Wealth Management Companies Addicted to 'New Share Subscriptions'? 53 Products Chase YuShu Technology, 23 Make Repeat Appearances, New Play for 'Fixed Income+' Emerges

Опубліковано о 2026-08-13Востаннє оновлено о 2026-08-13

Анотація

Unitree Tech announced its offline placement results, revealing a further expansion in the participation of wealth management firms in offline IPO subscriptions. A total of 53 products from 6 wealth management companies – CMB Wealth Management, Ningbo Bank Wealth Management, China Merchants Bank Wealth Management, China Post Wealth Management, China Minsheng Bank Wealth Management, and Bank of Nanjing Wealth Management – were initially allocated shares. They collectively received approximately 140,100 shares, amounting to 21.1277 million yuan.

On August 12th, YuShu Technology announced its offline allotment results, revealing an expanded lineup of wealth management companies participating in offline new share subscriptions:

A total of 53 products from six wealth management companies—Everbright Wealth Management, Ningbo Bank Wealth Management, China Merchants Bank Wealth Management, Postal Savings Bank of China Wealth Management, China Minsheng Bank Wealth Management, and Bank of Nanjing Wealth Management—were granted preliminary allotments, collectively receiving approximately 140,100 shares amounting to 21.1277 million yuan;

Compared with the previous ChangXin Technology allotment, the number of participating wealth management products increased by 24, with 23 products participating in both allotments consecutively. This indicates that offline issuance business has shifted from sporadic attempts to regular configuration within the product pools of some wealth management companies.

In terms of the number of products receiving allotments, wealth management subsidiaries of city commercial banks and joint-stock banks were the main participants in this round:

Ningbo Bank Wealth Management had 24 products allotted, Everbright Wealth Management had 21, with the two together accounting for over 80%; China Merchants Bank Wealth Management had 4, Postal Savings Bank of China Wealth Management had 2, and China Minsheng Bank Wealth Management and Bank of Nanjing Wealth Management each had 1.

Currently, products participating in new share subscriptions commonly adopt a portfolio strategy of "fixed income foundation + equity holdings + offline new share subscriptions." This involves building a foundational position with fixed-income assets like bonds to obtain stable coupon income, while holding a certain scale of stock positions to meet the market value threshold for offline inquiry and subscription, and then overlaying the price difference boost brought by offline new share allotments.

In an environment of low interest rates and narrowing yield space for pure bond strategies, this model is seen by some wealth management companies as a feasible path for "fixed income+" products to enhance returns.

However, the boost from new share subscriptions is not a guaranteed profit.

Statistics show that some products participating in new share subscriptions still recorded negative returns over the past three months or six months. This means that if equity holdings experienced drawdowns during this period due to stock market fluctuations, the resulting losses could offset or even exceed the net value boost contributed by the subscriptions. The final product return still depends on the combined performance of the fixed income, equity, and new share subscription components, not on the subscription segment alone.

Taking the YuShu Technology allotment as an example, even if the new stock's post-listing gain is high, the net value contribution to a single product from this subscription is likely only a few tenths of a percent, due to the relatively tiny allotment size and low allotment rate compared to the product's total assets under management;

In other words, a single subscription has limited pull on a product's net value. The actual value of the subscription strategy relies more on "sustained, multi-project accumulation"

. By frequently participating in multiple new share allotments, the scattered contributions of a few tenths of a percent can gradually accumulate to form an observable boost in returns.

This is also why 23 products participated in two consecutive allotments rather than just dabbling.

According to current rules, after the implementation of the underwriting new regulations in March 2025, bank wealth management products have been included as Class A priority allotment objects, enjoying the same priority allotment rights as public funds and social security funds. This provides the institutional foundation for wealth management companies to participate in offline new share subscriptions on a large scale;

However, the equity holdings themselves are affected by stock market fluctuations, introducing uncertainty into the effectiveness of the subscription strategy. When holdings draw down, the subscription boost may be eroded.

Up to the YuShu Technology allotment, both the number of participating products and the number of wealth management institutions involved in offline subscriptions have increased significantly compared to the first allotment, but single-allotment sizes are limited and return contributions are dispersed;

Subsequently, it will be necessary to observe the actual magnitude of net value boost that subscriptions provide to "fixed income+" products after more new share projects materialize, as well as the overall return stability amid fluctuations in equity holdings.

Пов'язані матеріали

The Era of Large Model Distillation is Over: Fable 5.1 Rewrites API, Cutting Off the Path of Distillation for Good

The era of large model distillation is ending. On September 2nd, Anthropic delivered a decisive blow by updating its API rules with Claude Fable 5.1, effectively cutting off the path for shell models and distillers. Previously, companies bypassed immense compute costs and lengthy training times by using API calls to extract the reasoning process of top-tier models like Claude, then using this data to train their own smaller "distilled" models. A key vulnerability was the "thinking blocks"—the model's internal Chain-of-Thought reasoning steps returned via API. Distillers exploited this by modifying the surrounding context (like system prompts or earlier messages) in multi-turn conversations, tricking Claude into revealing its hidden underlying logic. Fable 5.1 introduces a stringent "context consistency verification" mechanism. The API now strictly validates that the "thinking blocks" sent back by the client match the original system prompts, tools, and message history that produced them. Any modification causes the API to return an error. A "non-strict mode" is offered for legitimate developers who need to modify context (e.g., for compression), but it silently deletes all thinking blocks, forcing the model to answer without its prior reasoning. This crackdown was deemed necessary due to industrial-scale abuse. "Distillation hackers" used thousands of fake accounts and automated scripts to exploit the API, extracting high-intelligence reasoning capabilities while completely bypassing the costly safety and alignment training (like RLHF) built into models like Claude. This created a critical risk: "capability-safety decoupling," where distilled models gain advanced abilities but lack the ethical guardrails, potentially making them dangerous. The new rules are being rolled out in phases, initially targeting new API accounts created after August 31, 2026, UTC. Existing API accounts and consumer users (e.g., Claude.ai) are unaffected for now, giving legitimate developers time to adapt. Anthropic states the "thinking retention" mechanism will eventually apply to all accounts. An unexpected benefit for compliant developers is potential cost reduction and speed improvements. Enforcing context consistency allows for highly efficient prompt caching on API servers, slashing latency and compute overhead. This move marks a watershed for the AI industry, challenging the narrative of small models outperforming large ones through distillation and forcing a reckoning on innovation versus imitation.

