Author: Wall Street Journal Bu Shuqing

Original Title: "Goldman Sachs In-Depth Report: Who Will Become the Long-Term Winner in China's AI Large Model Industry?"

China's AI large models are at a historic turning point. Goldman Sachs believes that the intelligence performance of China's open-source/open-weight large models has approached the top proprietary models globally, and the adoption scale among domestic companies and global small and medium enterprises is rapidly expanding, creating a data flywheel effect that will further drive model iteration and upgrades.

According to the Chasing Wind Trading Desk, Goldman Sachs' latest report points out that this evolution trajectory can be summarized as "from the cost-efficiency moment of DeepSeek last year, to the model intelligence moment of Zhicheng GLM this year." The team led by Goldman Sachs analyst Ronald Keung systematically evaluates four core issues in this 50-page report: how Chinese AI models achieve high performance at low cost, why they choose the open-source route and how to monetize, where the core addressable market is, and who will be the long-term winner.

In assessing the competitive landscape, Goldman Sachs introduced a "competitive positioning framework" based on pricing ability, cost advantages, and financial strength, determining that in the field of foundational text models, Zhicheng (first coverage) and DeepSeek (unlisted) are strongly positioned; in the multimodal field, ByteDance (unlisted) leads. Goldman Sachs also maintains a buy rating for MiniMax and Kuaishou.

Big Wins with Small Investments, Efficiency Prevails

China's large models can achieve performance close to that of similar products in the US at a significantly lower cost, primarily due to breakthroughs in architectural innovation and parameter efficiency.

Goldman Sachs' report points out that the parameter scale of China's open-source models generally ranges from 200 billion to 1.6 trillion, only 2% to 10% of the global top models, mainly due to restrictions in obtaining high-end computing power. Meanwhile, innovations such as the Mixture of Experts (MoE) architecture and sparse attention mechanisms result in the ratio of actual activated parameters to total parameters being only 3% to 5%, significantly lowering training and inference costs.

On the specific model level, DeepSeek V4 Pro has a parameter count of 1.6 trillion, Zhicheng GLM5.2 has 0.7 trillion, and MiniMax M3 has 0.4 trillion.

Goldman Sachs attributes the recent leap in programming capabilities of Chinese models to the synergistic effects of data filtering, reinforcement learning post-training, and other factors. On June 27, DeepSeek launched the speculative decoding framework DSpark, which has been deployed in the online services of V4-Flash and V4 Pro, enhancing user-generated speed by 60% to 85% (V4-Flash) and 57% to 78% (V4 Pro) without changing model weights or output quality.

Meituan's LongCat 2.0 release on June 30 is viewed by Goldman Sachs as an important milestone for the localization of China's AI infrastructure — it is China's first completely open-source MoE model trained and deployed based on 50,000 domestically produced computing power cards, with 1.6 trillion parameters. Goldman Sachs believes this demonstrates the viability of a localized hardware stack during the computationally intensive pre-training phase, which has profound significance for Chinese AI models to free themselves from dependence on foreign high-end chips.

Market Polarization, Stronger Stronger

Goldman Sachs describes the Chinese AI model market as forming a "two-tier structure" and identifies two quadrants for ARR maximization.

In the high-end market, top models represented by Zhicheng GLM5.2 and Alibaba Qwen3.7 Max are priced at about $1 per million tokens, five times that of low-end models, with gross margins of approximately 10% to 20% (according to Goldman Sachs estimates). In contrast, the pricing for top US models is $4 to $8 per million tokens, while high-end Chinese models are only 10% to 25% of that, but still maintain positive gross margins due to a lower parameter activation ratio.

In the low-end market, models aimed at agent tasks are priced as low as $0.06 to $0.2 per million tokens, expanding into the price-sensitive global small and medium enterprises and individual user market. MiniMax derives 60% to 70% of its revenue from overseas. Additionally, it is noteworthy that DeepSeek has announced a peak-valley pricing mechanism for the V4 series starting mid-July, where peak rates are twice that of non-peak times, with mixed pricing around $0.35 per million tokens (V4 Pro) and $0.12 (V4 Flash).

Goldman Sachs predicts that the API and subscription revenue for Chinese AI models will increase from an estimated RMB 35 billion in 2026 to RMB 879 billion by 2030, corresponding to a daily token consumption increase from 350 trillion to 46,000 trillion, an approximate 25-fold increase.

Open Source Strategy: Broad Penetration, Monetization Path to Upgrade

Goldman Sachs' report thoroughly reviews the strategic logic behind the widespread adoption of open-source/open-weight routes by Chinese AI models and their monetization limitations.

