Nvidia's earnings are up 206% from last year as it continues riding the AI wave

Original Article ↗ Hacker News Discussion ↗

Nvidia’s latest earnings show revenue up more than 200% year-on-year, driven overwhelmingly by data center GPU sales for AI and machine learning rather than traditional gaming. Commenters attribute its dominance to long-term investment in CUDA and a tightly integrated hardware–software stack, while rivals like AMD and Intel struggle with weaker tooling and ecosystem lock-in despite competitive hardware. Some see Nvidia’s growth and forward valuation as justified by its strategic position in the AI “gold rush,” but warn that high dependence on a single trend and supply-chain concentration pose significant risks.

Nvidia’s Dominance and CUDA “Moat”

  • Many argue Nvidia’s edge is less hardware and more software: CUDA, cuDNN, libraries, and years of tooling and ecosystem work.
  • This creates vendor lock-in for AI/ML; porting away is costly, so competitors struggle despite decent hardware.
  • Some note Nvidia supports the whole stack, even low-end GPUs, which hooks developers early.

AMD and Intel as Competitors

  • AMD is seen as roughly competitive in gaming performance and often cheaper, but behind on AI/ML due to weaker software (ROCm, HIP, PyTorch support).
  • Intel’s ARC is viewed as a serious but very new attempt; they have decades of catching up to do in discrete GPUs.
  • Some say AMD and Intel repeatedly burned devs with short-lived parallel computing SDKs, while Nvidia just kept iterating on CUDA.

Drivers, Platforms, and User Experience

  • On Windows, several posters report AMD drivers and game optimizations as problematic, pushing them back to Nvidia despite higher prices.
  • On Linux, others report the opposite: Nvidia’s proprietary drivers, especially with Wayland, are painful, while AMD’s open drivers are praised.
  • There are mixed experiences; some say Nvidia Linux support has been fine for them.

Gaming vs Datacenter and AI Gold Rush

  • Thread emphasizes that Nvidia’s real money now is datacenter/AI, not gaming; ~80% of revenue cited as datacenter.
  • Nvidia is framed as the “pickaxe seller” for both crypto and AI booms. Even mediocre competitors are a threat now because demand exceeds Nvidia’s supply.
  • Some debate whether AI demand is sustainable; others see ongoing growth and potential emergent breakthroughs.

AMD’s ROCm and Future Prospects

  • AMD is said to be “caught with pants down” on AI but now doubling down on ROCm, PyTorch support, and products like MI300X, with major cloud partnerships.
  • Optimists expect AMD to close the gap; skeptics point to ROCm’s limited hardware support and history of half-hearted efforts.

Valuation, Margins, and Risk

  • Posters correct that the article’s title should refer to revenue up ~206%, while earnings are up ~1276% YoY due to datacenter margins.
  • Debate over Nvidia’s “crazy” P/E: trailing P/E is high, but some say forward P/E and PEG look reasonable given explosive growth.
  • Risks flagged: competition (including specialized accelerators), dependency on Taiwan fabs, and the chance the AI boom cools.