the number of large model projects won in january-august reached five times that of last year, with baidu leading the central state-owned enterprise market
2024-09-19
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
in 2024, as many large models obtained registration licenses, the commercialization process related to large models began to accelerate further.
therefore, we use the publicly available large-scale model-related winning projects as data to comprehensively track and analyze the trends and progress of large-scale model commercialization, and provide valuable market insights for related companies and practitioners.
from january to august 2024, the number of domestic large-scale model-related winning bids reached 475, which is about five times the number of large-scale model winning bids in 2023. the disclosed winning bid amount reached 1.535 billion yuan, which is about twice the disclosed amount of large-scale model winning bids in 2023.
with such a large market size, some large model manufacturers have gradually begun to take shape in terms of the number of winning bids, winning bid amounts, and industry coverage, demonstrating their leading position in the industry.
overall, baidu ranked first in comprehensive ranking, and won first place in three of the four dimensions that have attracted much attention in the market: number of winning projects, amount of winning projects, number of industries covered, and number of winning projects by central state-owned enterprises. iflytek and zhipu ai followed closely behind, and three major companies (tencent cloud, volcano engine, and alibaba cloud) were ready to go.
data on winning bids for large models from january to august 2024: the implementation of the large model industry is accelerating
entering 2024, china's large model market bidding market is very active. from january to august, a total of 476 large model winning projects were counted, and in the past eight months, the number of large model winning projects counted in a single month has set a new record. the number of winning projects represents the activity and business stability of companies in the industry, and best reflects whether an industry has the ability to develop sustainably.
in terms of the overall amount, although the monthly winning bid amount in the past 8 months has fluctuated, it has maintained a continuous upward trend.
from january to august 2024, the amount of winning bids disclosed by the projects counted reached 1.535 billion yuan, which is about twice the amount of winning bids disclosed for large-scale model projects in 2023 (789 million yuan). although september is only halfway through, preliminary data show that the number of winning bids in september should continue to set new records.
while the world is concerned about the commercial prospects of large models, chinese manufacturers seem to have found a blue ocean for the implementation of large models.
but it is undeniable that in this emerging new market, large model manufacturers are already very competitive. for example, in early september, guosen securities opened bids for the 2024 aigc investment advisory assistant project. the project amount was only 1.6 million yuan, but it attracted 7 manufacturers to bid. in some public projects, the bid price difference between bidders may reach 2-3 times.
winning bidders and competition situation: baidu ranks first in overall market performance
as of the end of july 2024, there are nearly 200 large-scale model manufacturers in china that have passed the registration. in such a fierce competition, a few large-scale model manufacturers have gradually emerged and become industry leaders.
after analyzing the large-scale model winning bid data from january to august 2024, we listed the number of winning projects, amount, industry coverage, and key customer distribution of 6 well-known domestic general large-scale model manufacturers.
through these data, we can see that from january to august 2024, baidu, iflytek, and zhipu ai have demonstrated their industry leadership, with the number of winning projects exceeding double digits. although the number and amount of winning projects of the three major manufacturers (tencent cloud, volcano engine, and alibaba cloud) are far behind the top three manufacturers, they can be seen in many large-scale model bidding projects in august and september, and the number of winning bids is also increasing rapidly.
baidu ranked first in overall performance, thanks to its first-mover advantage in the field of large models. but more importantly, baidu has accumulated experience in the field of ai for a long time and has a complete large-model product layout in the three major areas of ai computing power, model tools, and application development. it is precisely these technical accumulations and practices that allowed it to quickly break out in china when the large model exploded, and to lead the domestic large-model commercial market.
iflytek ranked second in overall performance, thanks to its winning bid for the gui'an new district computing industry cluster support project of east data west computing in august, which was worth more than 90 million yuan, which led to a rapid increase in the amount of its winning bid. iflytek has long focused on the b-end market as the focus of expanding the implementation of large models, and its nationwide business development system is also very mature, which has improved its implementation capabilities in the field of large models.
zhipu ai ranks third in overall performance, thanks to its willingness to actively participate in the bidding market. compared with several other star large-scale model startups, zhipu ai has invested the most in the bidding market, and of course the number of winning bids is also good.
currently, three major companies (tencent cloud, volcano engine, and alibaba cloud) are ready to go and are actively participating in the bidding projects for large models. as these major companies make efforts, the future market structure should see significant changes.
