How Nvidia's decade-old bet on GPUs for general-purpose computing is helping the company ride the ongoing AI boom
Nvidia, a maker of graphics processing units, is riding an artificial intelligence boom to put its chips in drones, robots and self-driving cars. Tweets: @nytimes and @semil Tweets: @nytimes : An artificial intelligence boom has helped turn this 24-year-old tech company into Silicon Valley's hottest prospect http://www.nytimes.com/... @semil : What if the *it* company riding the next wave (AL, ML, mining) isn't a tech startup but actually a public incumbent company? $NVDA http://twitter.com/...
Context & Ripple Effects
This 2017 Times piece sits mid-way through a longer arc. A year earlier, Bloomberg had already documented how Nvidia converted demand for GPUs in AI, self-driving cars, and data centers into rapid expansion after missing out on smartphone chips entirely. By 2020, the Wall Street Journal was tracing the same two bets — gaming plus AI — from a niche high-end graphics player in the mid-90s to the most valuable US chip company [[a:958124]].
What the article captures is the inflection moment: Jensen Huang has since framed the company's role through its early decisions leading into the AI boom, and later quarters show both the payoff and the cyclicality — the February 2023 report of revenue down 21% YoY even as Data Center grew 11% YoY to $3.62B — followed by quarterly free cash flow up 18x in three years by 2026 $48.5B being deployed as a capital moat.
First-order effects
- Cloud and data-center operators adopting Nvidia GPUs as their default AI compute lock in the CUDA-era ecosystem, making the same chips that served gamers the substrate for training and, increasingly, drones, robots, and self-driving cars.
- Every new AI workload that lands on Nvidia silicon compounds the installed base, which is exactly the ubiquity Huang argues rivals will find hard to replicate.
Second-order effects
- Inference-focused challengers such as Groq attack the low-latency end of the market, forcing Nvidia into non-exclusive licensing for Groq's inference technology rather than the reported $20B acquisition Huang denied in a leaked email.
- Process technology itself becomes a competitive weapon: sources say Nvidia tested Intel's 18A node for chipmaking but stopped moving forward, keeping manufacturing options open as leverage against suppliers.
Third-order effects
- If the pattern holds, semiconductor value consolidates around companies that own both silicon and the software layer above it, converting chip vendors into something closer to infrastructure utilities than component suppliers.
- Scale itself turns structural: free cash flow at $48.5B a quarter lets Nvidia underwrite capacity, partnerships, and licensing deals that raise the entry bar for any would-be successor, regardless of technical merit.
The trend: General-purpose GPU compute is consolidating AI infrastructure around a single incumbent whose software ecosystem and cash generation reinforce each other every cycle.