A 3-year-old startup that received 100 billion won: Investors 'bet on the future of enterprise AI'

[For more diverse corporate information on startups, visit Unicorn Factory's Big Data Platform Data Lab.]

Building an AI (artificial intelligence) system in-house requires advanced engineering capabilities, while relying on existing SaaS (software as a service) solutions raises concerns about internal sensitive data leaks and high subscription fees that increase with usage.

A startup has emerged to address this issue by offering 'on-premise' products. This is AetherX, which enables companies to rapidly adopt AI with low initial investment while allowing accumulated data and learning models over time to remain as core assets within the enterprise.

AetherX was founded in June 2023 by Kim Gwang-seop (CEO), a former CTO (Chief Technology Officer) of Kakao Brain. Shortly after its establishment, it received investment from AlToss Ventures, a Silicon Valley-based VC known as 'Midas Touch' in the venture capital industry during its seed round.

Recently, the company attracted a large-scale pre-Series A investment worth 100 billion won, further raising its profile in the industry. This round included participation from Smilegate Investment, AlToss Ventures, and KDB Industrial Bank.

AetherX develops 'customer behavior prediction AI infrastructure' that operates within a company's internal data environment. Its flagship solution, 'A2', is installed directly within enterprise systems and runs locally. Through this, it learns real-time customer behavior patterns and precisely predicts customers' 'next actions,' such as purchases, churns, or clicks.

The technical foundation lies in its self-developed 'BFM' (Behavior Foundation Model). While existing LLMs (large language models) primarily focus on text generation, BFM combines multimodal data including user behavior sequences, product images, text, and prices.

It is said that BFM possesses the capability to very precisely predict and infer individual users' next actions. In particular, AetherX realizes 'AI internalization,' enabling companies to move away from dependence on external public AI models or cloud-based SaaS solutions.

Operating without external communication within an intranet, it achieves ultra-low latency real-time response speeds. Additionally, the model has been lightweighted to run even in general CPU (central processing unit) environments without expensive GPU (graphics processing unit) servers, significantly reducing companies' initial adoption costs and long-term TCO (total cost of ownership).

Investment firms evaluated that AetherX precisely addressed the key issues for enterprises: 'data security' and 'AI internalization.' It is expected that a structure that resolves concerns about data leaks and allows each company to operate AI within its own independent data environment without excessive cost burdens will become the future of enterprise AI technology.

Lee Sang-geun, Senior Team Leader at Smilegate Investment, stated, "BFM developed by AetherX has superior performance in 'generalization' capabilities, combining multimodal data such as images, text, and prices to understand context and precisely predict and infer next actions."

Lee Sang-geun (Team Lead) emphasized, "The strength of this model lies in its implementation as an AI infrastructure form that includes ultra-low latency real-time inference and operational automation, enabling stable operation within customer internal data environments such as on-premise and VPC (virtual private cloud)."

In sectors like finance or large content companies, data is difficult to export to external clouds due to security regulations. However, AetherX's solution is installed as infrastructure within the client's firewall, satisfying the demand of companies seeking to protect data sovereignty while assetizing AI technology.

This team leader said, "We have already verified in numbers the actual revenue growth effects for adopting companies in the advertising sector." He added, "Another characteristic is that through 'modular architecture' rather than a massive single software, necessary functions can be flexibly assembled and installed according to the enterprise environment."

An AlToss Ventures representative noted, "While AetherX started with advertising as its entry point, it is essentially building AI infrastructure that predicts and optimizes user behavior. In the long term, there is potential for expansion into adjacent decision-making areas such as content recommendation, CRM (customer relationship management), and personalization."

It is also analyzed that the company has sufficient competitiveness in the global market. The representative stated, "In markets like Japan, where e-commerce scale is large but advertising infrastructure is not yet mature, there exist opportunities similar to those in Korea."

He further added, "We are closely monitoring whether it can establish itself as an infrastructure company that substantially improves platform revenue structures through advertising. The meaning of this investment lies not just in the technology itself, but in accompanying the process of creating a structure where that technology translates into actual business results."

[MoneyToday startup media platform 'Unicorn Factory']