Compared to the loss-making strategies in the ToC and ToB markets, despite its flaws, the ToG market undoubtedly offers the fastest path to quick returns.

@TechInsights Original

Local state-owned enterprises are propping up half of the domestic big model commercialization landscape.

OpenAI shook up the AI landscape with ChatGPT in late 2022, setting a new benchmark. However, in 2023, domestic policy support for large language models was relatively insufficient, focusing more on guiding the overall development of the AI industry and computational power support.

It wasn't until the Cyberspace Administration of China and multiple ministries jointly issued the "Interim Measures for the Administration of Generative Artificial Intelligence Services" that targeted policies began to emerge nationwide, culminating in an explosion of activity this year. The enthusiasm of local governments has driven centrally-administered state-owned enterprises supported by local state-owned enterprises to quickly launch big model-related projects and open up market tenders.

Among the list of winning bidders, familiar names like Baidu, iFLYTEK, and Huawei stand out, and ZhipuAI, one of the six leading domestic big model players, is also a formidable competitor in terms of the number of winning bids.

While the technological breakthroughs and application capabilities of AI big models are impressive, their commercialization has been slower than expected. Whether Baidu, iFLYTEK, ZhipuAI, and other big model enterprises can reach their destination straightaway or take a circuitous route by leveraging the ToG market remains to be seen.

01.

Polarization in the Race

The pie offered by the big model ToG market is undeniably substantial.

Since 2023, procurement demand for big models has surged, igniting a boom in the tender market. According to public tender market data, there were 190 big model procurement tenders initiated throughout 2023, with a total procurement value of 595 million yuan; in the first half of this year alone, there were 498 related tenders with a value of 1.34 billion yuan, more than double that of last year. In August, the number of winning bids for big model projects hit a new high of 127, with 84 projects disclosing a total value of 390 million yuan.

In July this year, the State Council Information Office held a press conference on the theme of "Promoting High-Quality Development" and announced that central enterprises are expected to invest over 3 trillion yuan in large-scale equipment upgrades and replacements within the next five years. This includes deploying advanced equipment with high technology, efficiency, and reliability to accelerate digital transformation and integrate new technologies like AI into the entire manufacturing process.

This surge in project demand coincides with the commercialization challenges faced by the domestic big model market. While the 3 trillion yuan investment isn't solely focused on big models, the prospect of AI big model applications gradually expanding from operational scenarios to decision-making and management scenarios is promising. According to the president of the Forward Industry Research Institute, Xu Wenqiang, the domestic AI big model industry is expected to exceed 100 billion yuan in market size by 2028, with a compound annual growth rate exceeding 50% over the next five years.

Although central enterprises operate under market principles, they are more motivated to cultivate industries and create demand due to state-owned capital support. Among the publicly disclosed tenders, many involve the digital upgrading of traditional industries such as energy, telecommunications, finance, and education, with frequent appearances by central enterprises like China Mobile, China Tower, China Southern Power Grid, and China Energy Investment Corporation.

However, among the winning bidders, attitudes towards bidding for big model projects have diverged. According to incomplete statistics from the self-media outlet "Silicon Valley Star Pro," ZhipuAI led the pack in winning bids in the first half of the year, followed closely by Baidu and iFLYTEK. Beyond the top three, subsequent players like Huawei, SenseTime, Alibaba, and Tencent lagged significantly in terms of winning bids, while other big model players comparable to ZhipuAI were nowhere to be seen.

Based on information from other public sources, Baidu, iFLYTEK, and ZhipuAI are undoubtedly the three main players in the big model ToG market. In particular, iFLYTEK won 112 and 127 projects in July and August, respectively, with the latter month alone surpassing 150 million yuan in winning bid value.

An industry analyst told TechInsights, "Large companies have a natural advantage in the ToG market. Government departments and central enterprises attach great importance to the qualifications and track records of bidders. Many small and medium-sized enterprises, even if they have the ability to implement and execute, are often limited to becoming subcontractors for larger winning bidders."

Baidu and iFLYTEK have accumulated rich and mature ToG experience in their respective core businesses, which they can directly leverage for their big model products. ZhipuAI, on the other hand, has adhered to a ToB strategy since its inception, bidding for a wider range of project types with varying contract values ranging from tens of thousands to millions of yuan.

In this game of "Three Kingdoms" among the minority players, how many winners will ultimately emerge?

02.

Eating the Meat or Sipping the Soup

In the ToG business of AI big models, what changes are the products and services delivered, while what remains constant are the G-end customers and market characteristics.

Yi Feifan, a renowned figure in China's venture capital circle, once wrote that among the top 60 enterprise service companies in China, nearly 70% focus on the ToG market, with half of these companies having a gross margin below 50%, compared to over 70% for their American counterparts.

Most demanders in the ToG market are government departments or large central enterprises, offering substantial order sizes. However, complex implementation scenarios and strict risk control standards significantly raise execution costs, ultimately leaving slim margins for winning bidders. With the improvement of laws and regulations such as the Government Procurement Law and Tendering and Bidding Law, the procurement price system has become increasingly transparent, stabilizing profit margins for most tenders.

Furthermore, G-end customers prefer integrators who can provide one-stop solutions due to the convenience of delivery. These integrators then either fulfill the demand themselves or subcontract it to other suppliers.