marsbit7 хв тому

The Era of Large Model Distillation is Over: Fable 5.1 Rewrites API, Cutting Off the Path of Distillation for Good

marsbit7 хв тому

South Korea Announces Securities Tokenization Timeline: First Batch of Tokenized Assets to Include Bonds, Funds, and Unlisted Stocks

South Korea's Financial Services Commission (FSC) has unveiled a three-phase roadmap for tokenizing securities, positioning itself as a potential first-mover with dedicated legislation. Following the formal enactment of amended laws in February 2027, Phase 1 will begin with tokenized private market assets, including bonds, institutional money market funds, and unlisted stocks (via trust beneficiary certificates). Existing licensed securities firms can operate without new permits, with specific rules for non-financial platform operators and investor limits. Phase 2 will expand to publicly issued securities, contingent on the stability of initial systems and market readiness. The final Phase 3 aims to enable on-chain settlement using stablecoins, pending separate stablecoin legislation. The announcement contrasts sharply with rapid, decentralized approaches like Robinhood's recent tokenization of stocks, which sparked controversy. South Korea's path prioritizes legal clarity and infrastructure, starting with controlled, institutional markets before broadening access. While this methodical approach may sacrifice speed, it seeks to establish a clear regulatory foundation. The global race for tokenization is highlighting divergent strategies between regulated, incremental models and faster, more open but less certain alternatives.

marsbit10 хв тому

South Korea Announces Securities Tokenization Timeline: First Batch of Tokenized Assets to Include Bonds, Funds, and Unlisted Stocks

marsbit10 хв тому

Just Now, Claude Proves Fermat's Last Theorem for the First Time, Led by Tsinghua Yao Class Prodigy

In a groundbreaking development, Claude has autonomously generated the first machine-verified proof of Fermat's Last Theorem in just 11 days. The project was led by Tianyi Peng, a researcher from Anthropic with a background from Tsinghua University's prestigious Yao Class. This achievement required Claude to write 13 million lines of Lean code, proving over 29,500 intermediate theorems and consuming 60 billion tokens—a volume exceeding the largest existing mathematical theorem library by fivefold. The process involved formalizing the 350-year-old theorem, which states that no three positive integers a, b, c satisfy a^n + b^n = c^n for any integer n > 2. While Andrew Wiles provided a human proof in 1995, its complexity made verification a years-long task for experts. Claude's formal proof builds from foundational axioms, autonomously constructing the entire logical chain and verifying it through the Lean compiler. Key to the success was the development of the "Prove2Me" platform, which managed dozens of Claude agents by organizing tasks into a theorem DAG (directed acyclic graph), separating statements from proofs, and maintaining natural language indexes. This addressed early collaboration inefficiencies and "hallucination" issues. The result is the largest Lean proof ever created. The accomplishment has stirred significant discussion in the mathematical community, highlighting AI's potential to automate the formalization and verification of complex proofs. While not replacing mathematicians, such technology could fundamentally change mathematical practice by providing absolute verification, checking human-generated mathematics, and enabling broader access to formal verification tools.

marsbit10 хв тому

Just Now, Claude Proves Fermat's Last Theorem for the First Time, Led by Tsinghua Yao Class Prodigy

marsbit10 хв тому

Raoul Pal: Why Has the Traditional Investment Portfolio Become Obsolete?

Raoul Pal argues that traditional investment portfolios (bonds, gold, real estate, index funds) are no longer effective for building real wealth. He posits that due to persistent currency devaluation from money printing (global liquidity expanding ~8% annually plus regular inflation), an investor needs an 11% annual return just to preserve purchasing power. He evaluates traditional assets: bonds fail as interest doesn't cover currency devaluation; real estate's historic wealth-creation window from falling rates is over; gold preserves purchasing power but doesn't create new wealth; and the S&P 500 barely meets the 11% threshold, relying on a historic bull market. The only assets consistently exceeding this benchmark, based on decade-long data, are technology stocks (NASDAQ 100: ~20% annualized) and crypto assets (Bitcoin: 58-70% annualized). Their outperformance stems from user adoption S-curves (Metcalfe's Law), not speculation. Pal explains that post-2008, traditional diversification lost its protective power because bonds, gold, real estate, and stocks are all now primarily driven by the same macro factor: liquidity. Thus, a diversified portfolio of underperforming assets offers false security. For crypto, he favors underlying protocols/L1 blockchains over applications, as they capture value from the entire ecosystem. A key, underappreciated future driver is AI agents, which will require programmable money and 24/7 settlement, a natural fit for blockchain. Key investment principles include: avoid leverage (it removes the ability to weather severe drawdowns), allocate a meaningful portion (not all) of capital to high-growth assets, and practice patience—"doing nothing" is a valid long-term strategy. The core opportunity cost is freedom. Returns below 11% annually mean your labor buys less freedom over time. The goal is to use this framework to audit your holdings, moving capital from assets that erode purchasing power to those with genuine compounding potential.

marsbit1 год тому

Raoul Pal: Why Has the Traditional Investment Portfolio Become Obsolete?

marsbit1 год тому

Торгівля

Спот
活动图片