Enterprise AI Spend Hits $37B as Coding Tools Capture 55% of Departmental Budgets

Budget is consolidating around proven use cases

Enterprise AI spend reached $37 billion in 2025, representing 6% of the total software market, according to Menlo Ventures' latest research. The standout finding: coding tools now command 55% of departmental AI budgets, dwarfing marketing AI at 9%. This is not incremental growth. Code completion usage grew 5.1x year-over-year, AI app builders 10x, and code agents 36.7x. For procurement teams, this means AI budgets are no longer distributed evenly across functions — they are concentrating in development organizations first.

The shift matters because it signals enterprise buyers have moved past horizontal experimentation and into function-specific deployment. CFOs evaluating AI line items should expect software engineering to absorb the largest share, with clear metrics tied to throughput, not abstract productivity claims. Security teams now face urgent questions about code provenance, IP leakage risk, and repo access controls as AI-generated code becomes a standard input to production systems.

OpenAI reports 900% seat growth and workflow capture

OpenAI disclosed ChatGPT Enterprise seats grew 900% year-over-year, with weekly enterprise messages up roughly 800% since November. The average worker now sends 30% more messages than a year ago. More importantly, usage of custom GPTs and projects — features that embed AI into specific workflows — jumped 19x, and now process about 20% of all enterprise messages. This is evidence of workflow capture, not just chatbot trials.

Sector growth varied widely. Technology companies saw 11x growth, healthcare 8x, and manufacturing 7x. The median sector grew 6x. Reported productivity gains include 40-60 minutes saved per day for typical users, and 60-80 minutes for data science, engineering, and communication roles. 87% of IT workers reported faster issue resolution, and 73% of engineers reported faster code delivery.

For buyers, this raises switching costs. Once custom GPTs are embedded in internal workflows, migrating to Microsoft Copilot or Google Gemini for Workspace becomes a process migration, not just a vendor swap. Governance becomes critical: who controls custom GPT access, how is data retained, and how is output audited? These are procurement questions, not IT questions.

The buy-versus-build pendulum is swinging back

Menlo's data shows 76% of enterprise AI use cases are now purchased from vendors, reversing the 2024 trend when 47% were built in-house. This favors packaged vendors with security controls, measurable ROI, and vendor-managed compliance over internal platform teams assembling AI stacks from scratch.

The implication: procurement will favor vendors that can demonstrate fast deployment and clear functional outcomes. Buyers should expect more pressure to standardize on fewer platforms rather than support fragmented internal projects. This also means more scrutiny on whether AI spend ties directly to engineering throughput, support deflection, campaign velocity, or compliance automation — not vague transformation promises.

Application spend now matches infrastructure spend

Menlo reports application-layer AI spend reached $19 billion, slightly exceeding infrastructure spend at $18 billion. This marks a shift from the model-and-GPU-focused narrative of 2023-2024 to a workflow-and-outcome-focused narrative in 2025. Vendors offering vertical SaaS with embedded AI, workflow automation, or function-specific agents are positioned to capture more budget than generic AI platform vendors.

The competitive landscape is sharpening. GitHub Copilot, Amazon Q Developer, Google Gemini Code Assist, and code agent vendors like Cursor and Replit are competing for the same coding budget. Microsoft Copilot, Google Gemini for Workspace, and Anthropic Claude for Enterprise are competing for the same productivity budget. Buyers should expect consolidation pressure as enterprises reduce the number of AI vendors they support.

What to watch

Three trends will define enterprise AI procurement in the next six months. First, expect more scrutiny on code AI governance as software supply-chain teams face audits on AI-generated code. Second, watch for custom GPT and workflow automation vendors to compete directly with horizontal productivity tools. Third, expect CFOs to demand clearer ROI metrics tied to specific functions, not company-wide productivity claims. The experimentation phase is over. Budget is now tied to measurable outcomes in identifiable workflows.