30/03/2026
Reacting to Jensen Huang’s recent statement around how engineers should hit a minimum token usage….
In a recent discussion, Jensen Huang suggested that engineers may be expected to reach a baseline level of AI token usage, framing it as part of how work will be measured in the future. While the direction reflects the growing importance of AI in engineering workflows, it raises important questions around practicality, economics, and performance measurement.
At senior levels, engineers are primarily responsible for leadership, decision-making, and delivering outcomes. Anchoring expectations around token consumption risks shifting focus away from value creation towards tool interaction. From a cost perspective, even aggressive usage scenarios appear unlikely to reach the levels suggested, which further challenges the premise.
More importantly, tying performance to token usage introduces a metric that may distort priorities, encouraging spend over productivity. This creates a potential misalignment with shareholder expectations, where capital allocation is expected to drive measurable outcomes rather than activity-based targets.
A useful discussion point as AI continues to reshape how engineering work is evaluated.
