The era of cheap AI is ending. Enterprise marketing teams using AI tools are suddenly facing costs they didn't anticipate. Microsoft recently cancelled internal Claude Code licenses because token-based billing made development prohibitively expensive. Uber burned through its entire 2026 AI budget within four months. These are not isolated cases - AI software prices have jumped 20-37% across the board, forcing a wholesale rethink of marketing tech stack economics.
What changed and why
For years, vendors offered flat-rate pricing models that hid usage cost volatility. That changed overnight as enterprises demanded transparency on actual token consumption. The shift is clear: GitHub has discontinued all flat-rate plans for usage-based billing. This pricing model makes every content generation, image creation, and analytics request visible in real time. Your marketing team may suddenly notice weekly AI costs rising 15% from the previous month.
Worth noting: The shift from flat-rate to usage-based pricing means your AI costs are now directly tied to output volume. More content, more spend. No ceiling.
Marketing team impact: the double hit
Two cost pressures now hit simultaneously. First, content generation via tools like ChatGPT or Gemini has become steadily more expensive per word. Second, image generation tools - DALL-E, Midjourney, and Stable Diffusion APIs - face similar price hikes. A marketing team that generated 1000 images monthly for £200 now pays £250. Combined with rising text costs, the total budget impact compounds fast.
🟢 The Signal
The AI subsidy era is ending. Token-based pricing is forcing enterprises to confront actual costs. Two likely outcomes: enterprises scale back AI usage (slowing revenue for AI labs pre-IPO) or labs cut prices and absorb losses (worsening unit economics). Both paths lead to tighter budgets for marketing teams.
Practical steps for UK SMEs and agencies
Don't wait for the bill to arrive. Start with three concrete steps:
- Calculate your actual per-task AI cost by tracking tokens consumed per piece of content generated. A 500-word product description that cost £2 six months ago might now cost £5 - and at volume, that adds up.
- Compare token prices across vendors to find the most cost-effective solution for standard workloads. Prices vary 3-5x between providers for equivalent output quality.
- Build fallback workflows using local models like ComfyUI for image generation that don't rely on cloud APIs. One-time hardware cost versus recurring cloud fees.
The Gemini Flash opportunity
Google's Gemini Flash model offers a natural hedge for marketing teams. It delivers sufficient quality for most non-critical content at a significantly lower price. Unlike enterprise-focused models, Flash's cost structure aligns better with marketing team budgets. Agencies using it report 25% lower costs for basic content generation without significant quality loss.
When AI is worth it and when it isn't
Stop assuming every task benefits from AI. Run a simple per-task cost audit: compare what AI costs now versus what a freelancer or in-house writer would charge. For high-impact client work, consider a hybrid model where AI handles first drafts and humans refine. That approach cuts costs without sacrificing the quality clients actually pay for.
The subsidy era is over. AI costs now reflect what the compute actually costs to run, and that changes how every marketing team should budget. Agencies that audit their AI spend this month - not next quarter - will come out ahead. Those that keep running last year's playbook at this year's prices won't.