Is iFLYTEK's "Bet" on Large Models Worth It?

Original Content by the Digital Economy Studio of Youshu

Author | Youshu

WeChat ID | yds_sh

The cognitive large models represented by ChatGPT are universally recognized as a significant technological leap in the history of AI development, and many domestic tech enterprises have also joined the battle for large models. In order to seize the opportunities of AI development, tech giants have continued to invest substantial amounts of money. However, despite the huge investments, tangible results are difficult to achieve in the short term, and the growth brought by AI is still in the brewing stage. This is the problem that leading tech companies represented by iFLYTEK are currently facing.

Giants Compete to Join the Game, the Battle of "Models" Rages On

Large models, also known as foundation models, refer to machine learning models with a large number of parameters and complex structures. The level of intelligence they possess determines whether the final products are "smart" or not. Therefore, insisting on promoting the independent research and development and training of large models, despite the huge investments required, helps to improve their quality and iteration, and drives the collaborative development of related industries in the entire AI field.

The emergence of ChatGPT has sparked the rapid emergence of hundreds of large models, igniting the chaos of large models: Alibaba Cloud released Tongyi Qianwen 2.5, with model performance surpassing GPT-4 Turbo; ZeroOne launched the closed-source large model Yi-Large with 100 billion parameters; OpenAI released GPT-4o, the most human-like AI large model so far; Google held its developer conference and updated Gemini across the board; on the same day, ByteDance released the Doubao large model family, including multimodal large models; iFLYTEK launched the iFLYTEK Spark large model, with its developer ecosystem growing rapidly, adding 1.279 million developers in the first half of 2024 alone...

Investment and financing highlights in large models by tech giants have been frequent. OpenAI has received $14.3 billion in funding, with a latest valuation exceeding $80 billion; Anthropic, which is also in the first tier of large model development alongside OpenAI, has raised over $7.5 billion, with a valuation of $18.4 billion. According to CBInsights data, generative AI startups globally received approximately $20.4 billion in funding in 2023, more than five times the $3.6 billion in 2022.

In China, according to IT Orange, there were 530 total financing events in the AI sector in the primary market in 2023, with a total financing transaction value of approximately 63.1 billion yuan. Zhipu AI, MiniMax, DarkSide Moon, and Baichuan AI became the "new AI Four Dragons" under the enthusiasm of capital.

According to incomplete statistics, by the end of May 2023-2024, Alibaba, including Ant Group, had invested in 7 large model-related companies, with a total investment amount exceeding 3.5 billion yuan. Among them, according to Alibaba's financial report, the company invested a total of approximately $800 million in the AI startup DarkSide Moon in fiscal year 2024, acquiring approximately 36% of its equity. In MiniMax's new round of funding, Alibaba led the investment of at least 600 million yuan and has also invested 1.64 billion yuan in Zhipu AI.

Tencent has invested in large model companies such as Zhipu AI, Minimax, Ba