AI has been the biggest variable in the cloud computing market in the past few years, and the maturity of the large model will undoubtedly enhance this variable exponentially.
I remember that at the end of 2022, when generative AI and large models began to explode, we discussed a problem: in this wave of AI, the cloud computing market will be the first to be profoundly affected.
This is because of a series of characteristics such as large-scale data, high training costs, and frequent updates of large models, which inevitably lead to the best application method to access the user interface based on the public cloud. As the large model takes a firm foothold in the process of industrial intelligence and plays an increasingly important role, the value of the public cloud platform as the source of the large model will also be greatly enhanced.
At present, this judgment is gradually becoming a reality. The specific performance is that the growth of AI public cloud is accelerating; large-scale models are beginning to drive a new round of cloud computing industry transformation; cloud computing vendors with large-scale model technology and application advantages are continuing to expand their leading positions.
(Graph from IDC Consulting)
IDC recently released the "IDC China AI Public Cloud Service Market Share 2022" report. According to the report, the growth rate of China's AI public cloud service market will reach 80.6% in 2022. Among them, Baidu Smart Cloud accounted for the first market share, with a growth rate of 69.7%. This is also the fourth consecutive year and the eighth time that Baidu Smart Cloud has ranked first in this market.
(Graph from IDC Consulting)
We know that machine vision CV and natural language processing NLP are the two core pillars of AI technology. In the AI public cloud map, Baidu Smart Cloud also ranks first in the two major markets of CV and NLP. With the remarkable success of the Wenxin large model, there is reason to believe that Baidu Smart Cloud will make greater progress in the AI public cloud market in 2023.
(Graph from IDC Consulting)
The IDC report is important not only because it summarizes the previous market structure, but also because it shows the future very clearly and firmly.
why would you say so? What will the big model bring to the cloud computing market? Baidu Smart Cloud's success in the AI public cloud market, what reference significance can it bring to cloud computing manufacturers, users, related industry chains and investors?
Let us take these questions into account and understand the coming new era of "big model + cloud" through the IDC report.
AI public cloud
Cloud computing returns to the main channel of high growth
If you want to predict the changes that large models will bring to cloud computing, you must first understand the role played by the AI public cloud.
Overall, the public cloud is currently in a development cycle where development is slowing down and industry pressure is increasing. According to data, the growth rate of China's public cloud in 2022 will only be 35.4%, a year-on-year decrease of 7.8% in 2021. The industry generally faces a series of problems such as slowdown in growth rate, high pressure on profitability, and lack of industrial fulcrum.
Under such circumstances, finding a sustainable and stable high-growth window has become the most important task of the cloud computing industry. After years of development, the AI public cloud has proved its feasibility as a new fulcrum of the cloud computing industry.
As mentioned above, according to the IDC data report, the growth rate of China's AI public cloud service market will reach 80.6% in 2022, compared with the entire public cloud territory. Such a growth rate is very rare. More importantly, the AI public cloud will not only exist as a short-term outlet.
From a technical point of view, cloud vendors have the best technical base for generative AI and large models. These technical capabilities in their infancy have received strong attention from public opinion, policy guidance, investment and construction, etc., and will bring great development momentum and room for cloud computing vendors. From a commercial point of view, training AI models, deploying AI models, and obtaining AI solutions and comprehensive services all need to rely on the public cloud model, which will greatly enrich the commercialization path of public clouds, so that cloud computing manufacturers have more ways Solve the most critical problem of profitability. It can be said that with the increasing role of intelligence in the industry, the upgrade of AI public cloud is irreversible.
IDC believes that in the next five years, the next generation of artificial intelligence driven by large models and generative AI is expected to bring the overall industry back to the era of high growth.
Under this trend, the AI public cloud will gradually become the core of the public cloud market by virtue of the value of the large model, and then evolve into the main channel for cloud computing to return to the high-growth stage. Baidu Smart Cloud, which has won the championship in this field for many years, It will also be the first to obtain the opportunity to start sustained rapid growth.
Under such a trend, the next explosive point of the AI public cloud is the large model. How to create large-scale model capabilities and implementation solutions that adapt to user needs will become the next main point of competition for cloud computing manufacturers.
Large base, wide application
Large-scale model + landing solution in the cloud era
The large model is the key to the AI public cloud and even the overall transformation of the cloud computing industry. We need to note that the IDC statistics quoted above are from 2022, and this year’s situation is that the demand for large-scale model applications has increased sharply, and the understanding and recognition of large-scale models by cloud users in various industries has rapidly increased, which will inevitably lead to large-scale models. The advantages of cloud computing vendors have been further expanded.
During the 2023 World Artificial Intelligence Conference (WAIC) held not long ago, the National Artificial Intelligence Standardization General Group under the guidance of the National Standards Committee officially announced the launch of the formulation of national standards for large-scale model testing, and technology companies such as Baidu entered the large-scale model "national team" .
