Per seat pricing is hardly dead in the age of AI. In fact, it’s driving the $1B+ vibe coding revoltion.
But the AI tooling industry just hit its first major pricing bump. Cursor’s June 2025 pricing adjustments to charge power users more on a variable / token basis —and the subsequent developer pushback—have exposed the fundamental economic tensions in how AI-powered software companies think about monetization.
This wasn’t just about price increases. It was about the collision between user expectations and economic reality in the AI era. And it may signal the beginning of the end of bargain per-seat pricing for AI tools. The controversy reveals why per-seat pricing, while still preferred by users and widely adopted by AI leaders, creates a dangerous vulnerability: power users can destroy your unit economics.
The Cursor Pricing Fiasco: When Power Users Break the Model
In June 2025, Cursor made what seemed like routine pricing adjustments. The company moved from unlimited usage to tiered consumption limits, fundamentally changing how developers could use their AI-powered code editor. The backlash was swift and brutal.
According to TechCrunch, the changes “upset users” to the point where Cursor felt compelled to issue a public apology for “unclear pricing changes.” But the real story wasn’t about communication—it was about economics. Analytics India Magazine documented a “developer exodus” as at least some power users fled to competitors, highlighting just how sensitive the developer community has become to AI pricing structures.
The company’s own blog post about the June 2025 pricing changes reveals the core tension: “Our users love per seat pricing. It’s just the cost side makes it harder.” This single sentence encapsulates the existential challenge facing every AI company today.
Per-Seat Pricing Isn’t Dead—It’s Still the Gold Standard
Here’s what most analysis of the AI pricing situation gets wrong: per-seat pricing isn’t fundamentally broken. In fact, it remains the preferred model for both users and AI companies. A look at current market pricing shows the industry’s commitment to this approach:
Despite the economic challenges, virtually every major AI coding tool still leads with per-seat pricing because it works for most use cases. Users love the predictability. Finance teams love the budgeting simplicity. Sales teams love the scalable revenue model. For the majority of users who consume AI resources within normal parameters, per-seat pricing delivers exactly what everyone wants: predictable costs and unlimited access within reasonable bounds.
The problem isn’t per-seat pricing itself—it’s the power user tail that can make individual seats economically catastrophic.
The Preference for Per-Seat Pricing Extends Far Beyond Coding Tools
AI companies across categories are gravitating toward hybrid models that combine subscription fees with usage-based billing. Granola, the AI meeting note-taking tool, offers a clean per-seat structure: Free trial, Individual at $18/month, and Business at $14 per user per month. ElevenLabs, despite being a heavily usage-dependent AI voice service, still anchors its pricing around monthly subscription tiers starting at $5/month for the Starter plan, with usage-based billing as an add-on for overages rather than the primary model.
This pattern reveals the industry’s recognition that while usage-based pricing might be more economically pure, per-seat pricing remains what users want and what sales teams can sell effectively.
The problem isn’t per-seat pricing itself—it’s the power user tail that can make individual seats economically catastrophic.
The Power User Problem: When 5% of Users Cost 80% of Revenue
Every AI company discovers the same brutal reality: user consumption follows a power law distribution. Our analysis of realistic usage patterns reveals the mathematical impossibility of sustainable per-seat pricing:
User Distribution Impact (1,000 users at $20/month seat price):
Healthy Mix (60% light, 30% typical, 8% heavy, 2% power users):
- Revenue: $20,000
- Costs: $8,868
- Profit: $11,132 (55.7% margin)
Power User Heavy (40% light, 30% typical, 20% heavy, 10% power users):
- Revenue: $20,000
- Costs: $20,075
- Profit: -$75 (-0.4% margin)
Enterprise Heavy (20% light, 40% typical, 30% heavy, 10% power users):
- Revenue: $20,000
- Costs: $24,288
- Profit: -$4,288 (-21.4% margin)
The math is unforgiving. Once heavy and power users represent more than 10-15% of your user base, per-seat pricing becomes unsustainable. This creates what economists call “adverse selection at scale”—the users who extract the most value are also the ones who make the business unprofitable.
The Venture Capital Masking Effect
The broader AI tooling industry has been masking this power user problem with venture capital. Companies subsidize the heaviest users to drive adoption and land enterprise deals, hoping to figure out the economics later.
