From 17 to 20 july 2026, 2026 world artificial intelligence (waic) will be held in shanghai. This session will be about world models, open sourcesSmartThere will be more than 300 products globally。
The national letter security analysis states that the current artificial intelligence technology cycle is in the second half of the cycle, i. ELarge modelThe current stage of centralized change in development and the application of smarts. The chinese securities indicate that in 2026 the ai application will usher in a “metanal year of gold”, that technology has matured and that a number of large models are already equipped with powerful tools for mobilizing and multimodular understanding。
Global ai innovation focus
It is understood that the world congress of artificial intelligence, 2026, with the theme “intellectual partners for the future”, has a total surface area of more than 100,000 square metres, with more than 1,100 companies participating in the exhibition, and more than 3,000 exhibits in focus, with over 300 products starting globally. There are over 200 companies in each of the two main tracks, with a focus on the cutting edge of artificial intelligence empowerment。
For the first time, the congress launched a high-level international academic conference, “waic academic”, which was chaired by toolin lau zhou zhi, and led by two faculty members to form an academic committee covering a wide range of countries and regions around the world, working to create a top-level academic exchange platform in the area of artificial intelligence. Nine turing and nobel laureates have been preliminarily identified。
The conference forum will focus on ai hotspots, and will explore new trends in industry development around issues such as world models, open-source smarts, ai coding, token economy and opc。
In the exhibition, the 950 hypernodes (atlas 950 superpod) will be presented for the first time in the true form of the largest ultranodes in industry. It's a sign of a massive surgeCountNew steps in the area of infrastructure provide a solid calculus support for the age of large models。
In addition, new items such as minimax m3 megamodels, step agent operating systems, the world's first ai smartphone, and multiple human robots, ai smart hand and others will be displayed at the general assembly。
Big model speeds upCommercializationThe continuous release of the full-chain computing needs
At present, the development of our artificial intelligence industry is strong. The data show that the number of artificial intelligence enterprises in our country exceeded 6,000 in 2025, with core industries exceeding 1. 2 trillion yuan. The applications have covered key industries such as steel, colour, electricity and communications。
By the end of 2025, the country was using more than 1. 73 million standard hangers in computing facilities, with 42 of the huanca chile calculator cluster, with an intelligent calculator size of 1. 59 million pflops (fp16), ranked second in the world。
In its research, widely issued securities states that as model capabilities continue to increase and reasoning costs decline rapidly, large models are accelerating towards the commercial landing phase of scale. From a cost perspective, it has benefited from enhanced hardware performance, optimization of model structures and improvements in systems engineering, continuous downscaling of reasoning costs and further opening of token quantities. According to the agency, capacity enhancement and cost reduction form a positive cycle and are expected to accelerate the scaling-up of agent, ai coding and enterprise process-type applications。
China & china (603296. Sh) has a clear all-branch technology advantage and is a very small number of producers in the industry with both computing and network nodes design capabilities that can work in depth with head csp manufacturers. In a recent study, chinese technology suggested that, as the demand for large model reasoning expands, csp manufacturers need to reduce the cost of token, while model path diversification creates a need for model and hardware matching, and therefore, represented by hypernodes, there will be different combinations of power clusters; and there are expected to be significant changes in gpu, storage, etc. And cpu requirements in response to agent's needs. It believes that the above-mentioned changes will provide a good opportunity for producers with strong systems integration capabilities, and that the company has in recent years been reserveing technological, production and market capabilities for overnodes. After practical testing and validation by the csp manufacturer, the delivery of the ultra-node will begin to be measured this year。
The extension information (002261. Sz) continues to deepen the intelligent computing layout, based on the 鲲鹏 + 腾 腾 腾 腾 算 算 算 算 算 算 算 算 算 算 算 算 算 算 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 00 00 。 。 。 。 。 。 。 。 。 。 。 00 00 00 00 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 on the centre side, the company launched the mega-rh220t-c high-performance server series, the mega-ra2300-c series reasoning server; on the side of the ai application, the company launched the mega-deepseek as one; on the side side, the company launched the mega-station x5 ai work station based on china for the products of the knpc/apn module; and on the end side, the company built the mega-l320 series of notebooks and the mega-d585 series。
The tianyang technology (300872. Sz) has recently worked with kai-han to get into the calculator lease track and actively nurture the second growth curve. In its research activities, tianyang technology revealed that the company, in collaboration with kai-han, provided computing leasing services to end-users and was expected to be fully tested by the third quarter. Prior to this, the company had been working online with the capital on the sun lake calculator park project, working with the national super calculator changsha centre to provide calculator services to clients, and had been investing in the field of computing in the past. Enterprises with the ability to compute services will also continue to benefit against the backdrop of rapid and iterative ai technology, commercialization of large models, and sustained surges in power demand。
