Spending on artificial intelligence applications, particularly generative AI, is driving up the cost of enterprise cloud computing. These costs climbed an average of 30%, according to a 2024 report commissioned by Tangoe in October, a technology expense management solution provider and conducted by Vanson Bourne.
In addition, 72% of IT and financial leaders believed that GenAI-led cloud spending had become unmanageable.
“GenAI is creating a cloud boom that will take IT expenditures to new heights,” Chris Ortbals, chief product officer at Tangoe, said in a statement. “With year-over-year cloud spending up 30%, we are seeing the financial fallout of AI demands. Left unmanaged, GenAI has the potential to make innovation financially unsustainable.”
Ortbals even described cloud costs as lethal to GenAI.
“The cloud’s hidden costs and unpredictable invoices can become the silent killer of GenAI,” added Ortbals. “The more urgently companies adopt comprehensive cost management FinOps strategies, the easier it is for them to turn GenAI’s promise into lasting innovation instead of runaway expenses and technical debt.”
Cloud costs are rising amid inflation and technical debt, Ortbals wrote in Forbes. He noted that it is the role of CIOs to pay for shared services as the “habitual corporate financier” even when costs increase. As these cloud costs climb, tensions rise between IT and finance, Ortbals wrote.
How AI Is Impacting the Cloud Landscape
Cloud spending is indeed increasing because of the demands of AI, explains Matt Hobbs, cloud, engineering, data and AI leader at PwC.
“If you look at the resource intensity of the very specific workloads you’re using it for, in combination with the fact that those resources are super constrained, it is keeping those costs really high right now,” Hobbs tells InformationWeek.
AI workloads are costly because organizations are hungry for capacity and they are using cloud resources to unify their data environment, he says.
“Speed matters a lot here, and so if you’re in the cloud, you have the ability to go a lot faster than if you’re running on prem,” Hobbs says.
In addition, as organizations move from on-prem infrastructure and shut down data centers to move to the cloud, even with AI driving up costs, companies’ cloud costs were increasing anyway, Hobbs suggests.
In addition, Hobbs notes the “duplicative costs” that occur as AI companies offer their own direct LLM services and cloud providers integrate them as well.
“If you look at AI as a driver toward cloud costs, that’s a question of, is it actually more expensive, or is it a shift toward cloud that’s happening because of AI?” Hobbs says.
As the life cycle of infrastructure gets shorter and GPUs get more powerful, cloud costs go up, explains Dmitry Panenkov, CEO and founder at cloud-management platform Emma.
“So basically, the life cycle is getting shorter, and each and every accelerator they release is more powerful, but on the other hand, is also more expensive, and this automatically drives up your costs,” Panenkov explains. “So, you need to pay more if you want to get these GPUs, and the providers need to pay more. And then if you train the models on top-notch accelerators, you pay more per hour to ramp up this capacity.”
Although cloud costs are increasing due to AI, organizations are not slowing down in spending on cloud or AI, according to Hobbs.
Nic Benders, chief technical strategist for New Relic, agrees that spending will be robust for infrastructure such as cloud amid AI’s growth.
“I believe IT spend is actually constrained by the amount of money in IT, not by the things to spend it on,” Benders says. “So, I believe that we will continue to see rapid growth in spending on infrastructure.”
How AI Tools Help Forecast Cloud Spending
Although AI may make cloud costs climb, AI tools can also help manage these costs and alleviate cloud spending. Organizations can use predictive analytics to study past usage patterns. In addition, machine learning can train models on past usage patterns and auto-scale use of cloud resources.
Emma uses AI to analyze the behavior of cloud workloads and allow organizations to adjust their environments to reduce their cloud bills, Panenkov says. He predicts that AI costs and thus cloud costs will go down as the price of GPU accelerators drop.
“We have a networking backbone that interconnects the clouds, and we have AI algorithms to define the best and most optimal route from one service provider to another, which is associated with a smaller cost,” Panenkov says.
Benders also sees the move to expensive infrastructure such as GPU accelerators as short-term.
Just as the tech industry moved from three nodes in a cluster to thousands of nodes in a cluster and hardware got less expensive, Benders sees a similar pattern with AI.
“I suspect that we’re going to see the same thing in the AI-driven load that, if it matures, will move away from those kinds of cutting-edge experimental systems, but that’s not going to be for some years now. So, I think we’re in a phase right now where people are going to be spending their money on those cutting-edge systems,” he says, referring to GPU accelerators.
How CIOs Should Approach AI and Cloud Spending Going forward
Panenkov recommends a hybrid model of on premises and cloud to manage cloud costs.
“The best model to work with is a hybrid model, where you have your on-premises environment where you can train your models,” Panenkov says. “But in case you need to scale and pick up more GPU instances to continue your training of your model, you can scale the workloads up into the cloud, and for short period of time, you can rent certain instances with the cloud service provider, so that we think is the right approach.”
Hobbs advises that organizations assess what they are using AI services for when choosing their infrastructure. By deploying workloads — whether cloud or AI — at the edge as part of a hybrid cloud setup, organizations can drive down overall cloud costs.
“When enterprise data is connected, companies naturally leverage the centralized cloud,” Hobbs explains. “However, when data becomes disconnected at the edge, placing computing power locally can significantly lower costs.”
For example, Hobbs notes that a telco company might serve its customers through both private and public clouds. In this arrangement, the private cloud delivers direct value to end users, while the public cloud offers operational efficiencies for enterprises.
