ENVALITH
株式会社ABEJA logo

ABEJA,Inc.

5574Growth MarketInformation & Communication

株式会社ABEJA logo
ABEJA,Inc.5574

Business

ABEJA, Inc. is an AI platform company founded in 2012 under the corporate philosophy of "Implementing a Bountiful World." Centered on its proprietary ABEJA Platform, the company supports enterprise companies in the societal implementation of AI on an end-to-end basis, from consulting through implementation and operation. Its business consists of two domains: the "Transformation Domain," which supports AI implementation, and the "Operation Domain," which handles operations through collaboration between humans and AI. The company has a track record of implementation at more than 300 companies across diverse industries including retail, manufacturing, finance, healthcare, and real estate, and is expanding its business domain from digital space to real space (AI robotics), with LLM (Large Language Model)-related demand as its main growth driver. The company listed on the Tokyo Stock Exchange Growth Market in June 2023.

Business Model

The Transformation Domain is centered on flow-type (per-engagement contracts), but because it involves long-term, planned processes, the repeat-customer ratio is high (enterprise repeat-customer revenue ratio of 88.8% for FY2025 (ending August 2025)). The Operation Domain is mainly driven by stock-type recurring revenue. Through a synergistic structure in which the knowledge and data accumulated on the ABEJA Platform are mutually leveraged across both domains, the company has achieved larger per-customer transaction sizes (clients with annual transaction value of ¥50 million or more increased 41.7% year on year) and deeper account penetration.

Company Strengths

The LLM revenue mix rose sharply from over 20% in the prior period to over 50% in FY2025 (ending August 2025). Under NEDO's GENIAC framework, the company succeeded in improving the accuracy of small LLMs, and favorable benchmark results have been promoting deal conversion. The company is accumulating a track record in the highly demanding area of introducing LLMs into mission-critical operations.

While maintaining an enterprise recurring customer revenue ratio of 88.8% (FY2025, ending August 2025), the number of major clients with annual transaction values of ¥50 million or more increased 41.7% year on year, reflecting a trend toward larger deal sizes. Transaction volumes increased for both new and existing clients, and revenue reached ¥3,585 million, up 29.6% year on year.

The company has accumulated knowledge gained from supporting AI implementation across more than 300 companies in a wide range of industries and business types, and has consolidated this knowledge into the ABEJA Platform. Its "digital EMS" model, which provides an integrated offering from consulting through to operation, serves as a differentiating factor from consulting firms and SIers and forms a barrier to entry. The company has also obtained a patent (PCT/JP2018/3824).

ENVALITH's Perspective

Cumulative revenue for the first nine months of FY2026 (ending August 2026) of ¥3,369 million represents 74.0% of the revised full-year forecast of ¥4,550 million, while operating profit of ¥566 million represents 79.7% of the full-year forecast of ¥710 million. The required revenue for Q4 alone stands at ¥1,181 million, representing a 32.4% increase versus the same period last year (¥892 million), but given the accumulation of recurring revenue in the Operation Domain, the likelihood of achievement is considered high. As an external factor, the robust market environment for corporate AI investment appetite serves as a tailwind.

Cumulative quarterly net profit for the first nine months of ¥713 million (up 101.8% year-on-year) includes a negative ¥131 million in income taxes resulting from a reassessment of the recoverability of deferred tax assets. On a pre-tax profit basis, the figure was ¥582 million (up 44.8% year-on-year), and it should be noted that the growth rate in net profit overstates the actual improvement in underlying profitability. Achieving the full-year net profit forecast of ¥852 million presupposes additional tax effects or an accumulation of operating profit in Q4.

The Transformation Domain, which accounts for 74.8% of revenue, is heavily dependent on demand related to LLMs, and as an external factor, amid the rapid evolution and commoditization of AI models, the platform's ability to maintain differentiation is being tested. In addition, the concentrated structure of enterprise customers carries the inherent risk that trends in large-scale order intake directly affect performance volatility. Attention should be paid to whether the expansion into Physical AI & AI Robotics contributes to diversification over the medium to long term.

Growth Strategy

LLMを現軸に据えつつフィジカルAIでリアル空間へ事業領域を拡張

Capturing demand for AI implementation in mission-critical operations, the company is advancing the continuous sophistication of enterprise customers through the ABEJA Platform. It is strengthening its capability to build advanced AI agents through integration with external state-of-the-art models, including the incorporation of NVIDIA Nemotron3.

By accumulating customers in the continuous operation phase through collaboration between humans and AI, the company aims to increase the proportion of recurring revenue, thereby improving the stability and predictability of earnings. In the third quarter cumulative period, sales in the Operation Domain grew steadily, with continued accumulation of recurring revenue.

The company has begun bank-wide DX/AI collaboration with Tochigi Bank, building an architecture incorporating AI governance, risk management, and Human in the Loop. By accumulating implementation track records capable of addressing the complex governance requirements unique to the financial sector, the company aims to expand horizontally into other industries.

The company is advancing the accumulation of technology to connect decision-making in digital space to execution in real space. Through the evaluation and verification of robot foundation models, it positions Physical AI as its next growth domain, aiming to integrate decision-making AI with robotics.

The company has been selected for NEDO's GENIAC program for four consecutive rounds, from the first through the fourth. In the fourth round, it is working with IDOM to build and verify the practical operation of an AI agent specialized in the automotive maintenance domain. In the third round, the company built a long-context-capable, compact LLM and AI agent, and has already verified its operational feasibility for specific business tasks.

Last updated: July 17, 2026