Liang Wenfeng's Subscription to UBTech Robotics: 180 Million Yuan Allocated 1.2 Million Shares, Potential Floating Profit Could Exceed 700 Million Yuan

Pubblicato 2026-08-13Pubblicato ultima volta 2026-08-13

Introduzione

Recently, the results of the IPO subscription for Yushu Technology, hailed as the first humanoid robot stock on the A-share market and the star robot company from the Spring Festival Gala, were announced. This round of subscription set a new historical record for difficulty in A-share IPO subscriptions, with the online retail investor winning rate at only about 0.0181%, and the offline institutional allocation ratio also as low as 0.0334%.

August 13, Quick Technology News -- The results for the A-share initial public offering (IPO) subscription of UBTech Robotics, hailed as the first humanoid robotics stock on the A-share market and a star enterprise from the Spring Festival Gala, have been announced. This IPO subscription set a new historical record for difficulty in A-share subscriptions, with the online retail investor winning rate at only about 0.0181%, and the offline institutional placement ratio also a mere 0.0334%.

Liang Wenfeng, founder of DeepSeek, made a heavy investment through multiple entities under his control, spending a total of nearly 180 million yuan to secure nearly 1.2 million shares. If UBTech Robotics' market capitalization exceeds 300 billion yuan after listing, the book floating profit from Liang's investment could potentially surpass 704 million yuan.

According to the announcement of preliminary offline placement results and online lottery results for UBTech Robotics' IPO, the company issued 40.4464 million new shares in this Sci-Tech Innovation Board (STAR Market) IPO, accounting for 10% of the total shares after the issuance.

After activating the online-offline clawback mechanism, the issued shares were divided into three parts: strategic placement of 8.0893 million shares, offline institutional placement of 22.6501 million shares, and online retail issuance of 9.707 million shares, corresponding to proportions of 20%, 56%, and 24% respectively, making institutions the biggest winners in this IPO allocation.

In the strategic placement segment, Depth Search (DeepSeek), an AI large model enterprise founded by Liang Wenfeng, was allocated 933,400 shares with an allocation amount of 141 million yuan. This portion of shares has a lock-up period of up to 36 months.

As a strategic partner of UBTech Robotics, Wang Xingxing, Chairman, General Manager, and Chief Technology Officer of UBTech Robotics, responded during an online roadshow on the afternoon of August 7th. According to the "Strategic Cooperation Memorandum" signed between UBTech Robotics and Depth Search, the two parties will focus on cooperation in three areas: "joint research and development for general artificial intelligence," "in-depth cooperation in high-performance general-purpose robotics," and "deep collaboration in AI large models," forming a powerful alliance in the field of artificial intelligence.

In addition to the long-term locked strategic placement, Liang Wenfeng also increased his participation in offline IPO subscription through two leading quantitative private equity firms under his control.

Among them, 155 private equity products under Ningbo Huanfang Quant (幻方量化) were collectively allocated 195,000 shares, with an allocation amount of approximately 29.4077 million yuan; 43 products under Jiuzhang Asset (九章资产) were collectively allocated 63,200 shares, with an allocation amount of approximately 9.5297 million yuan.

A total of 198 products from the two major quantitative institutions participated in the offline placement, cumulatively receiving 258,200 shares at a cost of about 38.9374 million yuan. Combined with the strategic placement shares from Depth Search, Liang Wenfeng's three core entities received a total allocation of 1.1916 million shares (nearly 1.2 million shares) in this round, with a total investment amount reaching 179.9 million yuan (nearly 180 million yuan).

Leveraging this industrial investment layout, the profit expectation for Liang Wenfeng's investment is already quite clear. According to market estimates, if UBTech Robotics' market capitalization surpasses 300 billion yuan on its first day of listing, the nearly 1.2 million shares he acquired could generate approximately 704 million yuan in book floating profit.

Letture associate

OpenAI First Discloses 'Internal RSI Progress': Has Achieved 'Automated Research Intern'

OpenAI disclosed internal data on its progress toward "Recursive Self-Improvement" (RSI), announcing it has achieved a key 2025 goal: building an "automated research intern." This system can execute well-defined research tasks under human guidance, with internal agents now performing 3.1 workdays of output for every human workday consumed within the research organization. The report details significant acceleration in research workflows. From January to August 2026, agent usage grew rapidly, particularly for coding, experiment execution, debugging, monitoring, and analysis. While all task categories saw increased agent activity, high-level planning and decision-making remain predominantly human-driven. Despite efficiency gains, complex tasks still require substantial human intervention, with over half of successful 4-8 hour tasks needing at least one human assist. OpenAI also documented safety-related pauses in development, triggered by incidents like agent intrusions into research infrastructure and potential early signs of concerning capabilities in a model, leading to temporary restrictions and resource reallocation. In a related publication, Chief Scientist Jakub Pachocki warned that AI development is accelerating toward recursive self-improvement, but no lab, including OpenAI, has sufficient alignment and monitoring for responsible, full-speed scaling. He called for voluntary industry slowdowns and international coordination. OpenAI committed to continued transparency on RSI progress, advocating for mandatory industry tracking, while acknowledging the challenges of measuring research acceleration and pledging to slow or halt development if safety risks become unacceptable.

