Token Shock Just Broke the Enterprise AI Playbook
Uber burned through its entire 2026 AI coding budget in four months. A major French insurer pulled the plug on Claude in two. Welcome to the era of “token shock” – where the per-seat SaaS model that built a trillion-dollar software industry is quietly dying on the CFO’s desk, and the Build vs. Buy vs. Partner decision has become the defining strategic call of the enterprise AI era.
New AIMG research reveals a market that has crossed the Rubicon – and a commercial rulebook being rewritten in real time.
The production wave is here.
55% of enterprises are already running AI in production. Another 35% ship within six months. Only 1% remain on the sidelines beyond a year. Experimentation is officially over.
Buy is beating Build – decisively.
68% of enterprises succeed with third-party solutions. Only 18% go pure in-house. Licensing specialized platforms can cut data science headcount by 50-75% and save north of $1M annually – making Buy the default path for non-core workflows, despite lock-in risk.
Partner is the scale multiplier.
The SIs are weaponizing talent at industrial scale: Accenture is training 30,000 people on Anthropic; EPAM is targeting 10,000 Claude-certified architects. Microsoft leads enterprise mindshare, trailed by OpenAI, Google Gemini, Anthropic, and AWS.
Budgets are new, not recycled – yet.
91% of organizations are creating entirely new AI budget lines rather than raiding IT. But the cannibalization is coming: 39% expect to displace traditional software spend, and 35% will offset non-software IT.
The per-seat model is finished.
When one human orchestrates dozens of agents running 24/7, seats stop making sense. 37% of CIOs now prefer usage-based pricing versus just 21% for seat-based. AIMG projects subscription pricing collapsing from 60% to 30% of SaaS this decade, while outcome-based pricing rockets from 10% to 60%.
The bottom line
Build gives you control but starves for talent. Buy delivers speed but exposes you to token shock. Partner scales expertise but creates dependency. In H2 2026, the winning enterprises aren’t picking one – they’re engineering a portfolio across all three, with consumption economics at the core.
Token shock isn’t a pricing problem. It’s a strategy problem.
Read the full AIMG report: Enterprise AI Strategy – Navigating the Build vs. Buy vs. Partner Decision Framework