Author: ThinkAI, Evan
“Whoever can connect with DeepSeek for financing will receive a reward of 500,000.”
At the beginning of April this year, DeepSeek's financing plan was exposed for the first time, shaking the capital market, with fund managers even publicly offering 500,000 in their social circles for connections.
Now that DeepSeek has successfully secured financing, the total raised amounts to 7.4 billion USD, approximately 50 billion RMB, with founder Liang Wenfeng investing 20 billion, becoming the largest investor.
In addition to Liang Wenfeng, Tencent contributed 10 billion, CATL 5 billion, and JD, NetEase, and IDG Capital each contributed 3 billion RMB.
The National AI Industry Investment Fund has also participated, contributing 1 billion RMB this round and holding voting rights.
After this financing, DeepSeek's total valuation exceeded 350 billion.
Yet, just a year ago, Liang Wenfeng was on stage confidently proclaiming the “Three No Theorems”: No financing, No IPO, No commercialization.
Why has DeepSeek, which once aimed to create a company “not for making money through an IPO,” now turned around?
From Top Scholar to Quantitative Genius
In 2002, the current founder of DeepSeek, Liang Wenfeng, scored 806 points to become the top scholar in the Zhanjiang college entrance examination.
With scores reaching the admission line of Tsinghua University, he chose to apply for the Electronic Information Engineering program at Zhejiang University as his first choice.
Liang Wenfeng was born into a family of teachers and attended Wuchuan No. 1 High School.
He excelled in science during high school, and under the focused training of his physics and chemistry teachers, his ranking jumped from the top 100 in his first year to the top 10, maintaining a position in the top 3 throughout his final year.
He began learning computers in middle school, often volunteering to repair school computer equipment.
When applying for the college entrance examination, Liang Wenfeng's scores met Tsinghua University's admission threshold, yet he ultimately chose to apply to Zhejiang University's Electronic Information Engineering program as his first choice.
According to relevant reports, Zhejiang University ranked third nationally in comprehensive strength at the time, while Electronic Information Engineering was his ideal major.
In 2008, Liang Wenfeng began his journey in quantitative trading with an initial capital of 80,000 RMB, founding Phantom Quant in 2015 at the age of 30.
“Phantom” originally comes from the Luo Book's Nine Palaces in the River Map and Luo Book, also known as the Longitudinal and Horizontal Diagram, expressing an ultimate pursuit of numerical algorithms.
In 2019, “Phantom Quant” became the first quantitative hedge fund in China to raise over 100 billion RMB.
According to related agencies, Phantom's annual net profit is estimated at 5 to 6 billion, far exceeding the annual financing total of the vast majority of domestic AI unicorns.
DeepSeek's Peak from the Start
After Phantom Quant's performance retracement in 2021, founder Liang Wenfeng made a bold move, spending heavily on purchasing high-end NVIDIA A100 graphics cards priced at over 100,000 each.
He made an early bet on the large model track.
By 2023, only five companies in China possess over 10,000 GPUs, with Phantom being the only independent entity, aside from major internet firms, to stockpile over 10,000 A100s.
This batch of pre-stocked high-end computing power has become the core strength for the development of self-researched large models, directly giving rise to DeepSeek.
On December 26, 2024, DeepSeek V3 was officially released, and the user growth curve post-launch is considered a miracle in the industry.
By January 2025, the platform had accumulated 125 million new users, securing 100 million users within just 7 days,
surpassing ChatGPT, which took two months to reach the same milestone, and suddenly becoming a phenomenal dark horse among domestic large models.
New Crises Following the AI Boom
Behind the impressive user data lies a deep-seated hidden danger capable of shaking foundations, with three major pressures overwhelming DeepSeek, completely shattering the long-held myth of “not relying on external financing.”
Core talent continues to drain, and without an equity mechanism, it cannot retain top research and development personnel.
Luo Fuli was lured by Lei Jun to join Xiaomi, and Guo Dayan switched to ByteDance to take on the role of head of Agent business; counting these two, at least five core R&D personnel have left successively.
Currently, competition for talent in the AI industry is fierce, and equity incentives are standard for top firms to retain elite engineers, while DeepSeek, which has long refrained from financing, lacks equity option incentive tools and has no competitiveness in the talent war.
The competition for computing power is intensifying, and self-funding struggles to support huge consumption.
The competition among large models is no longer about algorithms but rather about frantically stacking graphics cards and burning money to compete in computing power. Giants like OpenAI, ByteDance, and Alibaba use tens of thousands of high-end graphics cards for training, with monthly expenses on data centers and electricity reaching hundreds of millions.
DeepSeek is set to launch a new version that requires 2-3 times the computing power than V3 for training and daily operations. The procurement price for a high-end H100 graphics card is over 200,000, and expanding additional graphics cards will require an investment of several billion.
The thousands of A100 computing power previously stockpiled by Phantom are already insufficient, and relying solely on self-funding earned from quantitative trading cannot keep pace with the endless computing power investments by big firms; the funding gap will only grow larger.
The window for IPO valuation is limited; missing it will mean losing development opportunities.
After Zhipu AI and MiniMax successfully went public, their stock prices surged, and Starry Sky is also advancing the preparations for an IPO.
If DeepSeek misses this capital market boom, it will face significant risks of a sharp decline in first-tier market valuation and a direct crash.
Comparatively, the survival pressure brought by sticking to non-financing has long surpassed the constraints imposed by financing.
Once famous for its thousands of graphics cards' computing power, DeepSeek now finds itself at a critical crossroads where it must reevaluate its financing options.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
