ClaimsJensen Huang

“Over the past two years, the computational demands for inference have increased approximately 10,000 times, while usage has increased approximately 100 times.”

Jensen Huang, NVIDIA CEO, keynote 17 March 2026 Jensen Huang's GTC Speech (Full Text) (PANews)

Factual AI compute, hardware and data

Over the past two years, computational demand for inference rose about 10,000 times while usage rose about 100 times.

Mixed Partly true and partly false in roughly equal measure.

Factual: Statements about the present or the past, checked against the record: did they hold up?

The second judge would not resolve this claim; the verdict shown is the first judge's.

Said for an organisation they lead or work for; it counts on both records.

What would count as true or false

Judged as of 2026-03-17. The claim is hedged ('approximately'), so order-of-magnitude agreement is enough. True: credible data (NVIDIA-cited or independent, e.g., token-volume and compute-per-query figures from major AI providers or industry analyses) shows that over roughly the two years to early 2026, inference compute demand grew on the order of 10,000x (roughly 3,000x–30,000x) and usage (users, queries or tokens served) grew on the order of 100x (roughly 30x–300x). Mostly true: both figures are in the right order of magnitude, but the period or the definition of usage differs somewhat, or one figure is near the edge of those bands. Mixed: one of the two figures is supported and the other is off by more than an order of magnitude or is unsupported. Mostly false: the figures rest on something real, such as reasoning models using many more tokens per query, but the multipliers are unsupported by any data source or greatly overstated (e.g., real growth is only around 10–100x compute). False: the evidence shows inference compute and usage growth far below these levels, or shows compute did not grow much faster than usage. If the only source is Huang's own assertion, with no supporting data, the claim is mostly false at best (unproven, not disproven).

Set before the research began.

Evidence (4)

  1. E1 Nvidia Gtc 2026 Keynote Deck Scribd (third-party upload of NVIDIA GTC 2026 keynote deck) Own record

    “10,000X ChatGPT Compute”

  2. E2 AI Token Demand vs Price Deflation: Why Volume Wins SAVRN (Chad Everett Harris, Founder) · 20 September 2026

    “Google, for example, went from 9.7 trillion to over 3.2 quadrillion tokens a month in 24 months, roughly 330 times”

  3. E3 The Inference Inflection: Why AI's Center of Gravity Has Shifted from Training to Inference Leading.AI (dev.to), open-source research series · 30 April 2026

    “Reasoning models (like o1) require ~100x more compute than standard generation. Agentic systems (like Claude Code) add another ~100x. Combined: ~10,000x.”

  4. E4 OpenAI Statistics — Shuai Guan Shuai Guan (personal blog compiling OpenAI statements) · 21 June 2026

    “DevDay 2025 listed 6 billion tokens per minute on the API platform; in 2026 OpenAI said its APIs process more than 15 billion tokens per minute”

Reasoning

The usage half has rough independent support: E2 reports Google's monthly tokens rising from 9.7 trillion to over 3.2 quadrillion in 24 months, about 330x. That is the right order of magnitude, though just above the 30x–300x band. E4 shows OpenAI API tokens rising from 6 billion to 15 billion per minute, but over a shorter period. The roughly 10,000x compute figure comes from NVIDIA's own keynote deck (E1). The only other source for it is E3, which gets 10,000x by multiplying two estimated factors of about 100x each, not from measured compute data. So one figure is reasonably supported and the other is not shown by data.

Cite this claim

Jensen Huang, “Over the past two years, computational demand for inference rose about 10,000 times while usage rose about 100 times.” (said 17 March 2026). Verdict: mixed. Reliability API, claim c_d00b59d67aaa, method 0.1.0+03d23c0e, dataset release v1.1.2. https://epistacklabs.com/claims/c_d00b59d67aaa

This claim is in dataset release v1.1.2, with the same verdict.

Reply to this claim

For Jensen Huang, or someone speaking for them: a correction or counter-evidence. Replies are read before anything is shown; an approved reply shows your name, role, date, statement and link. Your email address is never shown and is deleted 30 days after the decision (privacy).

More factual claims by Jensen Huang

  1. Mostly true Factual 4 September 2026
    “we invented a technology called a GPU, and basically a GPU does parallel processing tasks”

    NVIDIA invented the GPU, which does parallel processing tasks such as fluid dynamics, particle physics, quantum chemistry, graphics and AI.

  2. Resolved false Factual 1 June 2026
    “Today, NVIDIA’s token cost is the lowest in the world, not by 10%, by X factors, orders of magnitude, all because we did extreme co-design”

    NVIDIA's token cost is the lowest in the world, lower than others by large factors and orders of magnitude, due to extreme co-design.

  3. Mostly true Factual 1 June 2026
    “It’s the first one to be PCI Express Gen 6.”

    Vera is the first CPU to support PCI Express Gen 6.

  4. Mixed Factual 17 March 2026
    “We have demonstrated that NVFP4 can achieve inference without loss of precision, while delivering significant performance and energy efficiency improvements, and it is equally suitable for training.”

    NVIDIA has demonstrated that NVFP4 can achieve inference without loss of precision, with significant performance and energy-efficiency gains, and is equally suitable for training.

  5. Mostly false Factual 17 March 2026
    “overall performance is improved by 35 times, and a new level of inference performance previously unreachable by NVLink 72 is opened up.”

    Combining Groq with Vera Rubin under Dynamo scheduling improves overall performance by 35 times and reaches inference performance NVLink 72 could not previously reach.

All of them

Verdicts are made automatically; how verdicts are made.