CLOUD AT WORK
There are numerous compelling examples of analytics initiatives made possible through cloud services, as shared by industry observers. “Marketing teams can leverage cloud services to conduct customer behavior analysis, while banks can run credit risk assessments and fraud detection,” said Pattni. “Cloud infrastructure enables scalability and on-demand access to services like cloud data warehouses, data lakes, and machine-learning platforms.”
One well-known and very massive application, Google Earth engine, “has been killing it for years,” said Flynn. “They host petabyte-scale remote sensing catalogs, so researchers, students, and noncommercial users can study the effects of our ever-changing planet. This is something that couldn’t easily be replicated outside of the cloud. Anything similar would require investing billions in research, hardware, software, and imaging data for decades. The fact that this exists is pretty damn cool.”
Within banking, cloud-based analytics is being employed to “process real-time customer interaction data, such as call center transcripts, chat logs, or online behavior and applied machine learning models, to recommend the next best action during a conversation,” said Gago. “By running in the cloud, such systems [were] able to scale to handle large volumes of data, integrate multiple sources without any data silos, and deliver real-time insights.”
“A hybrid approach also served one transportation services provider with visibility across both its on-premises and cloud systems. This hybrid environment consisted of Azure-based analytics, data management, and automated data pipelines,” Kline explained. “Observability tools integrated across cloud and on-premises systems enabled the company’s team to achieve end-to-end visibility into operations and cut root cause identification time by 70%.”
In another case, healthcare organizations “use the cloud to support population health studies or accelerate clinical research,” said Gnau. “By linking data from EHRs [electronic health records], claims, and external sources, they can analyze patterns at scale while still keeping patient data governed within their own environments.”
OVERCOMING CLOUD ISSUES
As with all transformative technologies, there are issues or complications data managers need to be aware of as they deploy on cloud, observers agreed.
Long-term costs—leading to cloud sticker shock—is one of the leading considerations in cloud deployments. “While the cloud’s pay-as-you-go model can be flexible, it can also lead to unpredictable costs if usage isn’t closely monitored and optimized,” Morais warned. “The long-term cost can become a challenge compared to a one-time investment with on-premises hardware.”
For many organizations, the costs of supporting AI in the public cloud “are becoming unsustainable,” warned David Linthicum, managing director and chief cloud strategy officer at Deloitte, in a recent webcast. “High GPU pricing, data transfer fees, and hidden storage charges are forcing teams to rethink their approach as projects grow.” In response, some organizations are “turning to hybrid or multi-cloud strategies [or outright] investing in on-premises infrastructure or exploring emerging options like AI-optimized colocation.”
The key is “understanding your workloads, usage patterns, and long-term needs before locking into a costly path,” Linthicum asserted.
Careful mapping of data dependencies and attention to testing and compliance requirements need to be front and center when migrating to the cloud. “I know of one company where an analyst wrote and deployed a new SQL query on a Friday afternoon, then went home for the weekend,” Kline recounted. “When they returned on Monday, the query was still running—to the tune of around $60,000! In the cloud, you pay for the resources you consume, which means poorly coded applications and SQL queries cost more money. While long-term cloud expenses are not inherently prohibitive, achieving a sustainable return on investment requires disciplined planning, optimized workflows, and strong governance to maintain operational efficiency over time.”
Although cloud costs lead the list of issues, there are additional considerations that must be weighed. “The cost of long-term cloud is a barrier, but not the only rub,” said Flynn. “Look for identity sprawl and policy drift, governance gaps, lineage gaps, and compounding complexity in multi-cloud landscapes.”
As a result of overcoming these barriers, “Decision makers should accept premium prices for more secure SKUs [stock keeping units] and ultimately be OK with a shared security model, where misconfigurations can become major incidents,” Flynn cautioned.
There is also a need for constant fine-tuning to overcome “migration complexity, interoperability, governance, and team expertise,” Kline emphasized. “Costs escalate quickly when queries are inefficient, testing is limited, or incremental models are overused—particularly in GPU or compute-intensive workloads.”
Security and governance need to go with any cloud implementation—and this is something that should not be outsourced to a cloud provider. “While cloud providers offer strong security, it is still up to the customer to properly configure and manage their data to avoid breaches,” Morais advised. “Misconfigurations when setting up your cloud services can leave your systems vulnerable to data breaches and put your data at risk.”
Beyond security and compliance concerns, cloud sovereignty is also a factor, said Flynn “Some governments and regulations inhibit sharing constituent data outside of political boundaries.”
However, for most, “The benefits of the cloud, such as scalability and fast deployment, can outweigh concerns for some businesses,” said Morais.
It’s also important to be aware that not all data workloads belong in the cloud. Cloud may mean “less-comprehensive security, unpredictable consumption-based costs when data processing accelerates, and the inability to meet the more aggressive service-level-agreement deadlines,” said Gago.
Ultimately, the choice between cloud and on-prem AI and analytics comes down to unique business needs, said Gago. “The key is ensuring you have access to 100% of your data, no matter where it lives.”