Globale Freigabe von Rechenleistungskapazitäten: DeepYong Intelligence liefert die "Munition" für KI - Unternehmen.

The competition for large models has already reached its third year. The number of model parameters has expanded from billions to trillions, and the demand for computing power for training and inference has multiplied. Where there is demand, there is also a market, and that means new entrepreneurial opportunities.

Computing power accounts for a large proportion of the development costs of large models. In the research and development process, GPU and other computing resources in practical applications are often left unused due to insufficient planning or inefficient distribution. Resource utilization generally lies below 50%. On the other hand, many companies choose a hybrid - cloud or multi - cloud strategy, but the resource management methods of different cloud providers vary, and unified planning and optimization are extremely challenging.

These problems have directly created an urgent need for an efficient computing power economy management platform, which provides entrepreneurs with an entry point and has produced numerous promising AI innovation projects.

At the recently held second season of the "Lenovo New Business Innovation Ecosystem Roadshow", the AI computing power startup DeepFlow Intelligence was awarded as one of the TOP 10 projects and has become a new superstar in the Lenovo ecosystem. This is an acknowledgment of the technological strength of DeepFlow Intelligence and also the consistent support of the Lenovo Small and Medium Business Group for startups: providing comprehensive, high - quality and holistic services to solve the problems in enterprise development. The Lenovo Small and Medium Business Group will advance together with customers and accompany customers to success.

As the central instrument for Lenovo Group to leverage its leading role and drive the integrated innovation of startups, the Lenovo Star Program for Startups has built a three - dimensional support model that includes technological capabilities, industrial cooperation and capital acceleration. Startups that join the Lenovo Star Program have the chance to receive comprehensive and diverse support from Lenovo Group, the "leader", including but not limited to joint research and development, opening up the ecosystem, capital support, business development, brand and service promotion, etc. Since the launch of the "Star Program for Startups", it has already helped hundreds of startups to accelerate their development and has become an important part of the cooperation and mutual reinforcement between Lenovo and small and medium - sized enterprises. In the era of accelerating the development of new productive forces, the Lenovo Star Program for Startups will further expand the recruitment of ecosystem partners and hopes to build a community of shared interests with more innovative forces in the industry and jointly reach new heights in the global technology competition.

Put simply, DeepFlow Intelligence mainly helps AI companies or AI businesses in the market to better solve the problems in computing power economy management. It's like a super - administrator that, through the finely - tuned observation and control of the execution process of large models in a GPU cluster, better distributes computing tasks, completes the same tasks with fewer machines and fully utilizes the machine performance. In addition, it can also uniformly manage different cloud services and hardware devices and automatically adjust task distribution to make the utilization of computing power economy simpler and more efficient.

In the view of Huang Kecheng, co - founder and CEO of DeepFlow Intelligence, "the next decade is a very important period for the implementation of AI. We think it's a long - term matter. So we've decided to focus on serving the scenarios where large AI models are used for implementation."

"DeepFlow Select", intelligent computing power economy management

Entrepreneurship in the field of AI computing power is a vast untapped area. According to forecasts, the market for AI computing power will have a high growth rate in the next few years and could reach a value of hundreds of billions of US dollars by 2030. However, building AI computing power is not possible overnight, and the competition and risks one faces cannot be ignored:

First, high technical barriers. AI computing power economy management covers several complex areas such as distributed computing power, task planning, load balancing and hardware optimization, and the technical barriers are very high. To achieve efficient resource distribution, the platform requires strong algorithm support, such as task priority optimization, error prediction and real - time adjustment. Development and optimization are difficult.

Second, market dominance by tech giants and competition among similar startups. Many tech giants have already developed their own solutions for computing power economy management. Startups are at a disadvantage in terms of resources and ecosystem. A large number of AI infrastructure startups have emerged both in China and abroad, and the market competition is fierce. It's not easy to build a technological barrier and brand presence in a short time.

Third, the business model is still in the testing phase. Many small and medium - sized enterprises do not fully understand the value and actual effects of AI computing power economy management platforms. Marketing and market development require more resource investment. The needs and budgets of different customers also vary, and it's difficult to develop a suitable pricing model and make a profit.

Although one faces the challenges of technological accumulation and market competition, the Chinese market is in a phase of rapid growth, and the problems are obvious. This gives startups more entry opportunities.

DeepFlow Intelligence was founded in August 2023 and focuses on providing computing resources, models and holistic optimization management platform products as well as hardware - software integration solutions for the implementation of large models for AI computing centers and AI application enterprises.

