By 2029, half of cloud computing globally will be driven by AI workloads, up from less than one-tenth today. The surge in AI use will drive a shift in how organisations allocate compute resources to embed AI foundational capabilities into their IT infrastructures.
The speed of AI and ML growth will throw up the question of whether data centres and corporate cloud strategies will be ready to handle the surge.
The lack of interoperability between multi-clouds is a concern that can potentially slow cloud adoption. It is expected that more than half of organisations will not get the results they want from their multi-cloud implementations when we reach 2029.
Along the way, not all cloud implementations will succeed. A quarter of organisations are likely to find they are not getting the results they want from their cloud adoption by 2028.
These were the prognoses and predictions from Gartner’s recent report on the Top Trends Shaping The Cloud, released earlier this month.
IT infrastructures must be modernised to be ready for AI
When it comes to AI, Gartner believes cloud providers must address the challenge of growing power consumption, while organisations must manage their approach to AI workloads in the face of financial scrutiny and implementation difficulties.
Speaking with iTNews Asia on how ready organisations in APAC are in handling the AI/ML surge, Adrian Wong, Director Analyst at Gartner, said organisations in Asia Pacific face significant hurdles in realising their AI/ML ambitions due to the increasing complexity of infrastructure requirements and a shortage of skilled personnel.
“Building an effective AI infrastructure extends beyond merely deploying AI accelerators or GPUs, which have well-documented supply chain constraints. The surrounding infrastructure, including storage, networking, power delivery and cooling systems, must be optimised to prevent AI accelerators from experiencing data starvation, which can dilute their return on investment. Many of these components require modernisation to support AI workloads effectively,” said Wong, who is also Senior Principal Analyst for the Cloud Infrastructure and Operations team within the Gartner for Technical Professionals (GTP) research organisation.
He added that even with infrastructure solutions in place, the scarcity of skilled data scientists and AI specialists in the region remains a critical barrier. This skills gap will impede organisations from developing, deploying and managing AI/ML solutions at scale effectively.
To address these challenges, Wong encourages organisations throughout the region to turn to hyperscale cloud providers for their AI/ML workloads. Across the different countries, China remains an exception due to stringent regulatory requirements and data privacy concerns.
Across multi-clouds, Gartner advises that the cloud’s value can be increased and multi-cloud complexity reduced by identifying distributed applications, and adopting federated GenAI models and governance functions that benefit from a cross-cloud approach.
Making AI and sustainability work together
Wong said deploying AI/ML workloads in the cloud does provide opportunities to optimise power consumption and meet sustainability targets, adding that many cloud providers are already committed to powering their operations with renewable energy.
“When direct access to renewable energy is unavailable, they use renewable energy certificates (RECs) to achieve this goal. Additionally, most cloud providers publish Power Usage Effectiveness (PUE) metrics for their cloud regions, enabling organisations to identify data centres with efficient energy use. Cloud data centres typically have PUE ratings between 1.1 and 1.5,” he said.
“While the adoption of AI-specialised hardware and consumption optimisation is part of the sustainability narrative, the primary motivation for many APAC organisations remains cost reduction rather than achieving sustainability targets.”
Why will we see dissatisfaction with cloud adoption?
An interesting aspect seen from Gartner’s cloud adoption report was a prediction that a quarter of organisations will experience significant dissatisfaction with their cloud adoption by 2028. Wong said a key factor usually contributing to dissatisfaction is inadequate planning and the absence of a comprehensive cloud strategy, which invariably leads to high cloud costs.
“In the aftermath of COVID-19, many organisations rapidly migrated applications to the cloud without conducting adequate due diligence. This frequently led to applications being moved to the cloud despite not being well-suited for it, resulting in minimal benefits and higher costs compared to on-premises deployments. Legacy applications lacking a modernisation plan are a common example of this misalignment,” he said.
Wong added that significant challenges will come about when an organisation’s cloud strategy is poorly conceived or non-existent. “We are observing that ‘cloud repatriation’ is not as widespread as some on-premises vendors are claiming.”
(There are) case studies of organisations that have repatriated, which reveal a reliance on ‘lift and shift’ or ‘rehost’ strategies. These approaches typically fail to fully exploit cloud services, creating a gap between expectations and reality. The recurring theme in these cases is that costs become exceedingly high, yet the anticipated benefits fall short.
- Adrian Wong, Director Analyst, Gartner
Successful cloud adoption will require trade-offs
While the cloud offers numerous opportunities, Wong explained that not all benefits can be realised simultaneously. He explained that successful cloud adoption often involves trade-offs, and a well-defined cloud strategy is essential for establishing organisational priorities and clearly articulating the rationale behind these decisions.
“Organisations must prioritise their cloud adoption goals, whether driven by business agility, lower long-term total cost of ownership, or short-term cost avoidance. These trade-offs require careful consideration and alignment with business objectives.
“When the strategic approach is not effectively communicated, it can lead to organisational tension and unmet expectations,” he said.
Although a lack of expertise in cloud implementation can contribute to dissatisfaction, Wong recommends organisations address this talent gap by partnering with third-party managed service providers (MSPs) or using the professional services offered by cloud providers. These partnerships can provide access to specialised knowledge and support, enabling organisations to navigate the complexities of cloud adoption more effectively.
How critical will it be to align cloud strategies to address digital sovereignty concerns?
Looking at the global political and economic landscape, Wong said rising geopolitical tensions are today causing uncertainty in many countries, particular throughout Europe and across China's sphere of influence within APAC.
This uncertainty, he added, has heightened concerns over the ‘lock-in’ associated with public cloud services and the significant dependence organisations have on these providers. This is also driving the need for companies to align their cloud strategies to address digital sovereignty concerns.
“Events such as the 2024 CrowdStrike outage have made us painfully aware of the reliance on cloud providers (and IT systems in general) and the potential impact on organisations when they haven’t planned accordingly,” Wong said.
“Organisations are now evaluating ‘geopatriation’ – the relocation of workloads from one hosting environment (typically a hyperscale cloud provider that is perceived to carry increasing geopolitical risk) to an alternate (commonly local) hosting environment that offers greater sovereignty.”
He warned that one of the key concerns and significant impact will be the cessation of trade between polities, where economic sanctions are put in place and leading vendors to decide or be forced to stop doing business with customers in certain polities.
