Skip to main content
Mintive is the new name for MintMCP.Read the announcement
Jiquan Ngiam

28 LLM Spend Statistics (2027): What Enterprises Spend on Models and Where It Goes

Data-driven analysis of enterprise LLM investments, spending allocation patterns, and the governance infrastructure required to control AI costs at scale

Enterprise generative AI spending has exploded from $11.5 billion in 2024 to $37 billion in 2025, yet 62% of organizations remain stuck in experimentation. This reveals a critical market paradox: massive financial commitment coexists with widespread implementation failure. Mintive's LLM Gateway (formerly MintMCP) addresses this gap by routing all model requests through centralized access controls, enabling per-team budgets and model restrictions while providing the cost visibility security and finance teams require. The statistics below reveal where enterprise AI budgets flow, which categories capture the largest share, and why governance infrastructure determines whether organizations achieve ROI or burn capital on ungoverned AI sprawl.

Key Takeaways​

  • Market growth validates enterprise AI as strategic infrastructure - Gartner forecast worldwide AI spending at nearly $1.5 trillion in 2025, with updated projections of $2.60 trillion in 2026
  • Foundation model APIs represent the largest infrastructure category - API spending captured $12.5 billion in 2025, making model cost governance a finance priority
  • The execution gap separates winners from wasters - Only 7% of enterprises have fully scaled AI, while 62% remain in experimentation
  • AI coding tools dominate departmental spend - Coding assistants captured $4.0 billion, representing 55% of departmental AI budgets
  • Security concerns grow faster than budgets - 83% of AI leaders express major concern about generative AI security, an 8x increase over two years
  • Agent adoption accelerates governance urgency - AI agent deployment jumped from 11% to 42% in just two quarters

Market Growth and Infrastructure Spending​

1. Enterprise Generative AI Spending Reached $37 Billion in 2025, Up 3.2x From $11.5 Billion in 2024​

The 3.2x year-over-year growth in enterprise generative AI spending demonstrates that organizations have moved beyond pilot programs into production deployments. This acceleration reflects competitive pressure as early adopters demonstrate measurable productivity gains. The spending increase spans foundation model APIs, AI applications, and supporting infrastructure, with enterprises purchasing rather than building 76% of their AI solutions.

2. AI Infrastructure Spending Reached $89.9 Billion in Q4 2025, a 62% Year-Over-Year Increase​

Quarterly infrastructure data shows $89.9 billion flowed into AI hardware and compute resources in a single quarter. This spending supports model training, inference at scale, and the data center expansion required to meet enterprise AI demand. The 62% growth rate signals sustained investment momentum rather than a one-time spike, with servers accounting for 97.6% of quarterly infrastructure spend.

3. Full-Year 2025 AI Infrastructure Spending Totaled $318 Billion, More Than Double the 2024 Figure​

Annual infrastructure investment reached $318 billion, representing a 2.1x increase from the $153 billion recorded in 2024. This doubling reflects enterprise commitment to AI as core business infrastructure rather than experimental technology. The spending encompasses compute clusters, networking, storage, and the specialized hardware required for large-scale model deployment.

4. Gartner Forecast Worldwide AI Spending at Nearly $1.5 Trillion in 2025​

Gartner's forecast positions worldwide AI spending at nearly $1.5 trillion, encompassing AI software, services, infrastructure, semiconductors, and AI-enabled devices such as PCs and smartphones. This figure spans both consumer and enterprise technology markets rather than enterprise LLM spending alone. Organizations should distinguish this broader market estimate from their direct spending on models, applications, and governance infrastructure.

5. Gartner's Updated 2026 Projection Shows Worldwide AI Spending Reaching $2.60 Trillion​

Gartner's updated 2026 projection shows worldwide AI spending reaching approximately $2.60 trillion, representing 47% year-over-year growth. This trajectory validates AI as a multi-year strategic priority for enterprises across industries. Organizations planning 2027 budgets should anticipate sustained investment requirements rather than temporary initiatives.

6. AI Infrastructure Spending Is Projected to Reach $487 Billion in 2026​

Infrastructure-specific forecasts show $487 billion flowing into AI compute and hardware in 2026, representing 53% year-over-year growth. This projection accounts for expanded GPU clusters, specialized AI accelerators, and the networking infrastructure required for distributed training and inference. The growth rate suggests infrastructure spend may outpace overall AI market expansion.

