Author: Scott Galloway / Ed Elson / Mia Silverio
Compiled by: Deep潮 TechFlow
DeepSight summary: Nearly 50,000 people were laid off due to AI this year, but more and more companies are discovering that the cost of using AI is higher than human labor. Uber burned through its entire annual AI budget in four months; Microsoft revoked Claude Code licenses for multiple departments; an Anthropic employee used $150,000 in API credits in one month. Scott Galloway believes companies will ultimately shift to Chinese large models that are 10 to 30 times cheaper, forcing Trump to intervene.
Is AI more expensive than the people it replaces?
Nearly 50,000 employees have been laid off this year under the rationale of AI—a figure nearly matching the total for all of 2025. For companies adopting AI, the logic is straightforward: AI can do the work humans do.
But in recent weeks, this logic has hit a wall: an increasing number of companies have found that the actual cost of using AI is higher than the labor it aims to replace.
Chart: Enterprise AI Cost Impact — AI Spending and Cost Feedback from Companies like Uber, Microsoft, Nvidia, and Meta
Uber burned through its entire 2026 AI budget in four months. The COO said it’s becoming increasingly difficult to justify AI expenses internally. Microsoft is canceling Claude Code licenses across multiple departments—for one reason: cost.
An Nvidia executive said that compute costs are now "far exceeding employee costs." Meta, Pinterest, and Spotify all cited rising inference costs as a drag on profitability in their first-quarter earnings reports.
How large are enterprise AI budgets? According to a survey by cloud cost management company CloudZero, 45% of enterprises spent over $100,000 per month on AI in 2025, up from 20% the previous year.
Anthropic has an internal case that is even more extreme: one employee spent $150,000 on Claude Code in a single month. To justify this cost, the engineer would need to accomplish the work equivalent of 11 typical engineers.
In today’s market, the performative value of the word “efficiency” has been consistently rewarded, to the point where companies don’t even need to genuinely measure efficiency. Seventy-nine percent of S&P 500 companies mentioned AI on their recent earnings calls, but only 8% disclosed any AI-related revenue.
Chart: Comparison of AI Mentions by S&P 500 Companies and Actual Revenue Disclosures
The same CloudZero report found that only half of the surveyed companies said they could confidently assess the return on their AI investments. Match Group CEO Spencer Rascoff said AI costs the company $5 million to $10 million annually. When asked about ROI, he said, "I feel like we're benefiting from it, but it's hard to feel."
China's large models will become the biggest winners
Scott Galloway’s assessment is that enterprises will ultimately shift to the cheapest models, which are Chinese large models. Chinese models are 10 to 30 times cheaper than American models.
Data is validating this trend: the share of Chinese models in developer usage has surged from around 1% in 2024 to over 60% this May, with 80% of U.S. AI startups using Chinese open-source AI models.
Chart: Share changes of Chinese large models among developers versus usage by U.S. AI startups