Klarna's Revenue Per Employee Tripled Since 2022. NVIDIA's Grew 399%. Only One of Those Is About AI.

GTMnow — GTMfund’s editorial arm and the successor to Sales Hacker — published a scorecard arguing revenue per employee has become 2026’s defining efficiency benchmark, while warning against reading every gain as AI-driven. Klarna’s revenue per employee hit $1.24 million in 2026, up 3.6x since 2022, alongside a 49% headcount reduction over the same span. NVIDIA posted $5.14 million per employee, up 399% since 2022 — but on 701% revenue growth and 60% headcount growth, which GTMnow flags as demand-driven, not automation-driven. Shopify cut headcount from 11,600 to 7,600 (down 34%) while roughly doubling revenue, lifting revenue per employee 215%.

The clearest AI-attributable case in the data set is Salesforce: total headcount grew 9% company-wide even as its support organization shrank from roughly 9,000 to 5,000 employees, with AI now handling about half of all support interactions and cutting support costs 17% — lifting Salesforce’s revenue per employee 26%. That pattern — flat or growing company-wide headcount alongside a shrinking, AI-automated function — is close to what Forrester’s labor-market research on customer service specifically found in July: US customer service job postings are now roughly 10% below pre-pandemic levels even as overall US job postings sit above pre-pandemic levels, and Forrester forecasts that 38% of US jobs lost to generative AI by 2030 will come from office and administrative support occupations, customer service included.

GTMnow’s real argument is a caution against a lazy read of the metric: the piece conflates several distinct drivers — labor intensity, distribution model, pricing power, and deliberate headcount strategy — and treating every revenue-per-employee jump as proof of AI leverage risks misreading what’s actually happening inside a given company. Salesforce’s support numbers are a real AI story; Klarna’s and Shopify’s headcount cuts predate most of their current AI tooling; and NVIDIA’s number is mostly a hardware supercycle wearing an AI-efficiency costume. For anyone building a board deck around headcount-to-AI ROI, the Salesforce case is the one with an actual mechanism behind it — the others need a different explanation before they’re cited as one.