From the promotion of national standardization work and the latest trends in the AI industry, it is not difficult to find that the large model is rapidly developing in a new direction of standardization, base, and strong application. From the point of view of the connection with the cloud computing market, obtaining large models through public clouds is more conducive to the realization of large model standardization. Larger models use more clouds, and the services and support that need to be obtained on the cloud are more complex, and the business models such as MaaS that have evolved are more diverse. Therefore, the development of large models will greatly stimulate the rapid improvement of the cloud computing industry. Large model + cloud will become the main way to realize industrial intelligence in the future and the core value source of the cloud computing market.
In the new stage of big model + cloud, the logic of "the stronger the big model is, the stronger the cloud" will become more and more obvious. Due to the high cost of training, parameter adjustment, and continuous upgrading of large models, the meaning of users to repeatedly create large models will become less and less meaningful. Therefore, the method of using a mature large model, combined with data fine-tuning and formulation of an industry implementation plan, will become the main implementation solution of large model + cloud. The IDC report also believes that in the future, enterprises will directly use the existing general-purpose large models in the market.
That is to say, the era of large model + cloud will soon show a small number of large model bases, with a wide range of application directions and a structure of landing channels for diverse industries. In this case, the general-purpose large-scale model technology used as the base must stand the test.
For example, the Baidu Wenxin large model has been upgraded to version 3.5. Compared with version 3.0 in March, the training speed has increased by 2 times, and the inference speed has increased by 30 times. A number of public authoritative evaluations show that Wenxin Yiyan's comprehensive ability evaluation score supported by Wenxin Large Model 3.5 has surpassed ChatGPT, far ahead of other large models, and some Chinese abilities even surpassed GPT-4. The era of large model + cloud is the era of large base + wide application. How to make the large model base build a bridge to the application will also become the most important capability in the new stage of cloud computing.
Two horizontal and vertical slices
Determine the future pattern
The large-scale model will promote the exponential development of the AI public cloud, and cloud vendors want to compete for the key seat of the industrial intelligence base in the next five years, a key change cycle, mainly relying on two aspects of capabilities.
We can understand the AI capabilities that determine future changes in the public cloud market as two aspects, one horizontal and one vertical. The horizontal ability is to pull through the technical foundation and build a large-scale model ability with a full score experience. Vertically, it is necessary to truly open the way for large-scale models and AI technology to enter the industry, so as to ensure that large-scale models are not suspended outside the industry.
We can see that Baidu Smart Cloud has been able to win the AI public cloud market champion for four consecutive years, and insisting on the simultaneous development of both horizontal and vertical aspects is the key to its success. According to the "AI Large-scale Model Technical Capability Evaluation Report, 2023" released by IDC, Baidu Wenxin Large-scale Model 3.5 scored 7 full marks in 12 core indicators, and its comprehensive score ranked first among domestic mainstream large-scale models. Among them, the two major indicators of "algorithm model" and "industry coverage" are the most important indicators to measure the technical depth and industry coverage of large models. Baidu Wenxin large model is the only large model that has obtained full marks in these two key indicators.
The advantages of the Wenxin large model can also be fully aligned with the advantages of Baidu Smart Cloud. First of all, from a horizontal perspective, Baidu Smart Cloud integrates Baidu's full-stack technology advantages in the AI field, realizes the integration of advantages from self-developed chips, self-developed frameworks to AI solutions, and continuously breaks through at the level of AI algorithms, accumulating its own advantages Advantages of software technology.
From a vertical perspective, large models need to face complex scenarios and needs when entering the industry. Through continuous case building and industry exploration, Baidu Smart Cloud continues to shape its own industry intelligence capabilities from multiple aspects such as common platforms and industry services.
For example, the machine tool industry is a representative of heavy industry and has always been considered a typical traditional industry, but Baidu Smart Cloud has brought AI capabilities into such fields. In Jinan No. 2 Machine Tool, by applying the AI intelligent scheduling optimization engine of Baidu Smart Cloud, its scheduling ability has been greatly improved, and the previous challenges of high scheduling and time-consuming scheduling in the scheduling link have been solved, and the scheduling situation has been improved. Significant improvement.
In many fields such as industry, energy, finance, and government affairs, Baidu Smart Cloud has brought the capabilities of AI and large models to a height far exceeding external expectations. In order to meet the AI needs of industrial users, Baidu also launched the Wenxin Qianfan large model platform, which is a full-process tool chain for large models, including data management, model training, evaluation optimization, forecasting service deployment, and plug-in services, which can effectively help the industry The high-efficiency and low-cost application of users to Wenxin series large models.
At present, Baidu has cooperated with State Grid, Shanghai Pudong Development Bank, Taikang, Geely and other enterprises to release 11 large-scale industry models, realizing the largest large-scale industrial scale in China.
Stabilize the core technology of large-scale models horizontally, and open up the boundary between large-scale models and industries vertically, which constitute the two major factors for the success of large-scale model + cloud. It is precisely because of this that Baidu Smart Cloud, which is grasping and strong with both hands, took the lead in launching a new picture of large model + cloud.
To sum up, the cloud computing market urgently needs a new development engine, and the explosion and continuous development of large models can just make AI public cloud become such an engine. In such a transformation of cloud computing, AI basic technology and AI industry capabilities will determine the future structure of the cloud computing market.
The deterministic future of cloud computing occurs at the intersection of industry and big models.
