-
A new Ciena survey found operators are leaning toward using private cloud infrastructure for AI deployments
-
Korean operators are most likely to deploy on private cloud, compared with operators elsewhere in the world
-
Data security and privacy concerns are primary considerations in cloud decisions
While big-name artificial intelligence (AI) companies are partnering with public cloud hyperscalers for deployment, communication service providers (CSPs) are leaning more toward private cloud, according to a newly released survey.
According to a survey from Ciena and Censuswide, 43% of operators said they favor using private cloud infrastructure for deploying AI services on the network. Another 21% said they plan to use a hybrid cloud model. A bit over a third (37%) said they’re leaning toward public cloud offerings.
Ciena’s survey included responses from over 1,500 telecom and IT engineers and managers working for operators in 17 countries.
Ciena shared the raw results of the survey with Fierce. Interestingly, operators in Korea (70%), Mexico (58%), Norway (56%) and Indonesia (55%) leaned the most heavily toward private cloud infrastructure. The United Arab Emirates (52%), Germany (51%), Sweden (50%) and the U.K. (49%) were most in favor of using public cloud exclusively. Operators in India and Egypt were the most inclined to go the hybrid route.
The focus on private cloud stems from the risk-averse nature of most CSPs.
As Equinix noted in a recent blog, running AI on private cloud infrastructure allows telcos and other enterprises to keep their data on premises instead of pushing it to the public cloud to be fed into an AI model. This can yield a range of benefits, including increased data privacy and security, control over and visibility into model history, and even lower cost at scale.
The public cloud, however, has a lower barrier to entry and can be a better option for companies with a lack of in-house AI talent. It also offers access to “higher quality models,” Equinix said.
Ciena’s raw survey results backed up Equinix’s assertions, with just under 38% of respondents citing data privacy and security concerns as the biggest barrier to telco AI adoption.
"The choice between private and public clouds for AI workloads is a complex one for CSPs. Factors like data sensitivity, control, performance, availability, scalability, sustainability, governance, skill sets and cost all play a role. There is no ‘on size fits all; approach," Jürgen Hatheier, Ciena’s International Chief Technology Officer, told Fierce.
He added, "Many CSPs prefer private clouds for sensitive AI traffic due to concerns about data privacy, security, and governance due to regulated data isolation rules. Additionally, private clouds are fully customizable to provide optimal performance."
In terms of what they’re hoping to gain from AI deployments, the Ciena survey found 40% of CSPs think AI will generate revenue by opening up networks to third-party integrations. Another 37% each said they expect revenue from AI-enabled security and privacy services and new product offerings. And 35% said they think AI can generate revenue from the creation of tailored subscription packages.
However, CSPs think a few potential network-related hurdles could hinder AI growth. These include the need for core network upgrades of optical equipment and the like (37%) and insufficient data center interconnect infrastructure (30%).