The core advantage of the open-source strategy is flexibility in deployment and community ecology. Models such as Alibaba's Qwen series, DeepSeek, Zhicheng GLM, and MiniMax M3 adopt open-source or open-weight methods, while ByteDance's Seed model is the main exception, following a fully closed-source proprietary route. The open-source model allows the model to be deployed flexibly both inside and outside mainland China and accelerates iteration through community feedback.

However, Goldman Sachs points out that the ARR figures disclosed by open-source model companies are likely to severely underestimate the actual deployment scale and revenue potential. For example, Zhicheng aims for an ARR target of $1 billion by the end of 2026, but the actual global deployment of GLM5.2 will far exceed the token volume and revenue from Zhicheng's own API channels — Alibaba Cloud’s Bailian MaaS platform can directly host the GLM5.2 open-source model without paying any fees to Zhicheng.

Goldman Sachs expects that the industry will gradually shift from purely open-source (MIT license, completely free) to "open weights + community license" mode — that is, commercial use must sign a revenue-sharing agreement with the model companies. The MiniMax M series has already adopted this model ahead of others. Goldman Sachs believes this change will significantly improve the unit economic benefits for AI model companies, as model companies can benefit from revenue-sharing agreements with platforms like AWS Bedrock, Alibaba Cloud Bailian, without bearing the computing power cost for inference themselves.

From "Token Maximization" to ROI Priority

Goldman Sachs characterizes international market expansion as the most important upward space for Chinese AI models, especially in non-US markets.

Goldman Sachs' US research team estimates that by 2030, agent AI will drive a 24-fold increase in global token consumption, reaching 120 million trillion tokens per month, with enterprise agents contributing a 55-fold increase and consumer agents contributing a 12-fold increase. In the global (outside China) market, Chinese AI models have achieved significant token share growth due to performance improvements and price advantages.

The Goldman Sachs report points out that the AI usage paradigm across global enterprises is undergoing a fundamental shift from "token maximization" to "ROI priority." The former prevailed from late 2025 to early 2026, with companies equating high token consumption with organizational productivity; the latter focuses more on clear task boundaries, daily active agent numbers, backend process automation, and actual output. A Jellyfish AI engineering trend study shows that heavy AI users in enterprises consume 10 times the tokens but only double their output.

On the channel level, Alphabet's Gemini Enterprise Agent Platform and Amazon's AWS Bedrock have both provided hosting services for Chinese AI models such as DeepSeek, MiniMax, Moonshot, GLM, and Qwen. According to the Wall Street Journal, Microsoft’s CEO recently stated that Microsoft is considering hosting a version of DeepSeek on Copilot as an optional low-cost model, emphasizing that if DeepSeek is hosted, that model will operate within Microsoft's cloud ecosystem, ensuring customer data remains within Azure.

Who Are the Long-Term Winners?

Goldman Sachs has constructed a three-dimensional competitive positioning framework to quantitatively assess the long-term winning probability of various players, with the core formula being: ARR scale × gross margin advantage + financial strength.

Pricing ability dimension examines the speed of listing (compared to previous generations and peer models), LMArena contest scores (based on large-scale blind user evaluations), and mixed pricing levels per million tokens.

Cost advantage dimension examines throughput (tokens per second), cache hit rates, parameter activation ratios, and inference gross margins. Financial strength dimension examines cash on hand, net cash as a percentage of total assets, and valuation multiples.

In the foundational text model field, Goldman Sachs recognizes Zhicheng (first coverage, neutral rating, target valuation of $110 billion) and DeepSeek (unlisted) as the strongest positioned, both performing excellently in terms of pricing ability and cost advantages. The overall implied valuation of independent AI model companies exceeds $200 billion.

In the multimodal/video generation field, ByteDance leads with Seedance, with reports from LatePost and 36Kr indicating that Seedance has a gross margin of up to 70%, and the ARR run rate has exceeded $2 billion. Kuaishou's Keling and MiniMax Hailuo/upcoming H3 model are also viewed favorably by Goldman Sachs, with expectations of benefiting from functional breakthroughs in video generation and LLM integration in the second half of 2026, and healthy pricing arising from supply tension.

Goldman Sachs maintains a buy rating on MiniMax, with a target price of HKD 860, reasoning that its M3 model is in the ARR maximization quadrant of high token volume and attractive pricing, and its current valuation is only 13 times the estimated ARR by the end of 2026, showing a clear discount compared to the valuation multiples of similar companies in China and globally, with the risk-reward ratio skewed upward.

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。