central soes become the focus of competition
where will the future large model landing market be? where will the large model manufacturers compete for the market? if readers carefully observe our industry statistics from january to august, they can actually see that the central state-owned enterprise market is a must-fight place for manufacturers in the future. central state-owned enterprises are the pillar enterprises of the national economy and play a leading role in some key fields and industries.
from january to august 2024, central state-owned enterprises won 295 large-scale model projects with a total bid amount of 963 million yuan, accounting for 62% of the overall market and 63% of the amount.
the rapid growth of large-scale model bidding projects initiated by central state-owned enterprises has a lot to do with the fact that enterprises in these fields closely follow policy guidance and accelerate the layout of new quality productivity. of course, central state-owned enterprises with a high degree of marketization, such as finance, also have their own business innovations and demands for improving efficiency and reducing costs.
specifically, baidu currently has a prominent leading advantage in the central state-owned enterprise market, with 25 of its 30 winning bids being central state-owned enterprise projects, including leading central enterprises such as state grid, china southern power grid, china tower, china resources, and china iron and steel research institute, as well as financial clients such as postal savings bank of china, agricultural development bank of china, and taiping financial. baidu's leading position in the central state-owned enterprise market also proves that it has won customer recognition for its full-stack ai capabilities such as ai computing power management, ai platform, large models, and intelligent agent applications.
among the projects won by iflytek and zhipu ai, there are also many central state-owned enterprise clients, but the industry coverage is relatively narrow. iflytek's representative clients include china mobile, china coal, zhejiang tobacco, etc.; zhipu ai's representative clients include china mobile, postal savings bank, china general nuclear power, etc.
industry distribution of large model projects: ten major industries are targeted
in the "china large model winning bid project monitoring report", we usually analyze the industry distribution of large model winning bid projects. from january to august, we roughly divided the bidders into 15 major industry categories according to their business attributes. of course, this division is not strictly consistent with the industrial and commercial registration information.
judging from the data from the first eight months of 2024, the number of big model winning projects is growing, and the industry coverage is also increasing rapidly. from the beginning of the year, the projects were mainly in the energy and government affairs industries, but now the winning projects have begun to cover more than 10 mainstream industries such as education and scientific research, communications, energy, technical services, finance, government affairs, media, transportation, commercial services, tobacco, medical care, and manufacturing. this proves that big models are indeed being rapidly implemented in all walks of life.
among these industries, education and scientific research, technical services, finance, and media have been the fastest growing industries in the past two to three months. the rapid growth of the education and scientific research industry has a lot to do with the increased research, exploration, and application of big models by education and scientific research institutions; the rapid growth of the financial industry is not surprising either, as more financial institutions hope to use big models to improve the efficiency and quality of financial services.
some of baidu's large model winning projects (january-august 2024, public data)
some of iflytek's large model winning projects (january-august 2024, public data)
some of zhipu ai's large model winning projects (january-august 2024, public data)
in terms of manufacturers, baidu's big model winning projects have the widest industry coverage. baidu covers all the 10 mainstream industries mentioned above, especially central state-owned enterprises, finance, and education and science customers. iflytek's winning projects cover 8 industries, mainly finance and central state-owned enterprises. zhipu ai covers 8 industries, mainly finance and central state-owned enterprises.
in the future big model market competition, we tend to believe that it will be the big technology companies that will dominate the general big model field, and more ai startups will focus on building industry models.
the logic behind this is also very simple. as the training cost becomes higher and higher, the general large models will most likely gradually converge to large companies with abundant financial resources, deep technical accumulation, and abundant talents, which may further affect the commercialization of large models in the future. and compared with ai startups, large companies are more likely to gain customers' trust in the sustainability of large model services.
the big model will eventually create value only when it reaches customers. from the current perspective, central state-owned enterprise customers play an important role in the future market structure. whoever wins more central state-owned enterprise customers may truly benefit from this wave of technology.
of course, it seems inappropriate to talk about competition in the early stages of market development, but it is precisely this fierce competition that drives the commercialization of large models forward and breeds various innovations in the process.
note: the winning bid amounts of some projects were not disclosed. for the convenience of statistics, the project amounts were calculated as 0. in addition, there may be statistical omissions, and the project classification and industry classification are not strict, which may affect the final statistical analysis results. readers are advised to be aware of this.