7. AI Infrastructure Spending Is Forecast to Exceed $1 Trillion by 2029​

The long-term projection establishes AI infrastructure as a trillion-dollar category within four years. This trajectory reflects both expanded enterprise adoption and the increasing compute requirements of larger, more capable models. Organizations building multi-year AI strategies should plan for sustained infrastructure investment alongside application and governance spending.

Enterprise Spending Allocation​

8. AI Applications Captured $19 Billion in 2025, Representing More Than Half of Generative AI Spending​

Application-layer spending reached $19 billion, confirming that enterprises prioritize ready-to-deploy solutions over building from scratch. This category spans coding assistants, customer service automation, content generation tools, and vertical-specific AI products. The buy-over-build preference creates opportunities for platforms providing governed access to enterprise tools and data through centralized Virtual MCP endpoints.

9. Foundation Model APIs Captured $12.5 Billion in 2025, the Largest Single Infrastructure Category​

API spending for models from OpenAI, Anthropic, and Google reached $12.5 billion, making model costs a material line item for enterprise finance teams. This category covers spending on foundation model APIs, while Menlo's broader market estimate excludes separate inference and model-serving revenue. Cost tracking by model, user, and agent becomes essential when API spend reaches this scale.

10. The AI Infrastructure Layer Captured $18 Billion in 2025, Up 2.0x From 2024​

The broader generative AI infrastructure layer reached $18 billion in 2025, doubling from $9.2 billion in 2024. This category includes $12.5 billion in foundation model APIs, $4.0 billion in model training infrastructure, and $1.5 billion in AI data and orchestration infrastructure. These figures represent a different market scope from IDC's hardware-focused AI infrastructure spending estimates.

11. 37% of Enterprises Spend Over $250,000 Annually on LLMs​

Enterprise survey data shows 37% of organizations exceed the $250,000 annual threshold for LLM consumption. At this spending level, uncontrolled usage by individual teams or ungoverned agents can materially impact departmental budgets. Organizations at this scale benefit from per-team budgets and model restrictions that prevent cost overruns while maintaining development velocity.

12. 73% of Enterprises Spend More Than $50,000 Annually on LLMs​

The broader spending distribution shows 73% of enterprises investing at least $50,000 per year in LLM services. This threshold indicates LLM usage has become a standard enterprise technology expense rather than experimental discretionary spending. The concentration of organizations above this level justifies investment in governance infrastructure that provides visibility and control over model consumption.

13. The Average Enterprise Invests $1.9 Million in Generative AI Initiatives​

Investment analysis shows the average enterprise commits $1.9 million to generative AI programs, encompassing model costs, infrastructure, personnel, and integration expenses. This figure represents the total cost of ownership beyond direct API fees, including the security, governance, and operational support required for production deployments.

Spending by Application Category​

14. AI Coding Spend Reached $4.0 Billion in 2025, Representing 55% of Departmental AI Spend​

Coding assistants including Claude Code, Cursor, and GitHub Copilot captured $4.0 billion in enterprise spending, dominating departmental AI budgets. This concentration reflects measurable productivity gains from AI-assisted development. Organizations deploying multiple coding tools require visibility into what coding agents do across Claude Code, Cursor, and Codex to understand usage patterns and control costs.

15. Departmental AI Spending Hit $7.3 Billion in 2025, Up 4.1x Year Over Year​

Spending on AI tools deployed to specific departments reached $7.3 billion, growing faster than the overall market. This category includes coding tools, marketing automation, sales intelligence, and customer support AI. The 4.1x growth rate indicates departments are adopting AI faster than centralized IT can govern, creating shadow AI risk.

16. Vertical AI Solutions Captured $3.5 Billion in 2025, Nearly 3x the 2024 Investment​

Industry-specific AI applications reached $3.5 billion, tripling year-over-year. This category spans healthcare AI, legal document analysis, financial services automation, and other specialized applications. Vertical solutions often connect to sensitive enterprise data, making tool-level permissions and audit trails essential for compliance.

17. Healthcare AI Spending Reached Approximately $1.5 Billion in 2025​

Healthcare-specific AI investment reached $1.5 billion, tripling from $450 million in 2024. This growth reflects AI adoption for clinical documentation, diagnostic support, and administrative automation. Healthcare organizations require HIPAA-compliant infrastructure with audit logging and SIEM export capabilities for regulatory compliance.