“I think it matters more where an organization is on its cloud journey — that’s what truly drives the architectural decision — than merely following a fixed pattern of delivering an end service to a customer,” Hobbs says. Spending on artificial intelligence applications, particularly generative AI, is driving up the cost of enterprise cloud computing. These costs climbed an average of 30%, according to a 2024 report commissioned by Tangoe in October, a technology expense management solution provider and conducted by Vanson Bourne.
In addition, 72% of IT and financial leaders believed that GenAI-led cloud spending had become unmanageable.
“GenAI is creating a cloud boom that will take IT expenditures to new heights,” Chris Ortbals, chief product officer at Tangoe, said in a statement. “With year-over-year cloud spending up 30%, we are seeing the financial fallout of AI demands. Left unmanaged, GenAI has the potential to make innovation financially unsustainable.”
Ortbals even described cloud costs as lethal to GenAI.
“The cloud’s hidden costs and unpredictable invoices can become the silent killer of GenAI,” added Ortbals. “The more urgently companies adopt comprehensive cost management FinOps strategies, the easier it is for them to turn GenAI’s promise into lasting innovation instead of runaway expenses and technical debt.”
Cloud costs are rising amid inflation and technical debt, Ortbals wrote in Forbes. He noted that it is the role of CIOs to pay for shared services as the “habitual corporate financier” even when costs increase. As these cloud costs climb, tensions rise between IT and finance, Ortbals wrote.
How AI Is Impacting the Cloud Landscape
Cloud spending is indeed increasing because of the demands of AI, explains Matt Hobbs, cloud, engineering, data and AI leader at PwC.
“If you look at the resource intensity of the very specific workloads you’re using it for, in combination with the fact that those resources are super constrained, it is keeping those costs really high right now,” Hobbs tells InformationWeek.
AI workloads are costly because organizations are hungry for capacity and they are using cloud resources to unify their data environment, he says.
“Speed matters a lot here, and so if you’re in the cloud, you have the ability to go a lot faster than if you’re running on prem,” Hobbs says.
In addition, as organizations move from on-prem infrastructure and shut down data centers to move to the cloud, even with AI driving up costs, companies’ cloud costs were increasing anyway, Hobbs suggests.
In addition, Hobbs notes the “duplicative costs” that occur as AI companies offer their own direct LLM services and cloud providers integrate them as well.
“If you look at AI as a driver toward cloud costs, that’s a question of, is it actually more expensive, or is it a shift toward cloud that’s happening because of AI?” Hobbs says.
As the life cycle of infrastructure gets shorter and GPUs get more powerful, cloud costs go up, explains Dmitry Panenkov, CEO and founder at cloud-management platform Emma.
“So basically, the life cycle is getting shorter, and each and every accelerator they release is more powerful, but on the other hand, is also more expensive, and this automatically drives up your costs,” Panenkov explains. “So, you need to pay more if you want to get these GPUs, and the providers need to pay more. And then if you train the models on top-notch accelerators, you pay more per hour to ramp up this capacity.”
Although cloud costs are increasing due to AI, organizations are not slowing down in spending on cloud or AI, according to Hobbs.
Nic Benders, chief technical strategist for New Relic, agrees that spending will be robust for infrastructure such as cloud amid AI’s growth.
“I believe IT spend is actually constrained by the amount of money in IT, not by the things to spend it on,” Benders says. “So, I believe that we will continue to see rapid growth in spending on infrastructure.”
How AI Tools Help Forecast Cloud Spending
Although AI may make cloud costs climb, AI tools can also help manage these costs and alleviate cloud spending. Organizations can use predictive analytics to study past usage patterns. In addition, machine learning can train models on past usage patterns and auto-scale use of cloud resources.
Emma uses AI to analyze the behavior of cloud workloads and allow organizations to adjust their environments to reduce their cloud bills, Panenkov says. He predicts that AI costs and thus cloud costs will go down as the price of GPU accelerators drop.
“We have a networking backbone that interconnects the clouds, and we have AI algorithms to define the best and most optimal route from one service provider to another, which is associated with a smaller cost,” Panenkov says.
Benders also sees the move to expensive infrastructure such as GPU accelerators as short-term.
Just as the tech industry moved from three nodes in a cluster to thousands of nodes in a cluster and hardware got less expensive, Benders sees a similar pattern with AI.
“I suspect that we’re going to see the same thing in the AI-driven load that, if it matures, will move away from those kinds of cutting-edge experimental systems, but that’s not going to be for some years now. So, I think we’re in a phase right now where people are going to be spending their money on those cutting-edge systems,” he says, referring to GPU accelerators.
How CIOs Should Approach AI and Cloud Spending Going forward
Panenkov recommends a hybrid model of on premises and cloud to manage cloud costs.
“The best model to work with is a hybrid model, where you have your on-premises environment where you can train your models,” Panenkov says. “But in case you need to scale and pick up more GPU instances to continue your training of your model, you can scale the workloads up into the cloud, and for short period of time, you can rent certain instances with the cloud service provider, so that we think is the right approach.”
Hobbs advises that organizations assess what they are using AI services for when choosing their infrastructure. By deploying workloads — whether cloud or AI — at the edge as part of a hybrid cloud setup, organizations can drive down overall cloud costs.
“When enterprise data is connected, companies naturally leverage the centralized cloud,” Hobbs explains. “However, when data becomes disconnected at the edge, placing computing power locally can significantly lower costs.”
For example, Hobbs notes that a telco company might serve its customers through both private and public clouds. In this arrangement, the private cloud delivers direct value to end users, while the public cloud offers operational efficiencies for enterprises.
“I think it matters more where an organization is on its cloud journey — that’s what truly drives the architectural decision — than merely following a fixed pattern of delivering an end service to a customer,” Hobbs says.