As the bottom-tier small and medium-sized suppliers, while they must rely on integrators to secure contracts, their flexible business models allow for higher-quality revenue streams. Integrators, on the other hand, must not only manage the typically long payment terms from G-end customers but also ensure supplier delivery standards, relying on continuous order intake to expand their scale. However, it's not uncommon for integrators to pass on pressure to suppliers.

Therefore, internet giants like Baidu and Huawei have frequently emerged as leaders in the ToG market in recent years, primarily because they can simultaneously play the roles of integrator and supplier, ensuring a certain level of profit margin. Moreover, with diversified business layouts and sustainable ToC operations, their ToG businesses, even if unprofitable, can serve as bridges to strengthen government relations.

iFLYTEK has taken a unique approach, adhering to a dual strategy of ToG and ToC for many years. As co-founder Hu Yu explained, they use their core technology to solve national planning issues while also targeting the mass market with products like translation devices, which often puzzles outsiders.

However, despite Baidu and iFLYTEK's significant investments in big models and their leading positions in the ToG market, their financial contributions vary.

Baidu reported second-quarter revenue of 33.9 billion yuan and core operating profit of 5.6 billion yuan, up 23% year-on-year, primarily driven by the accelerated growth of its cloud business, which is also related to its big model initiatives. iFLYTEK's half-year financial report showed revenue of 9.325 billion yuan, up 18.91% year-on-year, but with a net loss of 401 million yuan, marking its first semi-annual loss since going public, partly due to investments in big model infrastructure.

Baidu founder Robin Li recently explained the disparity at a third-quarter director meeting, according to 36Kr. Li believes that ToB businesses must standardize their products, which often involve project-based work with diverse demands, requiring on-site personnel and backend development modifications.

While standardized products like Comate may not generate significant revenue or competitiveness at present, Li sees potential in their low starting point. Continuous investment can raise their thresholds and widen the gap with competitors, paving the way for future growth.

Li also stressed the importance of focusing on mid-sized customers, arguing that while large customers may not be highly profitable, very small or niche customers can be challenging due to limited budgets.

ZhipuAI, the only startup among the "Three Kingdoms" players in the big model ToG market, is still in the funding stage, having completed its 11th round of financing in September. Its frequent bids for ToG projects seem more like an extension of its ToB strategy rather than a sole reliance on large orders from central enterprises for financial recovery.

When the three main players may not be solely focused on winning the game, it becomes easier to understand the choices of other players who remain on the sidelines.

03.

Shortcuts and Pitfalls

Domestic entrepreneurs in the ToB sector often have mixed feelings about the ToG market. Some companies have soared to success by seizing the opportunities of government reform, while others struggle with the uncertainties of business relationships.

Jia Yangqing, former Chief AI Scientist at Alibaba, once stated that there are two dilemmas in the commercialization of big models: revenue flows differ from the past, and big models have a shorter window to generate revenue compared to traditional software.

Zheng Yu, Vice President of JD.com Group and President of JD City, likened the government to a "super B-end customer." The ToG market for big models inherently follows the rules of the ToB route, including the separation between product users and decision-makers.

A product manager with years of experience in the ToG industry told TechInsights that when designing products, their primary focus should be on interpreting policy documents to identify content that supports leadership performance, followed by considering the actual needs to be met after product delivery.

Furthermore, ToG projects involving digitalization, such as AI big models, often entail transforming traditional workflows. When these transformations affect organizational structures and job reassignments, internal project approvals can become a battleground between leaders from different departments.

This is similar to the profit model of traditional SaaS products. TechInsights learned from a SaaS entrepreneur that a digital project they once undertook could significantly reduce offline manpower. While the project smoothly progressed through initial demand refinement and demo development, it ultimately failed to secure final approval for implementation.

Unlike overseas SaaS companies that primarily serve small and medium-sized customers with highly standardized products, domestic SaaS companies rely heavily on revenue contributions from large customers, who often demand a high level of customization, indirectly increasing costs and lowering overall gross margins.

A typical example is the downsizing at ByteDance's collaborative office product, Feishu. Facing competition from Alibaba's DingTalk and Tencent's WeChat Work, Feishu attempted to create benchmark cases among large customers through a big client model, hoping to generate a demonstration effect. However, the growing team sizes required to meet the high customization demands of these customers hindered cost-effectiveness, ultimately leading to a one-third reduction in headcount from its peak and a temporary position as a market follower.

Beyond the unique subjective factors in ToG projects, big model enterprises must also confront the persistent cost challenges in the ToB market. Beyond government and central enterprise project-based orders, most B-end customers of AI big model products adopt an API call-based billing model, requiring continuous usage to ensure cash flow.

In the small and medium-sized customer market, large companies have unhesitatingly initiated price wars. As a representative of the startup camp, ZhipuAI officially entered the fray in June, reducing the call price of its entry-level product, GLM-3 Turbo model, from 5 yuan per million tokens to 1 yuan per million tokens, a staggering 80% reduction.

Compared to the loss-making strategies in the ToC and ToB markets, the ToG market, despite its flaws, offers big model enterprises the fastest path to quick returns. However, just as a momentary sip of fine wine may obscure the long-term path, businesses must avoid getting lost in the short-term gains and instead focus on laying a solid foundation for sustainable growth.