Consider GitHub Copilot’s strategy: at $10/month for individuals and $39/month for business users, they’re likely breaking even or losing money on heavy users while profiting significantly from light users. Our analysis suggests typical Copilot usage costs roughly $4.80/month in API calls, leaving healthy margins for most users. But power users consuming 10x the average quickly erode those margins.
But venture subsidies are temporary by definition. As companies approach profitability requirements, they must confront the power user economics problem. The question isn’t whether AI companies will need to address this—it’s when and how.
Cursor’s experience shows what happens when this transition occurs without sufficient preparation. Users who had grown accustomed to unlimited usage suddenly faced consumption limits. The reaction was strong because these users—often the most vocal advocates—felt they were losing something they had already paid for.
Why Pure Token-Based Pricing Misses the Mark For Most
Pure token-based pricing seems like the obvious solution—users pay for exactly what they consume. But this approach solves the power user problem while creating new ones.
Most users don’t want to think about token consumption. They want to focus on their work, not optimize their AI usage to minimize costs. Token-based pricing introduces cognitive overhead that many users find unacceptable.
Moreover, when your entire business model depends on marking up someone else’s tokens, you’re vulnerable to upstream pricing decisions. OpenAI drops their API prices by 50%? Your margins can evaporate overnight. This creates business fragility that many companies can’t afford.
The Hybrid Solution: Seat-Based with Power User Safeguards
The solution emerging from this chaos isn’t abandoning per-seat pricing—it’s making it sustainable by addressing the power user problem directly. This takes several forms:
Tiered Consumption Limits: Keep per-seat pricing for base usage but add consumption tiers for power users. Users get predictable pricing up to a threshold, then pay for additional consumption.
Power User Identification: Proactively identify high-consumption users and migrate them to appropriate pricing tiers before they become economically problematic.
Transparent Overages: When users exceed their seat-based allocation, charge transparent per-token overages rather than cutting off access.
Enterprise Power User Plans: Create dedicated pricing tiers for organizations with known power users, pricing in the expected consumption from day one.
The Token Markup Evolution
For companies that can’t solve the power user problem within traditional pricing, token markup models offer a nuanced alternative. Users bring their own API keys and choose their preferred models, while the software company applies a transparent markup on top of actual token consumption.
This approach preserves user control while ensuring sustainable unit economics. Power users pay for their consumption, normal users pay predictable amounts, and the company’s revenue scales with value delivered rather than seats sold.
The downside? Users don’t take advantage of the large discounts big players like Cursor & Windsurf get on tokens when they bring their own. So in the end, this may result in no savings. Or even — higher costs.
Industry-Wide Implications
Cursor’s pricing crisis is just the beginning. Every AI company currently subsidizing power users will face similar transitions. The key insight is that per-seat pricing isn’t fundamentally broken—it just needs power user safeguards.
The companies that can solve this problem while maintaining the simplicity users love will dominate their markets. Those that force users into complex token-based models will lose to competitors who preserve the per-seat experience for normal users.
Enterprise customers are already demanding this balance. They want predictable per-seat pricing for budget planning, but they also want transparency when power users drive costs beyond normal parameters.
The Strategic Lesson
Cursor’s experience teaches us that the AI industry’s challenge isn’t pricing model selection—it’s power user economics. The CEO’s honest admission—”Our users love per seat pricing. It’s just the cost side makes it harder”—should be a wake-up call for every AI company to identify and address their power user problem proactively.
The data is clear: per-seat pricing works perfectly until power users represent more than ~10% of your base. At that point, unit economics become challenging rapidly. Companies that proactively address this will avoid the user friction that Cursor experienced.
Per-seat pricing will remain the preferred model for most AI tools. The winners will be companies that can preserve this simplicity while building sustainable economics around power user consumption. The challenge is managing this transition thoughtfully rather than reactively.
The future belongs to companies that can demonstrate clear value above and beyond AI model access while maintaining pricing simplicity for normal users and economic sustainability for power users. Cursor’s pricing adjustment may have been bumpy, but it’s accelerating an inevitable industry evolution toward sustainable AI software economics.
