Nvidia is now more valuable than Amazon and Google

Original Article ↗ Hacker News Discussion ↗

Nvidia’s market value briefly surpassed Amazon and Google, prompting debate over whether its soaring share price and ~95 trailing P/E ratio are justified by AI-driven demand for its GPUs. Commenters highlight Nvidia’s deep software moat (CUDA, ecosystem, datacenter integration) and massive profit growth, but contrast this with geopolitical and supply risks, dependence on TSMC, and the likelihood that AMD, Intel and cloud providers will eventually erode its dominance. Many see echoes of past tech bubbles, arguing that current valuations assume years of exponential growth and that any slowdown in AI spending or credible alternative hardware could trigger a sharp correction.

Valuation, P/E, and Macro Context

  • Many note Nvidia’s trailing P/E near ~90–95 versus much lower ratios for Apple, Google, Amazon, Meta.
  • Some argue this is justified by explosive recent earnings growth and strong forward expectations (annualized forward P/E estimates in the ~30–40 range).
  • Others say even forward P/E is high for a cyclical hardware company and requires continued 2× profit growth, which they see as unlikely.
  • Debate over what drives valuations: money supply (M2) vs risk‑free rates; some link high multiples to loose monetary conditions, others to low bond yields, with disagreement on empirical strength.

Moat: CUDA, Hardware, and Integration

  • Strong consensus that Nvidia’s moat is primarily software: CUDA, mature libraries, tooling, and “it just works” integration with ML frameworks.
  • Hardware advantages cited: tensor cores, FP8/FP4 support, NVLink, SXM, high‑bandwidth memory, Mellanox interconnects, “GPU Direct”/data movement features, and turnkey datacenter clusters.
  • Several stress that performance at scale is increasingly about interconnect and power efficiency, where Nvidia is seen as ahead.

Competition and Alternatives

  • AMD, Intel, TPUs, Apple Silicon, and custom cloud chips are discussed as challengers.
  • Skeptics say AMD/Intel underinvested in software (ROCm, drivers), are late, and still trail H100/H200 in real‑world workloads.
  • Others argue the market is now big enough that multiple giants will invest heavily, eroding Nvidia’s share over time (analogy to Tesla in EVs).
  • Projects like ROCm and ZLUDA (CUDA-on-AMD layer) show some progress but are incomplete, especially for ML.

Market Structure, Supply, and Power Constraints

  • Reports of long lead times for A100/H100 and severe datacenter supply constraints (power, cooling, networking gear, new facilities).
  • Some believe GPU compute is becoming a commodity that will eventually compress margins; others think demand growth will outpace that for years.

Product Segmentation and Gamers vs AI

  • Frustration over Nvidia’s removal of NVLink from consumer GPUs (e.g., 4090), seen as protecting high‑margin datacenter SKUs.
  • Counterpoint: segmentation prevents AI/crypto buyers from vacuuming up gaming cards and keeps some SKUs affordable for gamers.

Monopoly, Antitrust, and Ecosystem Concerns

  • Worry about CUDA lock‑in, lack of strong GPGPU alternatives, and Mellanox/NVLink dominance in interconnects.
  • Comparisons to historic monopolies (Intel, Cisco, AT&T) and speculation about whether antitrust action is appropriate; others say Nvidia’s position is largely earned via R&D.
  • ASML is mentioned as a deeper, EUV‑tooling “monopoly,” though its dominance is framed as technology‑driven rather than abusive so far.

Geopolitical and Foundry Risk

  • Several highlight Nvidia’s dependence on TSMC and associated China–Taiwan geopolitical risk.
  • Some argue Nvidia could shift to Samsung or Intel foundries if needed; others see this as a major unpriced tail risk.

Bubble vs Fundamentals, Investing Takes

  • Some see Nvidia and broader “AI trade” as bubble‑like, likening it to crypto or dot‑com, pointing to high multiples, retail enthusiasm, and concentration in a few names.
  • Others emphasize Nvidia’s real profits, backlogged demand, and structural AI build‑out, arguing price is elevated but not pure “meme stock.”
  • Thread contains conflicting advice: index‑fund Boglehead approach vs concentrated stock‑picking in “obvious winners” like Nvidia; both sides acknowledge risk and survivorship bias.