In the current global frenzy over large models, ‘AI’ alone no longer stands as a scarce label. Instead, what’s truly rare is the capacity for stable, replicable, and sustainable commercialization.

In January 2026, Shandong Extreme Vision Technology Co., Ltd. (hereinafter referred to as ‘Extreme Vision’) formally filed for listing on the Hong Kong Stock Exchange, with CITIC Securities acting as the sole sponsor. Founded in 2015, this AI computer vision firm is striving to carve out a niche in the highly fragmented and fiercely competitive enterprise-level AI market through its ‘Algorithm Mall + Platform-Based Delivery’ model.

However, as investors pore over the prospectus, questions loom large. When will the company’s profitability truly stabilize as its algorithm count and industry reach continue to grow? Can its platform-based narrative hold its own against competition from industry giants and price wars?

From Engineering-Driven AI Firm to Platform Player

Originally launched as Shenzhen Extreme Vision in 2015, the company positioned itself as a provider of AI computer vision solutions, offering end-to-end services—from algorithm development and training to deployment, operation, and maintenance—for clients in industrial, energy, retail, and transportation sectors.

Unlike many AI firms that focus on ‘single industry + customized projects,’ Extreme Vision embraced a platform-based approach early on. Through its ‘Jishi’ developer platform, ‘Jixing’ deployment platform, and ‘Jizhan’ training platform, it aims to create a closed-loop ecosystem spanning algorithm development, trading, deployment, and optimization.

Technologically, these three platforms form a full-process tech stack covering ‘cloud-edge-end’ systems, with the goal of modularizing and commercializing AI capabilities.

The potential advantage of this strategy lies in scale effects and long-tail value. Traditional visual solution providers often concentrate on a few high-value scenarios, such as security and finance. In contrast, Extreme Vision has amassed hundreds of thousands of developers via its platform, attempting to tap into niche markets overlooked or deemed uneconomical by giants across various industries and scenarios, and catering to the more fragmented and customized needs of small and medium-sized enterprises.

As of the end of September 2025, Extreme Vision’s Algorithm Mall had launched over 1,500 algorithms, covering more than a hundred industry scenarios in smart energy, industrial manufacturing, smart transportation, retail, and consumption. It had served over 3,000 clients and completed over 6,000 projects. Its scale and strength are indeed noteworthy.

Behind this success lies the synergy between broad scenario coverage and development agility. Clients don’t need to build an AI team from scratch; instead, they can swiftly find algorithms tailored to their specific needs, akin to selecting apps from an app store. This approach theoretically lowers the threshold and cost of AI adoption significantly. It has also earned Extreme Vision the backing of industrial capital such as China Resources Innovation, Qualcomm China, and Qingdao Economic Control Group. The investment from industrial players not only brings funds but also signals recognition of its platform’s value within specific industrial ecosystems, like energy, manufacturing, and consumption.

However, from a business composition standpoint, despite the rapid growth of its large model business, Extreme Vision’s revenue still heavily leans on AI computer vision solutions. From 2022 to 2024, this segment accounted for over 75% of its revenue for extended periods, while large model solutions only began contributing revenue in 2024, with their share rising to 19.2% in the first three quarters of 2025.

At this stage, its large model business functions more as an incremental tool than a disruptive new engine. This is because Extreme Vision hasn’t ventured into general-purpose large model training; instead, it focuses on delivering customized capabilities for enterprise clients through methods like Retrieval-Augmented Generation (RAG), multi-agent systems, and scenario-based fine-tuning.

In other words, while Extreme Vision’s vision of algorithm platformization is distinctive and has achieved certain successes, it hasn’t yet formed an industry-level ‘moat.’ The company remains more of a mid-sized, technology-driven firm reliant on delivery capabilities.

Revenue Rising, Cash Draining

In reality, under relatively constrained conditions, there’s a natural tension between the breadth and depth of a platform. The prospectus reveals that Extreme Vision’s revenue shows significant fluctuations by geography and client type.

Notably, its core revenue source rapidly shifted from East China to South China within a year. This hints at a possible over-reliance on large projects in a few regions, rather than the stable and widespread long-tail revenue expected from a platform model.

Furthermore, its largest client contributed only about 8% of its revenue in 2024. While not overly concentrated, this indicates that its platform still has a long way to go before forming a strong network effect and a solid competitive barrier.

Under these limited competitive advantages, Extreme Vision faces a dilemma that many startup AI large model companies struggle to overcome quickly—revenue growth without corresponding profit growth, or even increased losses despite revenue growth.

According to the prospectus, from 2022 to 2024, its revenue surged from 102 million yuan to 257 million yuan, showing impressive growth. In 2024, it achieved a brief profit of 8.708 million yuan, thanks to the explosion of its large model solutions business.

