Survey Reveals the Reality of AI Infrastructure Development Supporting Corporate AI Utilization

SB C&S Corp. (hereinafter "SB C&S") conducted a "Survey on the Actual Status of Infrastructure Development Supporting Corporate AI Utilization" (hereinafter "this survey") targeting its sales partners to understand the current state of AI infrastructure and challenges in GPU procurement, which are becoming increasingly important with the expanding use of generative AI.

Survey Result Summary

- Over half of the main customer segments for generative AI projects are "mid-sized companies (100-299 employees)" and "small companies (20-99 employees)".
- Regarding full-scale adoption of generative AI, "high interest but little concretization" accounts for the majority at 52.5%, while "full-scale adoption is advancing" remains at only 7.3%.
- The biggest challenge in AI infrastructure operations is the "lack of operational personnel" at 70.4% (multiple answers allowed).
- For AI infrastructure building needs, "support for vendor selection and configuration proposals" is at 52%, and "provision of evaluation and verification environments for the latest GPUs" is at 42.5% (multiple answers allowed).
- Regarding needs for utilizing a GPU verification center, "use as a PoC environment for customers" is at 49.2%, and "use for technical demos and seminars" is at 44.1% (multiple answers allowed).

[Background of the Survey]

With the popularization of generative AI, corporate AI utilization is expanding rapidly, making the AI infrastructure supporting it, centered around GPUs, more important than ever. Especially for the introduction of generative AI requiring advanced computational processing, proper infrastructure development is key to competitiveness. On the other hand, various challenges are becoming apparent in both construction and operation, such as procurement difficulties due to soaring GPU prices and supply shortages, the increasing complexity of vendor selection and configuration design, and the rise in operational costs including power and cooling. Given this situation, we conducted this survey to grasp the current state of AI infrastructure, challenges in GPU procurement, and the needs for external support and verification environments.

[Survey Overview]

- Survey Content: Actual status survey on AI infrastructure and GPU procurement
- Survey Period: February 27 - March 13, 2026
- Survey Conductor: SB C&S Corp.
- Survey Method: Web-based
- Survey Target: Sales partners (179 people)

Main customer segment for Gen AI projects is "mid-sized companies"

The most common customer size for generative AI projects was "mid-sized companies (100-299 employees)" at 29.1%, followed by "small companies (20-99 employees)" at 22.3%, and "quasi-large companies (300-999 employees)" at 19.0%. Meanwhile, "large companies (1,000 or more employees)" accounted for 18.4%, and "micro enterprises (1-19 employees)" for 11.2%. Combining these, mid-sized and small companies account for over half (51.4%) of the total, indicating that generative AI utilization is spreading its base from a large-company focus to a broader audience. Particularly in mid-sized companies, there may be an increasing need for adoption aimed at operational efficiency and productivity improvement, and their relatively swift decision-making and implementation flexibility are thought to be boosting this widespread use.

AI adoption lacks concretization, many stop at PoC (Proof of Concept), full-scale adoption is limited

Regarding the full-scale adoption of generative AI projects, "high interest but little concretization" accounted for the majority at 52.5%, "PoC and trial introductions are increasing" was at 19.0%, and "still limited" was at 21.2%. On the other hand, "full-scale adoption is advancing" remained at only 7.3%. Combining these, it is clear that about 70% (71.5%) are still in the interest or verification stage. From these results, it appears that while PoCs and trial adoptions of generative AI are spreading in companies against a backdrop of growing interest, in many cases they have not reached the point of full-scale integration into business processes or company-wide rollout.

Biggest challenge in AI infra ops is "personnel shortage"

As for items with a high burden in the operational aspect of AI infrastructure, the "lack of operational personnel" was the highest at 70.4%, followed by "hardware renewal and maintenance costs" at 49.2%, "data center contracts and costs" at 32.4%, and "operational costs such as power and cooling" at 23.5%. The "lack of operational personnel" stands out even when compared to other items, suggesting that securing and training specialized talent is not catching up with the advancement and expansion of AI infrastructure. Additionally, costs related to hardware and data centers, as well as operational costs like power and cooling, ranked high, revealing the reality that the burden of maintaining infrastructure is increasing multi-dimensionally with the expansion of AI use. In AI infrastructure operations, both personnel and cost aspects have become major challenges, and efforts to build a stable operational system are deemed increasingly important.

High needs for "vendor selection/configuration proposals" and "verification environments" in AI infra building

Regarding external support expected in building AI infrastructure, "support for vendor selection and configuration proposals" was the most requested at 52%, followed by "provision of evaluation and verification environments for the latest GPUs" at 42.5%, "procurement and delivery adjustment support" at 41.9%, and "post-implementation optimization and maintenance support" at 37.4%. The top two items were key keywords.