Overview
- Huang floated giving engineers an annual allotment of AI tokens roughly equal to half their base salary as a recruiting and productivity lever.
- Tokens are the usage unit for large language models, typically representing part of a word, and model providers bill enterprises based on tokens processed and generated.
- Engineers increasingly ask about compute access during hiring, and some firms view token allotments as a fourth pillar of compensation beyond salary, bonus, and equity.
- Organizations such as Zapier and Kumo AI are tracking token consumption by employee, with real costs already spiking, including a reported single-day $10,000 bill for a Vercel engineer and a venture firm citing tripled AI expenses since late 2025.
- OpenAI’s Sam Altman has promoted a future of Universal Basic Compute and says the core business will be selling tokens, while Microsoft’s Satya Nadella urges measurable public benefits and Nvidia touts leadership in token cost efficiency.