Domestic Large-Scale Language Models Target Overseas Markets, but the Window of Opportunity Is Short (AI Translation)

【周刊提前读】国产大模型先征海外 但窗口期并不长

As domestic AI models push internet giants into price wars, overseas markets have emerged as critical battlegrounds for AI applications. Major applications are incorporating AI-generated content (AIGC) features to solidify their market positions, while startups are seeking opportunities to compete with tech giants. ByteDance, dubbed China's most successful "app factory," is rolling out AI features across core applications, including TikTok and CapCut, which have seen significant user engagement and spending. [para. 1][para. 2][para. 3]

ByteDance’s TikTok, the first non-gaming app to surpass $10 billion in user spending, has over 1 billion monthly active users and is integrating AI features like chatbots and music generation amidst the AI large model craze. ByteDance has also launched CapCut, achieving over 600 million annual downloads and $130 million in user spending. Beyond TikTok and CapCut, ByteDance has introduced several products overseas including the chatbot "Doupao Overseas Edition" (Cici), AI education product Gauth, and image-sharing app PicPic. [para. 2][para. 3][para. 4]

Not limited to giants, many AI application companies are expanding overseas, utilizing large models like GPT. These include AI photo editing, AI video editing, and text-based AI office assistants. Overseas markets show a higher willingness to pay for AI applications and SaaS (Software as a Service) models, making early profitability more attainable. [para. 4]

Prominent Chinese internet companies like Kuaishou and Tencent are intensifying application development and updates. Kuaishou’s text-to-video model "Keling" has increased video durations and expanded internal testing, while Tencent's "Yuanbao" introduced a feature for generating 3D characters, reflecting their marketing and distribution advantages. [para. 9][para. 10]

In 2023, a surge in AI photo editing applications demonstrated significant competition in AI applications. Over 490 new AI photo editing apps emerged, a 192% yearly increase. The text generation field saw new players like copy.ai, Writesonic, and Microsoft's Copilot, alongside Chinese counterparts such as BowWave Intelligent and Lingyi Wanwu’s siuuu.ai. [para. 10]

However, international expansion isn't straightforward due to cross-border data and regulatory challenges. Companies must establish separate servers and align AI applications with regional models, such as domestic large models in China and OpenAI's models overseas. The United States is also considering restrictions on AI investments related to China. [para. 5][para. 15]

Startups have unexpected opportunities overseas with improvements in large model capabilities. For instance, HelloBoss, a Japanese recruitment app, quickly integrated GPT for AI-generated resumes and job postings. This competitive edge led to substantial user growth and partnerships with major companies, aiding funding and expansion. [para. 12][para. 13]

Following the release of GPT-4, internet giants are ramping up foundational large models and attracting developers through services like public cloud and APIs. Pao Ding Technology’s ChatDOC, which integrates over 20 large models, focuses on user-specific prompts for optimization. Meitu’s AI-enhanced photo editing tools have expanded globally, leading to significant revenue from international users. [para. 14][para. 19]

While AI applications see success overseas, replicating this in China remains challenging. Differences in market dynamics and user behavior are evident, such as Remini's dominance abroad compared to limited success domestically. Despite this, the willingness of overseas users to pay is a crucial survival factor. [para. 21][para. 22]

Startups like MiniMax and Waveform Intelligence demonstrate the benefits of simultaneous domestic and international launches. This approach facilitates user feedback, model optimization, and broader market engagement. Enterprise software firms also find profitability in mature overseas markets compared to domestic competition. [para. 28][para. 30]

As various software and SaaS companies expand overseas, they're leveraging lower prices and broader service coverage to withstand competition. For instance, Laiye Technology has adapted its focus towards regions like Hong Kong and Singapore to tap into larger markets in Southeast Asia and the Middle East. [para. 31]

AI application companies face intensifying competition, with investors emphasizing overseas market potential. Key strategies involve aligning product capabilities with local needs and leveraging established cloud and hardware ecosystems for global distribution. However, the decoupling of foundational models between regions necessitates distinct operational setups, significantly increasing costs and complexity. [para. 38][para. 40]

Ultimately, the regulatory environment demands clear "identity" resolutions for global operations. Investors encourage timely exits, suggesting that selling to larger corporations may be the optimal strategy for startups in an increasingly competitive and regulated AI landscape. [para. 43]