Summary of Key Points
This summer of 2026, AI open platforms have become a new battleground for major tech companies: Qianwen AI has launched an open platform that connects with businesses in over a dozen industries, while Dubao has increased the commission it charges on hotel bookings made through its AI channels (the overall fee rate has risen from 8% to 12%). This marks a shift in the AI industry from a phase of “free spending to generate revenue.” However, businesses are concerned about the additional costs due to multiple platform commissions and whether these orders represent new business or merely a redistribution of existing traffic. The industry also faces issues such as severe homogenization and redundant development. Whether this competition will ultimately lead to the creation of new demand or just a fight over existing market share remains to be seen.
Detailed Analysis
1. AI Platforms Finally Need to “Make Money”
In the past two years, AI assistants (for chatting, writing, drawing, etc.) have been free, but the underlying costs were staggering: For AI products with tens of millions of daily active users, the cost of GPU computing power alone can reach millions per day; if the daily active user count exceeds 100 million, the cost increases significantly. Moreover, previous revenue models (subscription fees, advertising) have not been successful due to low conversion rates and limited scalability. Therefore, tech companies must find new ways to generate revenue, and open platforms provide a solution. They can either build platforms to attract businesses or charge commissions on transactions made through their AI services, essentially shifting from “spending money to acquire traffic” to “converting traffic into income.”
2. Major Companies Employing Different Strategies
- Qianwen AI: Acting as an “Intermediary” with an Open Platform
Qianwen has launched an open platform that integrates with companies like SF Express, Ziru, and Hello. Users can directly use third-party services (such as booking hotels or ordering deliveries) within Qianwen without having to switch to other apps. This turns Qianwen into an “AI service supermarket,” where businesses can set up shop and Qianwen generates revenue from the traffic generated.
- Dubao: Charging Commissions on AI-Enabled Bookings
Dubao has raised the commission it charges on hotel bookings made through its AI assistant from 8% to 12%. This additional fee is considered a “channel fee” for using Dubao’s AI recommendations. Although they describe it as something other than a rent, it essentially means that users have to pay more for transactions facilitated by Dubao’s AI services.
3. Business Challenges
Businesses face practical concerns:
- Increased Costs: Platforms like Ctrip, Meituan, and REDnote already charge commissions, so adding another AI platform would mean additional costs for businesses.
- Quality of New Orders: There is concern that orders generated through AI platforms may not be from new customers but rather from users who were originally intended to use other channels (e.g., users who would have booked a hotel through Ctrip but now do so through Qianwen), resulting in extra commissions with no additional revenue.
- Homogenization and Operational Complexity: AI platforms offer similar services (conversations, services, transactions), but each platform requires separate integration, accounting, and adaptation, increasing operational costs. Additionally, partnering with one platform means becoming part of a larger ecosystem, which can be cumbersome if businesses want to maintain multiple options.
4. Will This Competition Lead to “Sharing the Pie” or “Creating New Pie?”
The industry faces two major issues:
- Redundant Development: Multiple tech companies are creating similar open platforms, leading to resource waste (e.g., multiple teams developing similar AI products).
- Competition for Existing Traffic: If AI platforms merely shift existing user demand from old channels, they are just competing for the same market share. This could result in a situation similar to the mobile internet era, where initial subsidies were followed by price increases.
The true breakthrough lies in “creating new pie” – AI platforms should use algorithms and interactions to generate new demand (e.g., recommending nearby hotels based on users’ travel plans) rather than just distributing existing traffic. Otherwise, this open platform craze might just be a different form of the same old battle for market share.
Conclusion
AI open platforms are a necessary step for tech companies to generate revenue, but their success depends on whether they can bring additional orders and new value to both businesses and users. If they merely lead to redundant development and a redistribution of existing traffic, the industry may repeat the mistakes of the mobile internet era (spending heavily followed by price increases). Only by creating new demand can AI platforms truly make progress. Both businesses and users are waiting to see how this competition will unfold.