Cloud compute has become the primary cost of goods sold for AI-native companies. Most of them have no way to manage it.
Enterprise software used to be cheap to serve. Once the code was written, the next customer cost almost nothing. Gross margins of eighty per cent were ordinary.
Artificial intelligence has ended that arrangement. Every model query costs money. Every agent interaction. Every inference call. The cost scales with use, which means it scales with success.
This shift, quiet but structural, is rewriting the economics of an entire industry. The companies absorbing the change are doing so with financial systems built for a world that is ending. The consequences are now visible in the data.
The scale of the problem
A 2025 study by SaaS benchmarking firm Benchmarkit, conducted with cost governance platform Mavvrik, surveyed 372 enterprise organisations. Eighty per cent miss their AI infrastructure forecasts by more than twenty-five per cent. Eighty-four per cent report measurable gross margin erosion tied to AI workloads.
IBM’s Institute for Business Value found that average computing costs rose 89 per cent between 2023 and 2025. Every executive in the IBM study had cancelled or postponed at least one AI initiative because of cost concerns.
Pricing research firm Monetizely places typical AI-native gross margins at fifty to sixty per cent. Traditional SaaS sits at eighty to ninety. The thirty-point gap represents compute embedded in cost of goods sold.
Gartner projects US public cloud spending at $723 billion in 2025. Goldman Sachs estimates generative AI will account for ten to fifteen per cent of global cloud spend by 2030. The scale of capital being deployed is enormous. The scale of capital being managed without adequate forecasting tools is the same.
Why the existing tools fall short
Cloud cost management has matured as a category. Platforms like Vantage, CloudZero, and Apptio show enterprises where their money goes. They flag waste. They surface anomalies. For the question of what was spent last month, they are excellent.
They stop at the billing layer. They have no view of pricing models. They do not know which customer pays which contract. They cannot tell a CFO whether a specific enterprise tier is profitable at its current contract value, or what happens to margin if a new AI feature ships.
The leap from cost visibility to pricing decision is something every AI company has to make. Most are still making it with spreadsheets and quarterly finance cycles. The data exists in different systems. Nobody has connected it at scale.
The work being done to close the gap
Abha Dalmia, an enterprise AI pricing strategist at a major US technology company, has been studying this gap from the inside for nearly a decade. Her work spans pricing architecture for enterprise AI products, strategy consulting at McKinsey across multiple industries, and earlier quantitative modelling for energy trading desks in Singapore.
She has now founded Cloudgin, an AI-native platform that connects cloud infrastructure cost directly to product pricing and margin intelligence. The platform is being built to solve the join that existing tools do not perform: live cost data tied to live revenue data, updated continuously rather than reconciled in quarterly reports.
“The data exists in different systems,” Dalmia says. “Cost data sits with infrastructure. Pricing sits with sales. Margin sits with finance. Nobody has connected the three in a way that gives a CFO a real-time answer to whether the business is actually working.”
Cloudgin operates across three principal functions. Live margin visibility shows what it costs to serve a specific customer or run a specific feature at actual usage levels. Probabilistic cost forecasting models how spending will evolve as usage grows. Pricing simulation lets teams model the financial consequences of pricing changes before committing to them.
The intended user is the Engineering. The CFO. The head of product. The pricing strategist. The decision-makers whose judgement determines whether an AI business survives at scale.
Why this matters at national scale
The financial sustainability of US AI companies is not only a startup concern. It is, increasingly, a policy one.
The National AI Initiative Act of 2020 and Executive Order 14110, signed in October 2023, both identify AI commercialization as a national strategic priority. The implicit assumption in both is that American AI companies will be commercially viable. That assumption requires financial infrastructure that does not yet exist at scale.
Foreign competitors, particularly state-backed AI development efforts, do not face the same commercial viability requirements. They can sustain losses indefinitely in ways that market-dependent companies cannot. For US AI companies to remain competitive, they must generate real margins on real products. Doing so requires understanding what those products cost to operate.
The public sector faces the same gap. A September 2024 audit by the United States Government Accountability Office found that most of twenty-four major US federal agencies were not meeting cloud cost governance requirements set by the Office of Management and Budget. The federal government spent $12.3 billion on cloud goods and services in fiscal year 2022. The GAO issued forty-seven recommendations addressing forecasting, attribution, and real-time adjustment.
These are the same capabilities Cloudgin is being built to provide.
Where this goes next
AI agent deployment is moving from pilot to production across industries. Investor scrutiny on AI unit economics has intensified. The compute cost curve continues climbing in aggregate, even as individual model workloads become more efficient.
Each shift increases the urgency of tools that connect cost to pricing in real time. The category of cloud cost intelligence, distinct from cloud cost visibility, is beginning to form. The companies and platforms that establish leadership in this category during the next eighteen months will likely define how AI businesses manage their economics for the next decade.
The bill keeps arriving every month. The question of what it actually means is one that more enterprises will be forced to answer.
Abha Dalmia, an enterprise AI pricing strategist at a major US technology company and the founder of Cloudgin, an AI-native cloud cost intelligence platform, has worked on this problem across multiple industries. She holds a STEM MBA from Carnegie Mellon University and has previously worked at McKinsey and Company and Morgan Stanley.
