ZipCodeAI's "Turnaround": From "Selling Resources" to "Providing Services" - A Detailed Analysis of the Business Logic Behind the $5 Billion Financing
Hello everyone, I'm your financial journalist. Today, we're talking about one of the hottest up-and-coming companies on the Hong Kong stock market – ZipCodeAI (2513.HK).
On September 16th, ZipCodeAI held a teleconference to share the latest developments with analysts and investors. In simple terms, they communicated to the market: "The funds are in place, we have enough computing power, and we're ready to expand our business. This time, we're not just selling models; we're selling 'labor power.'"
As a result, the company's stock price soared by 5.88%, pushing its market value to HK$335.2 billion. What exactly happened behind this? Don't let the jargon like "ARR" and "PFLOPS" confuse you. I'll break down the core logic of the teleconference into five key points, so you can understand ZipCodeAI's current strengths and future prospects.
---
1. Funds in Place, Bottleneck Eliminated: The Power of $5 Billion
Key Insight: Previously, they had the resources but lacked the means to deliver them. Now, with $5 billion in hand, they have the capacity to meet the market demand.
In the AI industry, the most valuable asset is computing power – the “energy” and “muscle” of AI systems. In February this year, ZipCodeAI released the GLM-5 model, which became an instant hit, increasing demand tenfold. It's like a popular restaurant suddenly receiving ten times the usual number of customers, but with only ten cooking stations – the food ran out, and they even had to temporarily stop selling their main product, Coding Plan.
Money is crucial at this point. On September 11th, ZipCodeAI completed a massive $5 billion financing round. Where did this money go? It was mainly used to purchase more computing power and build data centers.
The management outlined the benefits:
- Investment: Raising 30 billion RMB to acquire 100,000 P of computing power (a significant amount).
- Allocation: 40% for developing new models, 60% for generating revenue.
- Return: With the high cost-effectiveness of the GLM-5.3 model, the gross profit from business operations could theoretically reach 80%. At full capacity, they could earn 40 billion RMB annually, not only recouping their costs but also covering research and development expenses.
In other words, before, ZipCodeAI was like a small horse pulling a heavy cart; the business demand was too much for their limited resources. Now, with a “heavy truck” (new computing power), they can not only handle the load more efficiently but also reduce costs per unit of work, increasing their profit margins. Computing power is no longer a limiting factor; it has become a critical infrastructure for revenue growth.
---
2. Business Model Upgrade: From Selling Tokens to Selling Results
Key Insight: Previously, they were like selling water by the liter; now, they're offering professional services.
This is the most significant announcement from the teleconference. How do large-scale AI companies make money?
- Stage One: Selling API interfaces, charging per word (like selling water).
- Stage Two: Providing coding assistance (like hiring a chef to write code).
- Stage Three (current): Collaborative work (Co-work).
ZipCodeAI aims to integrate AI directly into businesses’ professional workflows.
For example:
- Previously: Companies bought security software, and employees monitored alerts; AI provided supplementary alerts.
- Now: AI automatically identifies vulnerabilities, audits code, and strengthens systems. Companies are buying the results of AI’s work, not just the right to use it.
The management reported that within a month of the GLM-5.3 release, the Co-work service generated over 1 billion RMB in orders. Although this isn’t yet full revenue, it shows that businesses are willing to pay for AI’s services.
What does this mean? AI companies’ revenue sources are expanding from software purchases to professional services and even labor costs. If AI can replace high-paying professionals, their market potential is enormous.
---
3. Why Start with Cybersecurity?
Key Insight: They chose cybersecurity as the first target because it’s a high-value, urgent, and easily quantifiable area.
- High Barriers: Tasks like vulnerability detection and code auditing require specialized expertise, which AI (especially GLM-5.3) excels at.
- Frequent Use: Security scans and alerts are continuous, so the more AI is used, the better the results and the stronger the customer loyalty.
- Clear Budget: Companies have fixed cybersecurity budgets, and AI can help reduce costs (by reducing staff or preventing incidents).
Over 100 security companies have already integrated GLM into their products, proving that ZipCodeAI is delivering real value.
In simple terms, cybersecurity is like the emergency department of a hospital – there’s a high demand, urgent cases, and customers are willing to pay. By establishing themselves in this area, ZipCodeAI demonstrates AI’s effectiveness before expanding into other sectors.
---
4. Partnering with Cloud Providers: Expanding Globally
Key Insight: Instead of building their own sales network, they’re leveraging existing cloud platforms.
ZipCodeAI has partnered with top cloud service providers (AWS, Azure, Alibaba Cloud, etc.) to distribute their services.
- How it works: They host their models on these platforms, and users access them through the clouds; ZipCodeAI shares the revenue.
- Revenue Start: Revenue is expected to start in October.
This strategy is smart:
- Cost Savings: No need to build a large sales team; cloud platforms have a vast customer base.
- Fast Expansion: It allows them to quickly enter global markets without complex international setups.
- Stable Cash Flow: Cloud providers are reliable customers, ensuring steady revenue.
In other words, ZipCodeAI is like a brand collaborating with major retailers like Walmart or Amazon. Although they share profits, the increased sales volume leads to significant benefits.
---
5. Risk Warning: Don’t Celebrate Too Soon
Key Insight: While the prospects are promising, we need to see the actual results.
- Orders vs. Revenue: The 1 billion RMB in Co-work orders are just contracts; revenue hasn’t been fully realized.
- Fierce Competition: The AI market is highly competitive, with companies like OpenAI, Anthropic, and domestic players like Baidu and Tencent.
- Computing Power Risks: The large investment in computing power could become a burden if business growth doesn’t meet expectations.
- Replication Challenges: While cybersecurity is a successful model, replicating this in other fields (finance, law, etc.) is uncertain due to different industry-specific challenges.
---
Summary: ZipCodeAI’s New Story
The teleconference outlined ZipCodeAI’s transition from a technology-driven to a business-driven company:
- Foundation: $5 billion in financing provides the necessary hardware for expansion.
- Core: Moving from selling models to providing professional services, opening up new market opportunities.
- Channels: Partnerships with cloud providers for global expansion.
- Goals: The management has raised their annual recurring revenue (ARR) forecast from $2.4 billion to $3 billion, showing confidence in the future.
For investors, this means ZipCodeAI is evolving from a model developer to a provider of AI-based services. If they can prove that AI can replace human professionals and generate sustainable revenue, their valuation could shift from a tech stock to a high-growth service/software stock, with significant potential.
However, we need to monitor key metrics:
- Revenue from cloud partnerships in October.
- Conversion of Co-work orders into actual revenue.
- Customer retention and new customer growth.
ZipCodeAI’s stock price already reflects some of these optimistic expectations, but the real test will come in the coming months.