18. Horizontal AI Captured $8.4 Billion in 2025, Expanding 5.3x Year Over Year​

Cross-functional AI tools spanning multiple departments reached $8.4 billion in spending, growing faster than any other category at 5.3x. Horizontal AI includes general-purpose assistants, knowledge management tools, and productivity applications used across the organization. This distribution pattern creates the need for role-based tool access that can govern usage across diverse teams and use cases.

Adoption, ROI, and Implementation Challenges​

19. 78% of Organizations Use AI in at Least One Business Function​

Adoption data shows 78% of organizations use AI in at least one business function, up from 55% in 2023. This mainstream adoption transforms AI from competitive advantage to competitive necessity. Organizations without governance infrastructure risk falling behind as competitors gain compounding benefits from earlier, more controlled deployments.

20. 62% of Enterprises Remain Stuck in the Experimentation Phase​

Despite high adoption rates, 62% of enterprises have not progressed beyond pilots and experiments. This execution gap reveals that spending alone does not determine success. Organizations that scale successfully typically address governance, security, and cost control early rather than treating them as afterthoughts.

21. Only 7% of Enterprises Have Fully Scaled AI Across Their Operations​

The 7% full-scale deployment rate highlights the difficulty of moving from experiments to enterprise-wide production. The gap between adoption and scaling often stems from missing governance infrastructure: no centralized access controls, no audit trails for compliance, no cost visibility for finance. Centralized MCP governance addresses these barriers by providing the controls required for enterprise-wide deployment.

22. Organizations Reporting Positive AI Returns Average $3.70 Per Dollar Invested​

Organizations reporting positive AI returns average $3.70 back for every dollar invested, with top performers reaching $10.30. This variance emphasizes that implementation quality determines outcomes. Organizations with centralized governance, clear cost attribution, and security controls achieve higher returns than those with ungoverned AI sprawl.

23. 46% of Enterprises Cite Integration With Existing Systems as the Main Challenge​

Implementation research identifies system integration as the primary barrier to AI deployment success. AI agents need secure, governed connections to enterprise databases, APIs, and internal tools. Private network tunnels that reach on-premises systems without public exposure address this challenge while maintaining security controls.

24. 83% of AI Leaders Express Major Concern About Generative AI Security​

Security concern among AI leaders has grown 8x over two years, with 83% now expressing major or extreme concern. This shift reflects awareness that AI agents can access, process, and potentially exfiltrate sensitive data. Runtime security screening for prompt injection, PII detection, and secrets exposure addresses these concerns at the tool-call level.

AI Agents and Autonomous Systems​

25. AI Agent Adoption Jumped From 11% to 42% in Just Two Quarters​

The rapid acceleration in agent deployment indicates enterprises are moving beyond assistants to autonomous systems. This shift creates new governance challenges: agents make decisions, call tools, and access data without human review of each action. Per-agent identity and scoped permissions become essential when autonomous systems operate at scale.

26. 84% of Enterprise Leaders Expect to Increase Spending on AI Agents in the Next 12 Months​

Forward-looking investment data shows 84% of leaders planning increased agent spending, validating autonomous AI as a strategic priority. This investment trajectory makes agent governance infrastructure a prerequisite rather than an optional addition. Organizations planning agent expansion benefit from agent identities with independent credential rotation before deployment scales.

27. 72% of Organizations Plan to Increase Their LLM Spending in 2025​

Budget projections show 72% of organizations planning higher LLM investments, indicating sustained growth rather than plateau. This spending increase makes cost governance more important as budgets expand. Organizations without visibility into model consumption by user, agent, and session risk budget overruns as usage scales.

28. The Enterprise LLM Market Is Valued at $8.65 Billion in 2025, Projected to Reach $33.72 Billion by 2035​

Market sizing establishes the enterprise LLM segment at $8.65 billion, with 10-year projections showing nearly 4x growth. This trajectory confirms enterprise LLM adoption as a sustained trend rather than a temporary spike. Organizations building AI infrastructure should plan for multi-year scaling rather than point solutions.

How Mintive Addresses Enterprise LLM Governance Challenges​

The statistics above reveal a consistent pattern: enterprise AI spending grows faster than governance capabilities. Organizations spending millions on models, infrastructure, and applications often lack the visibility to know which teams consume what resources, the controls to enforce budgets and policies, and the audit trails required for compliance.

