Easily build complex reports
Monitoring and efficiency metrics
Custom cost allocation tags
Network cost visibility
Organizational cost hierarchies
Budgeting and budget alerts
Discover active resources
Consumption-based insights
Alerts for unexpected charges
Automated AWS cost savings
Discover cost savings
Unified view of AWS discounts
COGS and business metrics
Model savings plans
Collaborate on cost initiatives
Create and manage your teams
Automate cloud infrastructure
Cloud cost issue tracking
Cloud data through AI
Detect cost spikes
Fresh off the quarterly earnings for Microsoft, Google, and Amazon, we are releasing the Q1 2025 Cloud Cost Report, an analysis of cloud usage based on anonymized Vantage customer usage. Vantage is a cloud cost visibility and optimization platform, with a unique view into industry trends, thanks to tens of thousands of connected infrastructure accounts across 20+ cloud providers. To discuss this report in more detail, join our growing Slack Community of over 1,000 engineering leaders, FinOps professionals, and CFOs. View past reports here.
Get a PDF of this report.
Protected by reCAPTCHA. The Google Privacy Policy and Terms of Service apply.
AWS, GCP, and Azure have increased their year-over-year run rate by an estimated 17%, 28%, and 35% respectively. While AI services are growing in percent of spend, the reality is that staple cloud services are still making up the majority.
Services like compute, storage, and databases are still driving the majority of cloud spend. Other supporting features like logging are growing as well, with Amazon CloudWatch, Google Cloud Logging, and Azure Log Analytics all showing increased percent spend from last quarter. It makes sense that logs are expanding because as more services are used, more logs are produced proportionally, creating a compounding effect where infrastructure growth directly drives observability costs higher.
Input/Output requests make up 83% of OpenAI spend.
Over 90% of Azure Virtual Machine Spend comes from On-Demand instances.
Only 37% of S3 Storage spend is on non-Standard tiers.
More than 40% EC2 spend comes from just three instance families. (see full section).
Non-Production accounts for 23% of cloud costs. (see full section).
As cloud costs trend up, observability’s share of total spend trends down. (see full section).
The data reveals significant differences in GPU adoption and spending patterns across cloud providers. GCP leads with GPU instances representing 14% of total spend for organizations using GPU resources, followed by AWS at 8%, while Azure lags considerably at just 1.5%.
AWS maintains a solid middle position at 8%, which aligns with their broad enterprise customer base gradually adopting GPU instances for machine learning and data processing tasks.
Data transfer costs represent a significant and often overlooked component of AWS spending, with nearly 90% of transfer costs concentrated in just two categories. DataTransfer-Out-Bytes leads at 45% of total data transfer spend, closely followed by DataTransfer-Regional-Bytes at 43%, while AWS-Out-Bytes accounts for 9% and DataXfer-Out represents a minimal 2%.
DataTransfer-Out-Bytes costs stem from data leaving AWS to the internet or external destinations, often indicating heavy API usage, content delivery, or data synchronization with external systems. DataTransfer-Regional-Bytes costs occur between different Availability Zones within the same region, which commonly occurs in multi-region architectures for disaster recovery, global content distribution, or compliance requirements.
AWS Data Transfer costs make up over 2% of total AWS costs.
Cognitive Services, which contains Azure OpenAI and other Azure AI services, is 13th and growing on Azure's services by spend.
Graviton-based instances account for over 6% of EC2 instance spend.
Dads v5 is the Azure VM with the highest amount of spend.
m6 has finally passed c6 in terms of percent spent. Q4 of 2023 was the time that c6 surpassed the formerly dominant m5, as organizations embraced the improved price-performance and enhanced networking capabilities of the sixth-generation compute-optimized instances. Since then, m6 has been gradually gaining share.
This migration pattern suggests organizations are prioritizing balanced compute and memory performance over pure compute optimization instances, as the m series offers a more balanced ratio of CPU, memory, and networking resources.
The relationship between total cloud spend and observability investment reveals an inverse correlation as organizations scale. Smaller environments spending $1k to $10k monthly allocate 17.91% of their budget to observability tools, while larger environments with $1M+ monthly spend dedicate only 11.95% to observability services.
One possible explanation is that larger organizations typically negotiate volume discounts on observability platforms, reducing their relative spending percentage as they achieve better pricing tiers unavailable to smaller customers.
Enter your email to view last two sections
This distribution reflects a mature infrastructure approach where production workloads receive the majority of resource allocation, while non-production environments are strategically right-sized for their specific purposes.
Teams can further optimize costs by implementing practices such as shutting down development or staging instances during evenings and weekends, using smaller instance types for test environments, adopting on-demand or spot instances for ephemeral workloads, and automating environment lifecycle management through infrastructure-as-code.