AI is crucial for driving adoption at the network edge, finds STL Research in latest survey
3 min read- Network edge locations are increasingly seen by developers as the optimum location for AI inferencing
- Two of the top three applications hosted at the network edge today leverage AI
- 24% of respondents anticipate GPU to be prevalent in edge adoption by 2029
AI is seen as a key driver for adoption at the edge – and inferencing is better suited to edge environments in comparison with model training, according to the latest survey conducted by STL Partners.
LONDON – 11 November 2024 – Based on its annual survey of more than 100 experts in the edge computing sector, conducted in October 2024, the research and consulting firm has uncovered the current status of the network edge, and challenges and opportunities in this area.
The network edge is increasingly seen by developers as the optimum location for AI inferencing because it has a latency, sovereignty and cost advantage over centralised cloud infrastructure.
A total of 61% of respondents cite video analytics/computer vision as the top application type deployed at the network edge. Specific examples of such solutions include customer behaviour analysis, connected car driver assistance, and machine monitoring and control.
‘Computer vision use cases often require less bespoke integration work than other use cases, reducing the time to implementation and helping to create a more attractive business case. The data volumes that high-quality video creates plus the need for real-time alerts or insights lead to a clear reason for deploying at the edge’, comments Tilly Gilbert, consulting director and Edge Practice lead at STL Partners.
‘There is a real and growing appetite for infrastructure that supports AI solutions in a way that is performant, cost-effective and for some regions – sovereign. Network edge is increasingly seen as a choice to best satisfy these needs. Recent examples, including Deutsche Telekom’s deployment to support an autonomous vehicle solution for Volkswagen at the Port of Emden, Germany, are indicative of this trend’, adds Anna Boyle, senior consultant at STL Partners.
Another finding from the survey suggests that executives are bullish about the potential of graphics processing units (GPUs) being deployed at the edge.
‘Demand for AI and generative AI (GenAI) applications at the network edge have mobilised telecom operators to deploy GPUs and other specialised silicon at these sites. More than 40% of respondents said that non-CPU hardware will make up between 40% and 100% of edge hardware by 2029’, comments Boyle.
Zooming into specific verticals, STL Partners’ survey has discovered that the industrial sector remains the top adopter of network edge computing in 2024, closely followed by the media and entertainment vertical (see figure below).
Notably, this year the public sector and financial services have emerged as new adopters of network edge
You can find out more takeaways from the annual telco edge computing survey by downloading an extract here.
For media enquiries, click here to get in touch with our team.
