “up to a 30x performance increase compared to the same number of NVIDIA H100 Tensor Core GPUs for LLM inference workloads” while reducing “cost and energy consumption by up to 25x.” Jensen Huang framed the launch in the grandest terms: “Generative AI is the defining technology of our time. Blackwell is the engine to power this new industrial revolution.”
The announcement landed in the middle of an extraordinary demand surge. In its own results for the quarter ended April 28, 2024, filed with the SEC, Nvidia reported record quarterly revenue of $26.0 billion, up 262 percent from a year earlier, with Data Center revenue of $22.6 billion, up 427 percent year over year. Those first-party numbers make the H100 demand story concrete: the chips that train and serve frontier models had become the bottleneck of the entire industry, and Nvidia sat at the center of it. The CUDA software stack that the company had cultivated for a decade and a half (see 2007-cuda) meant that for most labs there was no easy substitute.
The financial consequence was a historic run-up in Nvidia’s stock. Over the course of 2024 the company’s market capitalization rose into the trillions of dollars, briefly ranking it among the two or three most valuable public companies in the world. (Note on sourcing: a specific market-capitalization figure is a daily market price set by exchanges, not a number Nvidia itself publishes in its filings, so this entry describes the rise qualitatively. The first-party numbers cited above are Nvidia’s own reported revenue; valuation milestones widely reported in mid-2024 are not first-party and are not treated as primary here.)
For business readers, Blackwell and the demand behind it marked the moment AI’s “compute race” stopped being an abstraction and became a question of who could secure chips, power, and data-center capacity. The same logic runs straight into the half-trillion-dollar infrastructure commitments that followed (see 2025-stargate), and it explains why control of GPUs, and of the platforms that compete with them such as Google’s TPUs (see 2016-tpu), became as strategically important as the models themselves.
Note on sourcing: Nvidia’s newsroom and SEC archive pages return an HTTP 403 error to automated fetchers. The Blackwell GTC announcement page was fetched and verified live and is cited as the primary source for the product claims and the Huang quote; the first-quarter fiscal 2025 financial results (the source of the revenue figures) were corroborated through search against the canonical nvidianews.nvidia.com release and Nvidia’s SEC Form 8-K for the quarter ended April 28, 2024.