Akamai Targets a Niche in the AI Infrastructure Market With Distributed Networks
TECHWORLD ·
✦ AI Summary
The AI infrastructure market is shifting from training to inference in actual services.
Akamai is targeting demand for edge inference and AI inference by emphasizing cost competitiveness and strengths in distributed cloud.
Akamai proposed a structure that connects the edge and GPU data centers through a network, and said data transmission costs are about 5% of those at hyperscalers.
As AI use expands, the AI infrastructure market is growing rapidly. According to an industry assessment on the 23rd, the center of gravity in the AI infrastructure market is shifting from training to inference, which is used in actual services. As a result, competition in AI infrastructure, previously centered on hyperscalers, is expected to broaden.
At present, the main competitive landscape is being led by global hyperscalers. In this market, Akamai is positioning its cost competitiveness and strengths in distributed cloud as key advantages. Based on this, Akamai is targeting a niche in the AI infrastructure market.
The domestic market is characterized by strong enterprise demand for AI. In addition, demand is also growing in the domestic market for greater data sovereignty and for sovereign AI. Accordingly, Akamai is presenting performance and cost optimization, along with niche-market targeting, as its strategy for the domestic market.
Against this backdrop, edge inference is emerging as a promising area. The growth in edge inference is notable. Fortune Business Insights projected that edge inference will account for 70.76% of the AI inference market in 2026.
Akamai is expanding its business scope from AI training to AI inference. Its target market consists of companies facing burdens from rising compute resources and data transmission costs due to surging inference workloads. Akamai says it aims to capture this demand by emphasizing performance and cost efficiency.
The market is currently led by large hyperscalers based on massive centralized infrastructure. By contrast, Akamai is proposing a distributed structure that connects its existing edge infrastructure and GPU data centers through its own global network.
For AI inference, Akamai says its focus is not only on GPU computing performance but also on reducing data transmission costs and network latency. This structure is designed to target computing power as well as lower transmission costs and latency in processing user requests.
Akamai said it receives requests at user-adjacent edge locations and then forwards the received requests through its own network to GPU-built data centers. It added that the computation results are then delivered back to users.
Akamai said it has chosen a structure in which GPUs are not deployed directly at the edge, but the computing tasks of GPUs, which require high-power and cooling facilities, are handled at separate GPU data centers. Instead, it has focused on minimizing network latency between the edge and GPU data centers.
Akamai emphasized cost as a market competitiveness factor for this structure. It said data transmission costs are about 5% of those at hyperscalers, and that network costs are waived for some customers depending on the actual contract size.
It also said some customers have reduced their total cloud costs to about one-fourth to one-fifth of previous levels. Based on this, Akamai said it believes its strategy has proven effective in the market.
Kang Sang-jae, executive director at Akamai Korea, said at a media briefing on the 22nd that while the company initially expected strong demand in gaming when entering the market, demand has continued across various industries since the middle of this year. He added that demand for AI infrastructure among companies seeking lower cloud costs and better service performance is expanding.
Some results from this strategy are beginning to emerge. An Akamai official said that when the company launched its cloud computing business, its goal was to secure a 1% share of the hyperscaler market, and that the target has now been achieved after 4 years. Akamai's cloud computing revenue was tallied at USD 504 million in 2023, USD 630 million in 2024, and USD 708 million in 2025, and cloud computing revenue continues to rise.
Along with cloud computing, revenue from cloud infrastructure services is also growing rapidly. Cloud infrastructure services revenue reached USD 314 million in 2025, up 36% from the previous year.
Akamai plans to expand its GPU infrastructure in the future and also widen the scope of its AI business. Its main target at present is AI inference, and as demand is emerging among enterprise customers for training that uses high-performance GPUs, it is pursuing advance preparation and market response to meet training demand.
Akamai and NVIDIA are continuing to deepen their collaboration. The two companies are combining NVIDIA GPUs with Akamai's distributed cloud infrastructure and are building a structure that efficiently deploys distributed GPU resources and AI workloads based on an AI grid.
Akamai is strengthening the security of its distributed AI infrastructure by combining this with Akamai Zero Trust Segmentation technology. It also plans to secure next-generation GPU infrastructure in the future. Its current GPU support covers inference and relatively light training, and after securing next-generation GPU infrastructure, it plans to respond to full-scale AI training demand.
Source: TECHWORLD · Kim Hye-jin
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407359
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.