FinOps Practice at Bilibili: IT Cost Management and Optimization

This article presents Bilibili's approach to IT cost management and optimization through the adoption of FinOps (Financial Operations). It outlines the background of cost reduction as a key focus for internet companies, the challenges of balancing cost, quality, and efficiency, and the results achieved in 2022 where IT spending was reduced by hundreds of millions of yuan.

The historical process of IT budgeting is described, highlighting how business goals are translated into cost targets and the common problems encountered, such as insufficient involvement of the technical platform during budgeting, weak budget control, and lack of business‑level cost visibility.

An efficiency dashboard is introduced that aggregates data from monitoring, asset management, and hybrid‑cloud systems to provide resource utilization metrics for bandwidth and compute resources. The dashboard supports resource‑level forecasting and cost‑target setting.

The concept of FinOps, as defined by the FinOps Foundation, is explained. FinOps brings together engineering, finance, operations, and business teams to make data‑driven spending decisions. Bilibili applies FinOps across both public and private clouds in a multi‑cloud environment.

The FinOps lifecycle is divided into three stages: cost insight, cost optimization, and cost operation. Specific actions include quantifying cost through detailed billing, driving technical and operational cost reductions in parallel, and embedding cost metrics into business proposals and procurement decisions.

Financial concepts such as CAPEX and OPEX are introduced, along with a Total Cost of Ownership (TCO) model that spreads one‑time hardware expenses over the asset’s lifecycle. The article provides formulas for calculating server TCO, depreciation, network equipment costs, and IDC expenses.

Pricing (SKU) design is discussed, showing how Bilibili aligns internal SKU pricing with public‑cloud equivalents and calculates unit prices as TCO divided by theoretical service capacity. Billing is performed as "cost = unit price × usage", with usage measured differently for shared and exclusive resources.

The billing workflow is described: a unified billing system generates cost statements for both public and private clouds, distributes them to business owners, and collects review feedback. Cost allocation relies on CMDB‑derived APPIDs to map resources to projects, departments, and owners.

Cost optimization strategies are summarized, covering bandwidth, server, and public‑cloud cost reductions. For bandwidth, the focus is on reducing bitrate, shifting to cheaper CDN providers, and peak‑shaving. Server optimization emphasizes hardware iteration, virtualization, container‑based pooling, and mixed‑workload placement to improve utilization. Public‑cloud optimization includes selecting appropriate billing models (bandwidth vs. traffic), right‑sizing instances, and evaluating self‑built versus cloud services.

Operational optimization is divided into cost operation (budget control, variance analysis, billing review, cost modeling) and resource operation (monitoring idle resources, reclaiming unused cloud assets, containerizing low‑utilization servers, and improving ancillary resource usage such as SMS).

In conclusion, Bilibili has built a closed‑loop FinOps process—cost insight → cost optimization → operational optimization—that enables business growth without proportional IT cost increase, saving billions of yuan. Future work will focus on real‑time data‑driven decision making and cost prediction.