AI FinOps④：AI 成本管理服務競爭升温

Bespin Global says it provides a seven-layer full-stack AI offering. (Source: Bespin Global)

Bespin Global and AI Gateway X are also emerging as notable players in artificial intelligence (AI) FinOps.

Bespin Global is expanding its FinPlex multicloud FinOps platform to support AI workloads. The company plans to provide AI cost management and control alongside cloud cost management through FinPlex.

“AI cost management must include unified visibility into GPU and token costs, cost allocation by organization and service, budget and limit controls, real-time anomaly detection and automatic blocking, model routing-based optimization, and value measurement based on unit economics,” said Yoon Hyun-jip, a manager at Bespin Global. “FinPlex already provides visibility, allocation, budgeting and real-time policy controls in multicloud environments, and we plan to gradually expand these capabilities to AI workloads, including GPUs and LLM tokens.”

Bespin Global’s FinPlex Secures FinOps Foundation Certification

According to Bespin Global, FinPlex consolidates costs from five cloud platforms — AWS, Microsoft Azure, Google Cloud, Oracle Cloud Infrastructure (OCI) and Naver Cloud — into a single dashboard. It manages the entire cost lifecycle through three stages: Inform, Optimize, and Operate & Govern. The company said it also combines the platform with managed FinOps services, taking responsibility for cost-saving results rather than merely providing tools.

Bespin Global claims FinPlex is the only solution in South Korea built on a FinOps Foundation-certified platform engine that uses AI to analyze, explain and execute cost management tasks. According to the company, only 29 platforms worldwide, including its own, hold the FinOps Foundation’s FinOps Certified Platform (FCP) certification.

Bespin Global describes FinPlex as a FinOps platform in which AI analyzes, explains and executes tasks. LLMs automatically summarize cost increases and decreases and explain their main causes in natural language. Administrators can also ask questions such as “Why did costs increase last month?” and receive answers accompanied by charts.

AI also handles anomalous cost detection, root-cause analysis and cost-saving recommendations. “FinPlex uses AI to identify unusual cost changes by account, tag and service, trace their causes, recommend commitment purchases and resource rightsizing, and execute those actions upon approval,” Bespin Global said. Through Model Context Protocol (MCP) integration with AI agents, the system also allows agents to directly retrieve, analyze and report cost, anomaly detection, commitment and optimization data.

Bespin Global plans to combine its AI-powered HelpNow Cloud Build & Run service with FinPlex to control costs from the design stage through ongoing operations.

HelpNow Cloud Build & Run is an integrated managed service provider (MSP) offering that covers cloud design and deployment, stabilization, operations and optimization as a single lifecycle. During operations, the OpsNow Pilot AI agent automatically performs cost optimization tasks alongside infrastructure as code (IaC), continuous integration and continuous deployment (CI/CD), and monitoring.

According to Bespin Global, AI agents handle repetitive operations in this environment, allowing customers to receive professional-level cost controls without dedicated operational staff. Integrating deployment and operations also eliminates duplicate costs and optimizes total cost of ownership (TCO). The company said a major South Korean game developer is currently operating a centralized, proactive and real-time cost control model by combining a long-term cloud MSP contract with FinPlex-based FinOps services.

The customer receives daily infrastructure cost comparisons before and after major new game launches for management reporting, while weekly cost change reports and regular billing reviews help identify potential cost risks in advance. Billing reports are automatically generated through AI agent and MCP integration. The company also manages costs based on actual billed amounts reflecting contracted content delivery network (CDN) rates and uses the FinPlex dashboard to track more than 12 petabytes (PB) of annual CDN traffic and associated costs by month, region and account. Bespin Global said the customer has standardized its internal cost reporting data on FinPlex.

AI Gateway X Plans AI FinOps Launch This Month

AI Gateway X is developing an AI FinOps platform alongside its AI gateway service. The company is preparing the platform as a global software-as-a-service (SaaS) offering accessible to overseas businesses, with a launch planned for this month.

“We are developing an AI FinOps service that manages different AI models and APIs through a single gateway layer and provides visibility into AI usage and costs by project, provider and model,” said Lee Jung-woo, CEO of AI Gateway X. He emphasized the need to distinguish cost visibility from cost control.

Conventional cost dashboards provide retrospective visibility into spending after APIs have been used. AI Gateway X said its system goes further by assessing the estimated cost of each AI request and the current budget status before forwarding the request to the provider. Its core design allows requests to be approved or restricted according to enterprise policies and budget limits. The goal is to predict costs before they arise and enforce controls at the point of spending.

Cost management through AI gateways is particularly effective when different departments or business functions use multiple AI providers. Billing structures vary across providers, covering not only input and output tokens but also caching, long-context processing, service types and API features. Model prices also change frequently, making it difficult to calculate actual costs simply by multiplying token counts by unit prices.

“Companies are rapidly moving toward environments where they use multiple AI models and providers according to business objectives, performance and cost,” Lee said. “AI agents, in particular, may invoke models and tools multiple times internally to complete a single user task, making the relationship between the number of user requests and actual AI costs increasingly complex.”

AI Gateway X is likewise focusing on how enterprises can accurately predict and centrally control AI costs in environments using multiple models and services. It positions a centralized gateway that monitors and controls both AI requests and costs as a control point for enterprise AI operations.

To achieve this, companies must be able to identify and track costs by project, model and provider. AI Gateway X said it plans to provide proactive controls that allow businesses to set budgets not only at the enterprise level but also by organization and project, issue alerts when spending reaches specified thresholds, and block requests before they reach providers.

The company also said it normalizes different billing structures across multiple providers and distinguishes between models whose APIs can be called and those whose costs can be reliably controlled. It links specific models, API features and billing conditions to evaluate costs across providers on a consistent basis, excluding models or API functions for which sufficient cost calculation data is unavailable.

The technical ability to call a particular model’s API is different from the ability to accurately predict and control the cost of that request. Without reliable data to validate estimated costs, proactive cost controls lose their effectiveness. While the number of AI models supported by a gateway remains important, AI Gateway X intends to differentiate itself through accurate cost calculations and dependable controls.

“In the future, enterprise AI cost management will evolve toward making real-time decisions about whether a particular AI request justifies the expense,” Lee said. He founded application service provider (ASP) JLogos in 2007 and operated it until 2024 before establishing AI Gateway X.

FinOps Foundation Recommends Continuous AI Cost Governance

Globally, the AI FinOps market is attracting not only AI gateway providers but also data platform companies such as Databricks and Dataiku, as well as network security vendor A10 Networks. The trend reflects the growing need for enterprises to integrate AI FinOps into their internal IT capabilities. Just as DevOps and CloudOps became permanent responsibilities of IT organizations, AIOps and AI cost management are increasingly becoming essential operational capabilities.

The FinOps Foundation refers to this as the FinOps operating model (Ops Model) and recommends regular meetings involving finance, IT and business stakeholders, governance processes that periodically assess the costs and value of new AI spending, and the adoption of AgenticOps practices. It also advises organizations to establish guardrails that rapidly detect cost anomalies and limit financial exposure in generative AI environments characterized by high volatility and unpredictable spending.