Cloud usage continues to soar, as are its associated costs (especially recently AI-powered costs). Gartner analysts predict that global end-user spending on public cloud services will expand to $723.4 billion in 2025, up from $600 billion in 2024. 70% of executives surveyed in the IBM report listed Generative AI as a key driver of this growth.
Meanwhile, China’s DeepSeek claims it took only two months to train its AI model, which caused a sensation. Whether these characters are wondering whether these characters tell the whole story, but Microsoft and Nvidia still sprinkle whether there is any sign of stock prices, the announcement awakens the Western world until cost-effective AI systems are needed.
So far, companies have been able to view the increasing costs of AI as R&D. However, AI costs, especially those associated with successful products and features, will eventually hit the cost of goods sold by the company (COGS), thus losing its gross margin. AI innovation is always destined to face indifferent scrutiny of business awareness. DeepSeek’s bombshell announced a shorter timeline.
Just like with other public clouds, companies will need to manage their AI costs, including training and consumption costs. They need to connect AI spending with business outcomes, optimize AI infrastructure costs, optimize pricing and packaging strategies, and maximize returns on their AI investments.
How do they do it? with Cloud Unit Economics (CUE).
What is cloud unit economics (tips)?
CUE includes measurement and maximization of cloud-driven profits. Its basic mechanism is to link cloud cost data with customer demand and revenue data, revealing the most profitable dimensions of the enterprise, thus showing companies how to optimize the company. CUE is suitable for all sources of cloud spending, including AI costs.
The basis of the prompt is Cost allocation – Organize cloud costs based on who and/or what drives them. Common allocation dimensions include cost per customer, cost per engineering team, cost per product, cost per functional, and cost per microservice. Companies using modern cost management platforms often allocate costs in a framework that reflects their business structure (its engineering hierarchy, platform infrastructure, etc.).
Then, the prompting heart is Unit cost measurementcompare cost data with demand data to show the company’s full cost service. For example, a B2B marketing company may want to calculate “every 1,000 messages” sent through its platform. To do this, it must track its cloud cost and the number of messages sent, feed the data into a system, and instruct the system to divide its cloud cost by its message and draw the results in the dashboard.
Since the company starts with cost allocation, it can view its cost per 1,000 messages through customers, products, features, teams, microservices, or any other view it is considered to reflect its business structure.
result:
- Flexible Business dimension They can filter unit cost metrics to show them what areas of their business are driving their cloud cost
- A lighting Unit cost measurement This shows them how effective they can meet customer needs
- Ability to make targeted efficiency improvements, such as refactoring infrastructure, adjusting customer contracts, or improving pricing and packaging models
Tips from the AI era
In the prompt model, AI costs are just one source of cloud spending that can be incorporated into the enterprise’s allocation framework. The way AI companies disseminate cost data is still evolving, but in principle, cost management platforms handle AI costs in the same way as they treat AWS, Azure, GCP, and SaaS costs.
Modern cloud cost management platforms allocate AI costs and show their efficiency impact in the context of unit cost indicators.
Companies should allocate their AI costs in a few intuitive ways. One will be the cost of each team mentioned above, which is the allocation dimension shared by all cloud expenditure sources, showing the cost of each engineering team being responsible. This is especially useful because leaders know exactly who to notify and assume responsibility when the expenses of a particular team soar.
The company may also want to know theirs Cost per AI service type – Machine learning (ML) models and basic models and third-party models such as OpenAI. Alternatively, they can calculate the cost of each SDLC stage to understand how the cost of AI functionality changes as it goes from development to testing to staging, to production. A company may gain more fine-grainedness and calculate the cost of each of its AI development lifecycle stages, including data cleaning, storage, model creation, model training, and inference.
Shrink a little from the weed: Tips means comparing organized cloud cost data with customer demand data and then figuring out where to optimize. AI costs are just another source of cloud cost data, and with the right platform, it seamlessly fits the company’s overall prompt strategy.
Avoid gear tsunami
As of 2024, only 61% of companies have developed formal cloud cost management systems (according to CloudZero survey). Unmanaged cloud costs quickly become difficult to manage: 31% of companies (similar to the part that does not formally manage their costs) suffer major gear hits, reporting cloud costs consume 11% or more of revenue. Unmanaged AI costs will only exacerbate this trend.
Today’s most thoughtful organizations use cloud costs like any other major expenditure, calculating their ROI, allowing ROI to undermine ROI for its most critical business aspects, and giving relevant team members the data they need to optimize the ROI. The next generation cloud cost management platform provides a comprehensive tips workflow that helps companies avoid gear tsunamis and enhance long-term feasibility.
