Recently, it has become undeniable that DeepSeek has taken the internet by storm. Its impact is on par with the launch of ChatGPT, and it seems everyone is integrating DeepSeek into their operations, leaving those who lag behind. However, amidst this surge in adoption, a pertinent question arises: Is the pursuit of self-developed large models still warranted?

I. The Universal Embrace of DeepSeek

According to China Wealth Network, DeepSeek has not only catapulted China's generative AI technology into the global spotlight but has also ignited the entire AI industry chain. Cloud service providers, chip companies, and downstream application developers are all actively deploying and adapting to DeepSeek. The industry anticipates the next year to be a golden era for AI applications, with Omdia forecasting a 5.5-fold increase in China's generative AI market size, reaching $9.8 billion by 2029.

Major domestic cloud providers such as Tianyi Cloud, Mobile Cloud, Unicom Cloud, Huawei Cloud, JD Cloud, Alibaba Cloud, and Baidu Intelligent Cloud have all integrated DeepSeek into their offerings. International players like Microsoft Azure and Amazon Web Services also launched DeepSeek before the Spring Festival.

WeChat has integrated the DeepSeek-R1 model, allowing users to access its deep-thinking capabilities through AI search in the WeChat search box. Baidu, too, plans to fully integrate DeepSeek and ERNIE Bot's latest deep search capabilities into its search business.

Not only private enterprises but also governments are jumping on the bandwagon. The Guangzhou Government Service and Data Management Bureau has officially deployed the DeepSeek-R1 and V3 671B large models on the government extranet, aiming to integrate DeepSeek into various government services, administrative offices, and urban governance initiatives. Shenzhen has also made DeepSeek model application services available to all districts and departments through the government cloud environment.

Financial institutions are not far behind. Taiping Life, PICC Property and Casualty, and PICC Asset Management have all announced the completion of local deployments of DeepSeek within their organizations, leveraging it to create AI assistants and exclusive interactive applications.

Similarly, the mobile phone industry has been swept by the DeepSeek wave. As of February 18, major domestic manufacturers such as Huawei, Honor, Nubia, Meizu, Xiaomi, vivo, and OPPO have all integrated DeepSeek into their offerings.

II. The Necessity of Self-Developed Large Models

With DeepSeek's rapid rise, a pertinent question emerges: Is the investment in self-developed large models still justified?

Firstly, the widespread adoption of DeepSeek is inevitable. Its low cost and open-source nature are significant drivers behind this trend. DeepSeek breaks the high cost barrier associated with traditional large models, making high-quality AI services accessible to even small and medium-sized enterprises. The open-source feature further facilitates customization and optimization, attracting a plethora of developers and enterprises.

Secondly, DeepSeek's core value lies in AI equity. While it boasts impressive technical capabilities, its true strength lies in its cost-effectiveness and wide applicability, akin to China's manufacturing prowess in the global market. DeepSeek offers large model solutions comparable to or even surpassing international high-priced products at a fraction of the cost.

However, enterprises face a dilemma. Abandoning self-developed models would mean losing the significant investments made in them, as well as the technological dominance crucial for long-term competitiveness. On the other hand, self-developed models offer the advantage of tailored solutions that align with unique business needs, fostering technical innovation and talent accumulation.

In this scenario, enterprises must learn from DeepSeek's technical logic and methods to reduce R&D costs while maintaining their own characteristics. Independent R&D remains vital in specific fields like healthcare and finance, where data security and privacy are paramount.

Lastly, while DeepSeek has disrupted the AI market, it has also spurred industry progress. The AI field is experiencing a "steam engine moment," akin to the Industrial Revolution, propelling humanity into a new era of intelligence. Whether embracing DeepSeek or pursuing independent R&D, the ultimate goal for enterprises is to harness AI technology for sustainable growth and innovation.

In conclusion, while DeepSeek offers compelling advantages, self-developed large models still hold significant value, particularly for enterprises seeking to maintain their technological edge and meet specific business needs.