As businesses across industries accelerate their adoption of artificial intelligence (AI), more and more are now shifting toward on-premise AI deployments, particularly industries which have sensitive data and specific operational requirements, compared to the past, wherepublic cloud adoption was traditionally the preferred choice due toits scalability and cost-effectiveness,
In an exclusive interview with iTNews Asia, Anand Chakravarthy, Vice President of Advanced Solutions at Tech Data Asia Pacific and Japan (APJ), gives his insights on the evolving trends in AI deployments, particularly the balance between public cloud, on-premise infrastructure, and hybrid models.
Why the public cloud for AI solutions?
The ongoing growth of public cloud adoption is reshaping the way businesses approach AI deployments, said Anand
He explained that the primary appeal of public cloud infrastructure lies in its scalability and flexibility. For example, businesses can scale their AI workloads according to demand, avoiding the need for significant upfront capital expenditure (CapEx) required for on-premise infrastructure.
"The flexibility that public cloud provides - whether it's storage, server capacity, or GPUs - allows businesses to grow as their needs evolve," Anand said.
"This scalability, combined with the avoidance of CapEx, makes the public cloud an attractive option, particularly for companies that need agility in their AI operations."
On-premise AI offers security, latency, and customisation benefits
According to Anand, despite the growing use of the cloud, on-premise AI deployments continue to offer unique advantages, particularly when data security and real-time execution are critical. For industries handling highly sensitive data, on-premise infrastructure provides greater control over data security, helping organisations comply with local regulations.
"The decision to adopt on-premise infrastructure often stems from the need to handle sensitive data and the desire to avoid potential latency issues in real-time applications," said Anand.
Industries such as banking, healthcare, and insurance are seeing a higher uptake of on-premise solutions for AI, particularly when the nature of their operations demands heightened security and speed.
Anand also emphasised the importance of customisation, noting that certain AI solutions are complex and require tailored environments that are more easily managed on-premise. For enterprises looking to quickly move their AI solutions into production while maintaining control over their data, on-premise deployments provide a distinct advantage over cloud-based infrastructure.
The commercial viability of on-premise infrastructure is also a key factor.
"If you're considering a long-term investment, the ongoing costs of using public cloud infrastructure can be significant," said Anand.
Over a five to six-year period, cloud services can become more expensive, particularly when factoring in continuous operational expenses. In contrast, on-premise solutions offer enterprises greater cost control over time.
Anand added that regulatory compliance also plays a major role in decision-making. Many industries often have strict regulations regarding where sensitive data must be stored. In these cases, on-premise deployments ensure that data remains within local jurisdictions, mitigating the risks associated with cloud storage in regions with different data protection laws.
The case for a hybrid infrastructure model
Looking ahead, hybrid infrastructure solutions appear to be the best of both worlds, offering businesses the flexibility to leverage both cloud and on-premise solutions according to their specific needs. Hybrid solutions enable enterprises to take advantage of the scalability of public cloud while retaining control over critical workloads in their own data centres.
Hybrid infrastructure is here to stay, particularly for large enterprises that need to balance cloud flexibility with the security and control provided by on-premise solutions. Smaller businesses may prefer the public cloud for its lower CapEx, but hybrid is ideal for organisations that have diverse needs.
- Anand Chakravarthy, Vice President of Advanced Solutions at Tech Data, Asia Pacific and Japan
Different industries in the APAC region face unique challenges when adopting AI infrastructure. Anand discussed how regulations, such as data sovereignty laws, play a significant role in shaping the decisions of businesses in various sectors.
In certain markets, local data protection regulations may restrict the use of public cloud for sensitive data, pushing companies in regulated industries to deploy AI solutions on-premise.
Conversely, countries with more flexible data privacy rules may see faster adoption of public cloud solutions, particularly for non-sensitive workloads such as customer service AI or chatbots.
Anand noted that hybrid models could offer a solution for organisations navigating diverse regulatory landscapes. "Companies operating across multiple countries with different data regulations will likely adopt hybrid infrastructures that enable them to comply with local laws while still benefiting from the cloud's scalability and cost-effectiveness," he explained.
However, he also pointed out that integrating and synchronising data between on-premise and cloud environments poses challenges. Ensuring seamless data flow across both platforms is essential for optimising performance, but these complexities require thoughtful planning and robust solutions.
As organisations look to scale their AI capabilities, it is clear that a one-size-fits-all approach will not suffice. Looking ahead, the balance between on-premise, cloud, and hybrid infrastructure will continue to evolve. While many businesses may initially experiment with on-premise solutions, Anand foresees a gradual shift toward the cloud in the coming years.
"In the next three to five years, a good percentage of AI deployments will move to the cloud," he said, noting that as organisations grow more confident in the security and scalability of public cloud environments, more businesses will make the transition.
However, the need for on-premise infrastructure will not disappear entirely. For large enterprises dealing with complex AI deployments, hybrid solutions will remain the preferred option.
The challenge, Anand points out, will be to maintain data synchronisation and ensure performance standards across hybrid environments –a task that can be facilitated by partners, who offer expertise in integrating cloud and on-premise systems.
