Is the AI Subsidy Era Over?
TL;DR · AI Summary
The monthly subscription era for AI services is ending as per-token billing becomes the industry default, with Microsoft canceling internal Claude Code and Uber burning through its full 2026 AI budget in four months showing severe cost pressures facing enterprises.
Key Takeaways
- Microsoft canceled internal Claude Code licenses due to high token-based billing
- Uber exhausted its entire 2026 AI budget in four months, prompting internal warn
- US AI software prices rose 20%-37%, GitHub abandoning fixed rates for usage-base
Outline
Jump quickly between sections.
The shift from monthly subscriptions to per-token billing has become an industry trend for AI services.
Cases of Microsoft and Uber prove that even large enterprises struggle with cost pressure under token-based billing models.
US AI software prices rose 20%-37% with platforms like GitHub shifting to usage-based billing models.
The mismatch between per-seat revenue and per-token computing costs leads to deeper losses with higher usage volumes.
Mindmap
See how the topics connect at a glance.
查看大纲文本（无障碍 / 无 JS 友好）
- AI计费模式转型
- 包月时代结束
- Claude Code取消
- GitHub转向用量计费
- 成本压力显现
- 微软案例
- Uber案例
- 盈利困境
- 席位收入上升
- Token成本激增
Highlights
Key sentences worth saving and sharing.
Microsoft canceled internal Claude Code licenses due to high token-based billing costs, even for a company with nearly unlimited cloud resources.
Uber's CTO internal memo warned that the company burned through its entire 2026 AI budget in just four months.
Green curve (Per-Seat Revenue) shows mild upward trend in per-seat billing, red curve (Per-Token AI Compute Cost) shows exponential rise in token-based computing costs.
As long as pricing units (per-seat) and cost units (per-token) don't match, higher usage volumes lead to deeper losses.
@HedgieMarkets believes: The "subscription era" of AI services is coming to an end, and token-based pricing is becoming the industry default · Microsoft canceled internal Claude Code: The reason was that token-based pricing made costs "untenable," even for a company with virtually unlimited cloud resources. · Uber's CTO internal memo: Warned that the company burned through its entire 2026 AI budget in four months https://t.co/K5BmGs5WgQ" / X
Is the AI subsidy era ending?
@HedgieMarkets believes: The "subscription era" of AI services is coming to an end, and token-based pricing is becoming the industry default · Microsoft canceled internal Claude Code: The reason was that token-based pricing made costs "untenable," even for a company with virtually unlimited cloud resources. · Uber's CTO internal memo: Warned that the company burned through its entire 2026 AI budget in just four months. · Industry pricing changes: US AI software prices have risen 20%-37%, and GitHub is abandoning fixed-rate plans across all products, shifting to usage-based billing. The two paths ahead are both difficult to navigate Path A: Maintain current pricing · Companies reduce AI usage to fit budgets · AI companies' revenue growth slows, while labs need revenue to support pre-IPO valuations Path B: AI companies lower prices · Company usage recovers · Unit economics deteriorate further, expanding losses Hedgie used a classic "profit scissors gap" chart · Green curve (Per-Seat Revenue): Subscription revenue charged per seat, showing mild upward trend; · Red curve (Per-Token AI Compute Cost): Computing costs calculated by token, showing exponential rise; · The two lines create a huge gap on the right side, labeled "Profit Collapse." As long as pricing units (per-seat) and cost units (per-token) don't align, the greater the usage, the deeper the losses. This is exactly the structural problem currently facing "subscription-based AI coding tools" like Claude Code, Codex, and Cursor, which also explains why GitHub is abandoning flat-rate pricing.
Quote
Hedgie
@HedgieMarkets
13h
Microsoft canceled its internal Claude Code licenses this week after token-based billing made the cost untenable, even for a company with effectively infinite cloud resources. Uber's CTO sent an internal memo warning the company burned through its entire 2026 AI budget in just
