- Many companies face challenges with cloud migration due to issues like security concerns, high costs and outdated infrastructure
- Incomplete cloud migrations are limiting AI effectiveness
- Experts recommend direct cloud connections, scalable workflows and private cloud solutions to enhance data movement and optimize AI performance
For many companies, cloud migration is a necessary step in the journey toward digital transformation and artificial intelligence (AI) adoption.
“We can expect a massive acceleration of data growth alongside the advancement of AI over the next few years,” said Ivo Ivanov, CEO of DE-CIX, one of the world’s largest internet exchange operators. “Developing a robust hybrid or multi-cloud scenario with effective connectivity will become a necessity.”
Yet, the move to cloud is proving complicated for many enterprises.
An MIT survey asked enterprises what aspects of their organization’s data are most limiting the speed to deploy AI. And about a third (34%) of respondents said incomplete cloud migration status is the culprit.
That is especially true for companies with revenues of over $1 billion, as “organizations presiding over larger data repositories and legacy IT infrastructure may be finding greater complexity and costs in transitioning to an AI-ready architecture,” the report said.
Ivanov cited major concerns including infrastructure, cost management and a lack of expertise in navigating complex cloud migration processes — with significant implications for businesses aiming to leverage AI across their operations.
Why is cloud migration stalling?
While cloud infrastructure offers the scalability and flexibility organizations need, there are hidden costs, such as cloud egress fees, which can inflate budgets unexpectedly, noted Ivanov. The shift often requires hiring cloud architects or reskilling IT teams, leading to additional challenges.
Vendor lock-in is another common pitfall. This can be mitigated with a robust multi-cloud strategy, but connectivity issues can still undermine the entire effort. Without a comprehensive cloud connectivity strategy, companies face performance bottlenecks that hinder the flow of data necessary for cloud and AI applications.
The lack of interoperability between different infrastructures in a hybrid or multi-cloud scenario, causes “an added layer of complexity,” Ivanov said. “One effective way to mitigate this risk is by using a software-defined cloud or AI routing service on an exchange platform.”
Without a well-thought-out connectivity strategy, the risk of losing control becomes "a real concern." According to a recent IDC Survey, 64% of organizations across Europe say network latency and performance is a relevant technical challenge for their cloud migration.
“Ensuring long-term success with AI initiatives requires a fully conceived infrastructure strategy,” he told Fierce Network. “This includes optimizing connectivity with a holistic approach across company locations, making sensible use of cloud resources, and avoiding unnecessary costs.”
Impact on AI initiatives
There is also the growing complexity of modern data environments to deal with. Many organizations rely on outdated, "brittle" pipelines, which are hard to maintain and slow down the cloud migration process, noted Jonathan Whitney, chief product officer at Airbyte. The fragmentation of data systems and the lack of standardized tools to move large volumes of structured and unstructured data add layers of difficulty.
“Legacy systems lack scalability, leading to bottlenecks and preventing seamless data movement,” Whitney told Fierce.
Data is the lifeblood of AI, and without seamless access to “diverse, clean, structured, and unstructured data sources,” these models lack the necessary inputs to function effectively, reducing accuracy and decision-making power, Whitney added.
Anything that undermines the flow of data to and from an AI model will impact the value enterprises can get out of the model. This is especially true for large language models (LLMs), which rely on massive datasets for training and re-training, all of which are often hosted in cloud environments.
When companies can't get their data to the cloud, AI models are left with insufficient inputs, leading to reduced accuracy and missed opportunities for AI-driven insights. Thus, without reliable cloud access, businesses cannot fully capitalize on AI's potential.
Navigating AI and cloud integration
To fully benefit from AI, Ivanov pointed to the importance of building direct cloud connectivity and leveraging cloud exchange platforms to reduce latency and costs.
“A company’s data transmissions, data exchange and network interconnections must be seamless, reliable, secure, and high-performance,” he said.
According to Ivanov, achieving this requires three key elements: direct cloud connections via private connectivity solutions at Cloud Exchanges, scalable cloud routing services and direct peering with AI-as-a-Service operators. These private solutions are more efficient and reliable than public Internet routes, he said, as they're covered by SLAs that guarantee performance.
Meanwhile, Whitney touted open-source platforms that support both structured and unstructured data across multiple clouds. He said companies should also implement scalable, automated workflows to ensure seamless data pipelines and consistent availability. Tools optimized for AI use cases, like vector databases and RAG models, enhance AI and LLM workflows, are also essential.