Mintive (formerly MintMCP) provides the governance infrastructure layer that addresses three connected challenges enterprises face as AI spending scales:

Cost Control and Budget Enforcement​

With foundation model APIs representing a $12.5 billion category and 73% of enterprises spending over $50,000 annually on LLMs, organizations need granular cost management. Mintive's LLM Gateway routes all model requests through centralized controls, enabling:

  • Per-team budget limits that prevent departmental overruns without blocking development velocity
  • Model-level restrictions that enforce approved model usage by role and team
  • Usage tracking that attributes spend to specific users, agents, and sessions
  • Provider flexibility that allows model swapping without changing application code

This centralized routing approach gives finance teams the visibility they need while maintaining the developer experience teams expect.

Security and Compliance Infrastructure​

The 83% of AI leaders expressing major security concerns recognize that agents accessing enterprise data create new attack surfaces. With AI agent adoption jumping from 11% to 42% in just two quarters, security infrastructure must keep pace with deployment velocity. Mintive's Mint Guard provides runtime security controls that:

  • Screen for prompt injection and jailbreak attempts at the tool-call level
  • Detect PII exposure before sensitive data reaches external model APIs
  • Block risky commands that could access unauthorized systems or data
  • Generate audit trails with SIEM export for compliance teams

Healthcare organizations requiring HIPAA compliance and financial services requiring SOC 2 attestation benefit from these built-in security layers rather than building custom solutions.

Operational Visibility Across AI Tools​

When 62% of organizations remain stuck in experimentation, the difference often comes down to whether IT and security teams can see what AI systems do. Mintive's Agent Monitor provides visibility into supported coding-agent environments, including Claude Code, Cursor, Codex, and GitHub Copilot, capturing prompts, commands, file access, and MCP tool calls where supported by the client and integration.

This visibility enables:

  • Shadow AI detection that identifies off-gateway usage patterns as departmental spending grows 4.1x year-over-year
  • Tool-level permissions that govern which agents can access which enterprise resources
  • Role-based access control that scales governance across diverse teams and use cases
  • Per-agent identities with independent credential rotation as autonomous systems multiply

Integration With Enterprise Systems​

With 46% of enterprises citing integration as their main challenge, production AI requires secure connections to enterprise databases, APIs, and internal tools. Mintive's private network tunnels reach on-premises systems without public exposure, addressing the gap between AI agents' need for data access and security teams' requirement for controlled connectivity.

Higher AI returns require more than model spending alone. Early investment in access controls, cost visibility, and auditability can help enterprises manage risk as deployments scale. Organizations that deploy governance alongside AI tools avoid the bottleneck that keeps 62% stuck in experimentation while the successful 7% scale enterprise-wide.

Frequently Asked Questions​

What are the primary cost drivers for enterprise LLM adoption?​

Menlo Ventures estimates $19 billion in enterprise generative AI application spending and $18 billion in infrastructure-layer spending in 2025. The infrastructure total includes $12.5 billion for foundation model APIs, alongside model training and AI data infrastructure. Organizations should budget for direct model consumption, the applications built on those models, and the governance infrastructure required to control usage at scale.

How can companies effectively manage and reduce their LLM API spending?​

Effective cost management requires visibility into consumption by user, agent, model, and session combined with policy controls that enforce per-team budgets. Organizations using centralized LLM routing can track spend, restrict model access by role, and swap providers without changing application code.

What is shadow AI and what are its financial implications for enterprises?​

Shadow AI refers to AI tools adopted outside IT governance, including personal accounts for ChatGPT, Claude, or coding assistants. With departmental AI growing 4.1x year-over-year, ungoverned adoption can represent significant untracked spend. Shadow AI detection capabilities identify off-gateway usage patterns.

How do governance platforms contribute to optimizing LLM costs?​

Governance platforms provide the visibility to understand where spend occurs and the controls to enforce budgets. This includes model-level routing, per-team spending limits, usage dashboards, and audit trails. Organizations with centralized governance achieve higher ROI by preventing waste while maintaining development velocity.

Why does the execution gap persist despite high enterprise AI adoption?​

The 62% stuck in experimentation often lack the governance infrastructure to move from pilots to production. Without centralized access controls, audit trails, and security screening, compliance teams block expansion. Organizations that deploy governance alongside AI tools avoid this bottleneck.