$25.5 Billion Funding Influx: How Can Large Models Monetize?
-
Large Model House estimates that the global large model market will reach $21 billion in 2023 and grow to $109.5 billion by 2028. In China, the large model industry market is expected to reach 14.7 billion yuan this year and expand to 117.9 billion yuan by 2028.
Data from PitchBook, a foreign venture capital analytics firm, shows that in the first half of 2023, there were 1,387 financing deals in the global AI sector, raising a total of $25.5 billion, with an average funding amount of $26.05 million per deal.
[Image: WeChat Image_20230809104207.jpg]
According to Titanium Media data, about 20 domestic large model companies disclosed investments in H1 2023, with funding typically ranging from tens of millions to hundreds of millions of yuan. Among them, the startup MiniMax secured the largest funding round - over $250 million on June 1, valuing the company at more than $1.2 billion.
[Image: WeChat Image_20230809104207.jpg]
Currently, domestic large model applications primarily focus on industry-specific implementations, shifting from general-purpose models to vertical industry models.
In customer service, large models significantly enhance chatbot dialogue capabilities and improve user satisfaction;
In recruitment, graduates use AIGC to write resumes, simulate interviews, and get job recommendations, while HR professionals leverage large models for job postings, resume screening, and interview coordination;
In design, fashion designers can generate countless clothing designs in minutes using AIGC, and interior designers use AIGC to provide clients with unlimited inspiration.
In the field of smart manufacturing, large models can reduce the training costs for industrial quality inspection, shorten deployment time, address cross-model multi-category defect generation and automatic labeling issues, and enhance the batch replication of AI. (Click to view "3 Million People to Be Replaced by AI: Who Will Break Through the 240 Billion Market First?")
The introduction of large model filing requirements in July highlights the government's emphasis on data security, challenging the model of relying solely on API interfaces to access foreign large models.
"As far as I know, To C large model apps that started development at the beginning of the year have encountered varying degrees of difficulties, including startups integrating ChatGPT APIs," said Wang Minghui of Jing Capital in a media interview. "To C large model startups must prioritize regulatory and compliance issues, covering both data security and AI safety. Their development strategies need a forward-looking perspective to align with the gradual refinement of regulations in the emerging tech industry. Under the current filing-based management approach for large models, products that obtain approval first naturally gain a first-mover advantage." "This has shifted investment focus toward vertical industry large models or those optimized for specific data scenarios."
"When we saw what seemed like a structural opportunity larger than the internet, we thought we could make a name for ourselves overnight, only to end up as cannon fodder." This reflects the sentiment of many application-layer entrepreneurs in the first half of the year. (Click to view "Entrepreneurs in the Era of Large Models")
Currently, four business models have emerged for B2B applications of large models, including transaction-based fees, custom development fees, service fees, and subscription fees.
Transaction-based fees — Charges are based on the number of API calls or transactions a customer makes per month. Pricing is typically calculated per transaction volume, such as a fee per thousand API calls.
Custom development fees — If a client requires an AI model tailored to a specific domain, companies usually charge a custom development fee. Pricing typically depends on the complexity and time required for development.
Service Fees – Charged based on provided data processing, labeling, and quality control services.
Subscription Fees – Customers can choose different subscription tiers (Basic, Standard, or Premium) as needed. Subscription fees are typically charged monthly or annually, with pricing based on the quantity and type of services required.
Additionally, the text-to-image field has pioneered three mature business models, representing three distinct corporate development paths:
- Stability.AI builds an open-source ecosystem
- Midjourney creates SaaS and MaaS ecosystems
- Adobe Firefly develops traditional tool ecosystems by integrating all AIGC functions into its tools
Midjourney, leading in image quality, is the preferred choice for most C-end users. Stability.AI suits startups and SMEs by offering private deployment and optimization to solve various long-tail problems. Firefly enters the toolbox of large enterprise clients through Adobe's comprehensive software suite.
Three paths are expected to coexist long-term in the fields of text-to-image, text-to-video, text-to-music, and text-to-novel generation, complementing each other.