Datadog, Inc. (NASDAQ: DDOG), a cloud application monitoring and security platform, today released a new report, What will cloud costs be like in 2024?According to the report, organizations using Graphics Processing Unit (GPU) instances have seen an average spending increase of 40% on these instances over the past year. This increase in spending on GPU instances comes as more companies experiment with AI and large language models (LLMs). The parallel processing power of GPUs is essential for training LLMs and running other AI workloads, and can be over 200% faster than CPUs.
“Today, the most widely used type of GPU-based instance is also the cheapest, indicating that many customers are still in the experimental phase of AI, applying GPU instances to early efforts in adaptive AI, machine learning inference, and small-scale training,” said Yrieix Garnier, vice president of products at Datadog. “As organizations scale their AI activities and move them into production, we expect they will begin to spend a larger percentage of their cloud computing budgets using the more expensive types of GPU-based instances.”
In addition to companies spending more compute on AI projects, the report found that containers have become a common theme of wasted spend across organizations. In fact, 83% of container costs were related to idle resources. Approximately 54% of this wasted spend was related to cluster idle (the cost of over-provisioning cluster infrastructure) and 29% was related to workload idle (resulting from requests that are larger than the resources required for the workload). This wasted spend is occurring as organizations allocate more EC2 compute to running containers on EC2 compute (up to 35% from 30% a year ago).
Other key findings from the report include:
- Outdated technology is widely used: While AWS' current infrastructure offerings generally perform better and cost less than previous generation versions, 83% of organizations still spend an average of 17% of their EC2 budget on previous generation technology.
- Fewer organizations are taking advantage of discounts: Cloud service providers offer commitment-based discounts for many services (for example, AWS has discount programs for Amazon EC2, Amazon RDS, Amazon SageMaker, etc.), but only 67% of organizations participate in these discounts, down from 72% last year.
- Green technologies on the rise for improved performance and cost: On average, organizations using Arm-based instances are spending 18% of their EC2 compute budget on them, double what they did a year ago. Instance types based on Arm processors use up to 60% less energy than similar EC2 instances and often offer better performance at a lower cost.
In this report, Datadog analyzed AWS cloud cost data from hundreds of organizations to examine how their use of emerging and previous generation technologies, cloud resource usage patterns, and participation in AWS discount programs are affecting cloud costs.
Datadog's What will cloud costs be like in 2024? is live now. For full results, visit https://www.datadoghq.com/state-of-cloud-costs/. To learn more about how Datadog is helping businesses optimize their cloud costs, visit https://www.datadoghq.com/product/cloud-cost-management/.
For more latest updates, follow us on Google News. Martech News