However, the good times were short-lived. In the first three quarters of 2025, it recorded revenue of 136 million yuan but posted a loss of 36.3 million yuan for the period, with a net cash outflow of 21.37 million yuan from operating activities. More alarmingly, after adjusting net profit turned positive at 20.49 million yuan in 2024, it plunged into a loss of 11.94 million yuan in the first three quarters of 2025.

Behind this, Extreme Vision’s sales, management, and research and development expenses have remained high for an extended period. In 2022, the combined total of these three expenses was nearly equal to its annual revenue. Even after its revenue doubled in 2024, the proportion of these three expenses rebounded to 74.39% in the first three quarters of 2025. This suggests that its business model still heavily relies on manual delivery and project-based sales, failing to truly achieve the scale leverage of a software company.

More serious than the losses is the quality of cash flow.

Cash flow is the lifeblood of a company. However, from 2022 to the third quarter of 2025, Extreme Vision’s net cash flow from operating activities remained negative, with a cumulative net outflow exceeding 190 million yuan. Another severe signal comes from its collection efficiency, with its total trade receivables and bills receivable soaring from 42.015 million yuan at the end of 2022 to 181 million yuan at the end of September 2025. Correspondingly, its accounts receivable turnover days deteriorated from 99 days to 379 days. A sales collection cycle exceeding one year means that Extreme Vision is increasingly trapped in the path of ‘working first and collecting money later,’ with business expansion consuming far more funds than its profit-generating capacity.

The company attributes this to an increase in public clients such as governments and state-owned enterprises, which have longer payment processes, as well as the seasonal nature of its business. However, this doesn’t fully allay market concerns. After all, in TO B and TO G businesses, a prolonged collection cycle not only ties up a significant amount of working capital and exacerbates cash flow pressure but also significantly increases the risk of bad debts.

Valuation Stagnation and Platform Anxiety: Extreme Vision’s True Test Lies After Listing

Extreme Vision operates in the emerging enterprise-level computer vision solutions market, a track brimming with opportunities but also fierce competition. According to a Frost & Sullivan report, Extreme Vision ranked eighth with a 1.6% market share in 2024, while ‘Company A,’ ranked first, held a 12.1% share. The market is highly fragmented, with no absolute leader yet, but the advantages of leading companies are already evident.

On one hand, Extreme Vision inevitably faces pressure from vertical giants and full-stack AI companies. In the security field, giants such as Hikvision and Dahua Technology deeply bind clients with their hardware advantages, often offering software solutions as attachments to their hardware. At the cloud service level, giants such as Baidu Intelligent Cloud, Alibaba Cloud, and Tencent Cloud provide integrated solutions from computing power and algorithms to platforms, wielding strong ecological influence.

Furthermore, compared to AI listed companies such as SenseTime and Innovusion, which have already established brand effects, Extreme Vision still lags in terms of capital scale and brand awareness.

More importantly, while Extreme Vision’s core narrative of building a platform ecosystem akin to an ‘AI Algorithm App Store’ is attractive, ‘algorithm quantity’ doesn’t necessarily equate to ‘bargaining power.’ At least at this stage, Extreme Vision hasn’t fully demonstrated this potential.

On the other hand, its valuation growth has shown signs of fatigue. The primary market is the most sensitive barometer. After completing its Series D financing in November 2024, Extreme Vision’s post-investment valuation was approximately 2.31 billion yuan, representing a growth of over 243 times compared to its angel round in 2015. However, compared to its valuation of 2.3 billion yuan after the Series C3 financing in October 2022, it has barely grown in over two years, with a mere 0.4% increase.

Amid the AI investment boom, this may suggest that the capital market is becoming more cautious in reevaluating its current business model and growth potential. Capital has begun to reprice AI companies based on ‘cash flow and replicability’ rather than purely technological narratives and scenario quantities.

Therefore, Extreme Vision’s ‘narrow path survival’ strategy hinges on its extreme ‘horizontal scenario expansion’ capability. Unlike giants focusing on ‘main tracks’ such as urban governance and autonomous driving, it gathers long-tail and fragmented visual application demands through its platform, seeking niche markets overlooked or deemed uneconomical by giants across various industries and scenarios.

Its newly launched large model solutions aim to leverage industry knowledge bases and multi-agent technologies to extend this ‘horizontal tapping’ capability from vision to broader enterprise intelligence. This may be the key to attracting industrial capital and seeking a second growth curve.

With this Hong Kong listing, Extreme Vision plans to use the proceeds for research and development, enhancing commercialization capabilities, and supplementing working capital. However, overall, issues such as profitability, collection cycles, and revenue models won’t be automatically resolved through listing.