news

Nvidia shows off Vera Rubin platform for tokenmaxxing

At an Nvidia lab briefing, Ian Buck pitched Vera Rubin as an AI-factory play: 10x the tokens per watt of a GB200 NVL72 in early CoreWeave runs on DeepSeek-R1, plus a Vera CPU said to double agent speed.

Published 2026-07-21Source: theregister
theregister source artwork

Why it matters

Token prices have a physical floor made of watts. If performance per watt really moves 10x, inference gets cheaper for everyone downstream, but only if suppliers pass the gain along instead of banking it as margin.

Tokenmaxxing read

Buck’s own arithmetic is the tell: 30-40 million developers, roughly $3 trillion in salaries, therefore nine trillion dollars of productivity from a claimed 3x. That is a multiplication, not a measurement, and the same piece notes firms are already capping employee token budgets.

Source takeaway

Vendor briefing, so every performance figure is Nvidia's own and the tokens-per-watt number covers initial CoreWeave runs on a single model. The Register itself flags the productivity math as a thought experiment rather than evidence.

Topic links

Related projects

Tools that match this angle

#2Direct
Observability

Langfuse

langfuse/langfuse

Open-source LLM engineering platform for observability, traces, metrics, evals, prompt management, datasets, and playground workflows.

33.2K3.6KSource-available
tracesevalscosts
#5Direct
Evaluation

promptfoo

promptfoo/promptfoo

A CLI and CI workflow for testing prompts, agents, and RAG systems across models, with evals and red-team style checks.

24.3K2.2KMIT
prompt-evalscirag
#6In spirit
Evaluation

DSPy

stanfordnlp/dspy

A framework for programming and optimizing language-model pipelines rather than hand-tuning one prompt at a time.

37.3K3.2KMIT
optimizationprogrammingevals
Related feed

More source-linked context

404 Media source artwork
news4M
news

Microsoft Tells Engineers ‘Tokenmaxxing Is Not What We Are Optimizing For’

Microsoft EVP Jay Parikh told staff that tokenmaxxing is not the goal, giving divisions AI token budget targets as of July 2026 and making OpenAI's cheaper GPT-5.6 the default model for internal use.

tokenmaxxingexplainerworkplace-ai
Read note
the Guardian source artwork
newsTG
news

Atlassian tightens tracking of staff AI use as other technology firms encourage ‘tokenmaxxing’

Guardian Australia saw an internal memo: Atlassian gave R&D staff monthly AI “wallets” of $500 to $2,000 spanning four tools including Claude Code. Alerts fire near the cap, usage pauses at zero, and no top-up has been refused yet.

tokenmaxxingexplainerworkplace-ai
Read note
AP News source artwork
newsAN
news

Workplaces look for cheaper AI as ‘tokenmaxxing’ fades as a corporate fad

AP reports the tokenmaxxing fad is buckling as costs climb without matching productivity. Moody's AI analytics head Vincent Gusdorf, author of a new report, says bills piled in and teams realized the tools need disciplined use.

tokenmaxxingexplainerworkplace-ai
Read note