For most companies, reducing technology costs remains a priority. Nvidia Chief Executive Jensen Huang sees things differently. Speaking on the All-In Podcast following GTC 2026, Huang argued that companies should not be trying to minimise AI token consumption but instead maximise the productivity those tokens unlock. His comments highlight a growing shift in how leading technology firms measure the value of artificial intelligence in software development.
A different way to measure productivity
Huang offered a simple benchmark for evaluating AI investment. If an engineer earning $500,000 a year uses less than half that amount in AI tokens annually, he said he would be “deeply alarmed.”
The remark reflects Nvidia’s belief that AI has become a fundamental productivity tool rather than an optional software expense. Instead of treating token costs as something to minimise, Huang suggested they should be viewed as an investment capable of generating significantly greater returns through faster development, better software quality and increased innovation.
Nvidia itself is reportedly working towards an annual AI token budget of approximately $2 billion for its engineering organisation.
The new economics of software development
Tokens represent the units processed by large language models whenever engineers use AI assistants to write code, generate documentation, analyse bugs or automate repetitive development tasks.
As AI models become increasingly capable, token consumption has become a growing operational expense for technology companies. However, Huang argues that the relevant comparison is not the cost of the tokens themselves but the productivity gains they generate.
If AI enables engineers to complete projects faster, identify software defects earlier or produce higher-quality code, the return on investment may substantially outweigh the additional computing costs.
From cost centre to competitive advantage
Many businesses continue to impose strict limits on AI usage in an effort to control cloud-computing expenses. Nvidia’s approach represents almost the opposite philosophy.
The company views AI inference capacity as a strategic resource that allows highly skilled engineers to operate more efficiently. Under this model, reducing token usage may actually limit innovation by discouraging developers from using advanced AI tools throughout the software development process.
The strategy aligns with Nvidia’s broader position as the world’s leading supplier of AI infrastructure. Greater enterprise adoption of generative AI inevitably increases demand for the graphics processors and accelerated computing systems on which many advanced AI models are trained and deployed.
A changing management philosophy
Huang’s comments also illustrate how executive thinking around artificial intelligence continues to evolve. Early discussions focused heavily on automation and labour replacement. Increasingly, leading technology firms are positioning AI as a collaborative tool that enhances human capability rather than replacing it.
Engineers remain responsible for architecture, judgement and verification, while AI increasingly handles repetitive coding, documentation, testing and debugging tasks.
This combination allows experienced developers to concentrate on higher-value engineering challenges while routine work is completed more rapidly.
The future of enterprise AI
As the cost of AI models continues to decline and their capabilities expand, businesses may increasingly evaluate employees not only by salary and output but also by how effectively they leverage artificial intelligence.
For organisations seeking to compete in software, semiconductor design and advanced engineering, Huang’s message is clear: the objective should not be reducing AI consumption but maximising the value created through it.
If that philosophy spreads beyond Silicon Valley, tomorrow’s technology leaders may judge success less by how much they spend on AI—and more by how much additional innovation those tokens help produce.
Newshub Editorial in Technology – July 14, 2026
Ask NF GPT
If you have an account with ChatGPT you get deeper explanations,
background and context related to what you are reading.
Recent Comments