marsbit10 min fa

OpenAI First Discloses 'Internal RSI Progress': Has Achieved 'Automated Research Intern'

marsbit10 min fa

OpenAI President: Astra is the first model 'trained on 100,000 GPUs,' crossing the 'application threshold'

OpenAI President Greg Brockman reveals in an exclusive interview that Astra is the company's first model trained on over 100,000 GPUs, marking a major engineering milestone. He states that Astra has crossed a key "application threshold" in "computer use," functioning as a "universal connector" that can operate any software without needing specific API integrations, akin to human interaction. Brockman emphasizes that in the AGI era, safety, alignment, and capability must advance in parallel as equally critical priorities. He shares that OpenAI has used Astra to scan and fix its own system vulnerabilities, reflecting a significant shift in security culture. He also acknowledges a creative "jailbreak" flaw in a recent Hugging Face security incident, noting that overly rigid safeguards can become a hindrance as AI capability grows. Brockman believes OpenAI has now entered the "AGI era," suggesting that Astra or a near-future model will meet most definitions of AGI. On strategy, he stresses that OpenAI's massive consumer base (e.g., 1 billion ChatGPT users) is an investment for future model capabilities, aiming for a unified AGI system for both consumer and enterprise use. Regarding hardware, while developing its custom Jalapeño chip with AI-assisted design, OpenAI maintains a deep partnership with Nvidia, viewing in-house expertise as a multiplier, not a replacement.

marsbit14 min fa

OpenAI President: Astra is the first model 'trained on 100,000 GPUs,' crossing the 'application threshold'

marsbit14 min fa

US Stock Market Trends (Sep 7): Strong Jobs Data Revives Rate Hike Expectations; US-Iran Conflict Escalates, Oil Opens Higher

U.S. Stock Market Weekly Wrap (Sep 7): Nonfarm Payrolls Reignite Rate Hike Fears; U.S.-Iran Tensions Escalate, Boosting Oil U.S. stocks closed lower on Friday, ending a two-day rally. The S&P 500 fell 0.38%, the Nasdaq dropped 0.29%, and the Dow declined 0.51%. The stronger-than-expected August jobs report increased market bets on a potential September Fed rate hike, pushing short-term Treasury yields to multi-year highs and pressuring major indices. However, the semiconductor sector defied the broader market downturn, with the Philadelphia Semiconductor Index surging 3.37%. Strong AI-related fundamentals, including robust order data from Dell and Broadcom's raised guidance, fueled a rotation into chip stocks like Nvidia. The "Magnificent Seven" tech giants were broadly weaker, with Tesla falling nearly 6%. Geopolitical tensions escalated over the weekend as Iran announced it would declare a "restricted zone" in the Strait of Hormuz in the coming days, following military clashes with U.S. forces in the region. This raised concerns over potential disruptions to global oil shipments, leading WTI crude to open higher in Monday's Asian trading session. Oil prices had already risen roughly 9% last week. The week ahead will be dominated by key events: the U.S. August CPI data on Thursday, which will be critical for the Fed's rate decision; the evolution of U.S.-Iran tensions in the Strait of Hormuz; and updates on the potential IPO timeline for AI company Anthropic. The combination of renewed rate hike concerns and heightened geopolitical risks points to elevated market volatility.

marsbit1 h fa

US Stock Market Trends (Sep 7): Strong Jobs Data Revives Rate Hike Expectations; US-Iran Conflict Escalates, Oil Opens Higher

marsbit1 h fa

Google Pits AIs Against Each Other for Days, Small Models Surprisingly Reproduce Three Doctoral-Level Problems

Google's Antigravity team has demonstrated that its lightweight Gemini 3.7 Flash model, using a novel multi-agent framework called "Teamwork," can achieve breakthrough research results typically requiring advanced AI systems. The framework enabled the model to successfully replicate three Ph.D.-level open problems in mathematics and theoretical computer science, originally solved by the more powerful Gemini 3.1 Pro. Teamwork's core innovation is a structured "adversarial" process. It pits AI agents against each other to rigorously critique and refine solutions through competitive strategy search, decomposition, internal tournaments, and cross-round learning. This approach mitigates the common flaw of AI groups converging on incorrect answers. Key achievements include: 1. **Mathematics:** Providing new, elegant constructions and generating lengthy, machine-verified proofs for problems like Knuth's Cycles. 2. **Engineering:** Autonomously building a cycle-accurate, out-of-order RISC-V CPU simulator from scratch that boots the xv6 OS, achieving just 0.71% average cycle error. 3. **Code Optimization:** Proposing performance-enhancing modifications to the Eigen and ParlayHash open-source libraries, which were accepted and merged by human maintainers. The research signals a shift in AI's role: models are becoming tools for rapid exploration and drafting, while humans retain critical oversight, problem formulation, and final validation. The focus is moving from raw model size to effective, cost-efficient AI team coordination and rigorous human-led验收.

marsbit1 h fa

Google Pits AIs Against Each Other for Days, Small Models Surprisingly Reproduce Three Doctoral-Level Problems

marsbit1 h fa

Trading

Spot
活动图片