Yanyu Technology Eyes Hong Kong IPO After 3-Year, 3B USD Valuation

As global enthusiasm for large AI models intensifies, can an "intermediary" bridging models and end-users sustain a massive capital narrative without building foundational models themselves? Recent media reports indicate that Evoken has begun preparing for a Hong Kong IPO, with its latest financing round valuing the company at $3 billion. Having risen rapidly over three years through various AI creation and application products, Evoken has reignited industry debates regarding the commercial value and competitive barriers of AI application layers.

From Model Community to "Intermediary" Business Logic: The Evolution of Liblib

Evoken’s journey started with the launch of LiblibAI, an image model community, in 2023. At the time, the open-source image model ecosystem was thriving. The platform gradually evolved into a vertical community akin to a code-hosting site by aggregating models uploaded and fine-tuned by creators. As multimodal large model capabilities advanced rapidly, Liblib upgraded to version 2.0, transforming from a single-model fine-tuning community into a multimodal creation hub. It integrated several mainstream multimodal large models at its core and introduced a user-friendly interface for one-click generation.

Despite expanding product positioning, its underlying business logic remains centered on service packaging by integrating multimodal model capabilities. As an "intermediary" for large models, its profit model primarily relies on securing API cost advantages from official or channel sources and selling them through bundled packages. For instance, the platform uses mechanisms like member point expiration to offset actual usage costs. This approach differs from traditional tool sales, focusing instead on refined operations driven by platform scale and actual user consumption.

Diverse Product Matrix: Rapid Breakthroughs from Design Agents to AI Video Creation

To move beyond the limitations of merely acting as a middleman, Evoken quickly established multiple vertical application products, building its own growth engine.

First, it launched a specialized agent assistant product targeting the design sector. Following the surge in general agent concepts, the team rapidly formed a dedicated product group, focusing on full-process automation and efficiency enhancement in design. Its core strategy involves using agent logic to break down design tasks, optimizing specific functions to address designers' pain points in text-image processing, text modification, and effect alignment, thereby deeply optimizing traditional design workflows.

Second, it introduced an AI video platform for the video creation sector. Capitalizing on the boom in AI short dramas and video generation, these products entered the market quickly by modularizing script generation, storyboard design, visual generation, and post-production editing. This has significantly contributed to the company’s Annual Recurring Revenue (ARR). By keenly capturing industry trends and rapidly iterating product features, the company has secured a first-mover advantage in multiple niche markets.

Moat and Time Gap: Real Challenges in the AI Application Layer

Although Evoken has demonstrated impressive revenue growth and product iteration speed, its business model still faces significant practical challenges and objective scrutiny.

On one hand, the core moat for AI application layer products largely depends on a "time gap." As large model capabilities continue to improve and internet giants enter vertical markets, the differentiated barriers created by first-mover advantages may be quickly compressed. Major manufacturers, leveraging vast ecosystems and resources, are accelerating their layout of similar creative intelligent agents and creation platforms.

On the other hand, user migration costs for AI tools are relatively low, and market competition often devolves into price wars and races in functional iteration speed. Maintaining rapid growth while improving long-term user retention and stickiness, and building more robust technical and ecological barriers, remains a key challenge for such AI application companies in their long-term development. As the IPO process advances, the market will ultimately determine the profitability quality and sustainability of this "intermediary" model.