DeepSeek Raised Its Registered Capital 65% in 2026, and Its Founder Says AGI Is the Only Goal

DeepSeek carried out two capital increases in 2026, raising its registered capital by 65% — from 10 million to roughly 16.45 million yuan. Founder Liang Wenfeng (梁文锋) told investors at a closed-door meeting that the company has "only one main line — the path to AGI." 36Kr reported this on July 23, 2026, citing the Tianyancha service and media publications.

How much did DeepSeek's capital grow?

The registered capital of Hangzhou DeepSeek (杭州深度求索), a company affiliated with DeepSeek, rose from 10 million to roughly 16.45 million yuan in 2026. That is a 65% increase, driven by two consecutive capital increases. According to Tianyancha, both capital changes were entered in the registry this year. Registered capital is a legal registration figure, but its growth reflects a revision of the lab's funding structure.

The company was registered in July 2023 in Hangzhou, and its legal representative is Pei Tian (裴湉). The main shareholders are Liang Wenfeng himself and the management consulting partnership Ningbo Chengen.

• Founded: July 2023, Hangzhou
• Registered capital: 10 million → ~16.45 million yuan (+65%)
• Capital increases in 2026: two
• Legal representative: Pei Tian
• Business scope: engineering and technical research, software and AI application development

What did the founder say about AGI?

Liang Wenfeng (梁文锋), DeepSeek's founder, said at a four-hour closed-door meeting with investors that the company is not spreading itself across multiple directions — it has a single goal. According to 36Kr, a recording of the meeting has circulated among tech and venture capital circles in China and become a subject of wide discussion.

"DeepSeek has only one main line — the path to artificial general intelligence (AGI)," —

Liang Wenfeng, DeepSeek founder, at a closed-door meeting with investors.

The statement stands out because many Chinese AI companies are diversifying aggressively into applied products and agents in 2026. DeepSeek, by contrast, is publicly committing to fundamental research — building AGI-level models — rather than monetizing through a portfolio of services.

Why does this matter for the market?

DeepSeek is one of the few Chinese AI labs whose open models (the V and R series) competed with flagship models from OpenAI and Anthropic in 2025–2026 at a fraction of the training cost. Its public "AGI only" stance, paired with the 65% capital increase, shows that the company intends to remain a research lab rather than turn into a product holding company.

For the industry, this is a rare signal: while competitors broaden their product lines, DeepSeek says the opposite — that it is concentrating resources on a single fundamental problem. That course also explains the capital increases: the race for AGI-level models requires stable funding for compute.

What does this mean?

DeepSeek is choosing a narrow strategy: rather than expanding in breadth, it is aiming at a single point — AGI. For the market, this is a rare example of a Chinese lab publicly rejecting diversification in favor of a fundamental goal, and a signal that investors are willing to back that course with money.

Frequently asked questions

What is DeepSeek?

DeepSeek is a Chinese AI lab (legal entity Hangzhou DeepSeek, 杭州深度求索), registered in July 2023 and affiliated with founder Liang Wenfeng. It is known for its open language models in the V and R series, which compete with OpenAI's models at lower training costs.

How much did DeepSeek increase its capital in 2026?

The registered capital of the affiliated company rose 65% — from 10 million to roughly 16.45 million yuan — through two capital increases recorded in the registry in 2026, according to the Tianyancha service.

What is DeepSeek's main goal?

According to founder Liang Wenfeng, the company has "only one main line — the path to AGI" (artificial general intelligence). He made the statement at a four-hour closed-door meeting with investors.