Currently, DeepFlow Intelligence has two main products: the Emerging AI Computing Platform for AI computing centers and the ENOVA Application Acceleration Platform for large AI models.

Huang Kecheng explains that the core technology of the Emerging AI Computing Platform is to develop a "process - based" performance monitoring technology for large models and GPU clusters. Compared with traditional monitoring methods, it can more accurately observe and control the execution process of large models in a GPU cluster and thus provide data for more efficient planning of large model tasks and optimization of GPU performance. "For example, in the past, I could only achieve the required parallelism with ten machines, but now two machines are enough," said Huang Kecheng. "Through our platform, we can fully utilize the machine performance."

The ENOVA Application Acceleration Platform can uniformly manage computing resources in a multi - cloud environment, support more than 20 common hardware models and thus achieve efficient resource utilization and improved computing performance. The platform automatically adjusts the distribution of computing tasks based on task requirements and the status of computing resources to achieve "atomic" resource planning efficiency and optimal resource configuration. "We jokingly call it 'DeepFlow Select'," said Huang Kecheng.

When it comes to the relationship between startups and large enterprises, Huang Kecheng is generally optimistic, although competition is inevitable. He believes that there is enough room in this market and the time frame is long enough, and that both the Chinese and foreign markets have a certain tolerance. So there are many opportunities for cooperation between startups and large enterprises. Startups are more suitable for the needs of small and medium - sized enterprises and complement the ecosystem of large enterprises.

"Hard - core user" experiences the Lenovo ecosystem

Since its founding more than a year ago, DeepFlow Intelligence has partnered with customers in AI computing centers in China, North America, Southeast Asia and some cross - industry application enterprises. Commercial operations are gradually moving in the right direction.

Huang Kecheng said that the company currently has an urgent need for industry customer resources and industry knowledge. "We are currently communicating with the intelligent manufacturing industry. In cooperation with them, we found that we need to find people who are on the same wavelength as us to quickly access the points that interest customers," said Huang Kecheng. "There are some key scenarios, such as finance, biomedicine, intelligent manufacturing, edge computing, etc. We hope to build a common understanding of these areas together."

For these reasons, the investor of DeepFlow Intelligence recommended the second season of the "Lenovo New Business Innovation Ecosystem Roadshow" to them. Huang Kecheng "found it very valuable" and therefore signed up for the competition.

As a consumer, Huang Kecheng is very familiar with Lenovo and even feels attached to the company. He is a hard - core user and has been buying Lenovo products for years. A few years ago, he also bought a Lenovo Xiaoxin 14 pro as his main computer. "The user experience of Lenovo's computer products is very good." He is particularly interested in the development of Lenovo's AI PCs and Lenovo's attempts in the field of edge AI. DeepFlow Intelligence is currently cooperating with an international edge chip manufacturer to provide monitoring and planning services for its edge tasks.

"Lenovo's AI PC strategy is a very correct decision. We believe that a large part of the technological dividend lies in high - quality data sets, and Lenovo has very comprehensive and powerful intelligent end - devices, which means that Lenovo has very high - quality data. This is a great advantage for the development of AI PCs," said Huang Kecheng.

Huang Kecheng prepared carefully for this roadshow and said that he went with the attitude of making friends. At the roadshow, he could first feel Lenovo's care for this event. "Lenovo's ecosystem thinking is very strong and directly shows the commitment and vision of an enterprise. For us startups, it's very helpful to find our value and grow in the Lenovo ecosystem in the future."

Thanks to the communication platform provided by Lenovo, Huang Kecheng was also impressed by the technologies and concepts presented by AMD. On the day of the event, he established contact with other entrepreneurs. "The ten startups that were finally on stage are all very excellent projects."

Huang Kecheng was very happy about winning the award. He believes that this is a double acknowledgment for DeepFlow Intelligence. On the one hand, the innovative technology of DeepFlow Intelligence is recognized, and on the other hand, the market potential of DeepFlow Intelligence is also seen.

The "big brother" for young entrepreneurs

Huang Kecheng studied at the Massachusetts Institute of Technology and is a young serial entrepreneur in the technology industry. "The entrepreneurial environment at MIT is very strong," said Huang Kecheng. Even during his studies, he made an attempt at entrepreneurship and gained a lot of experience.

After graduating from MIT, he quickly made his first attempt at entrepreneurship and joined as a co - founder of a brain - computer interface company that focuses on the research, development and production of devices for monitoring the nervous state. This experience helped him to gain team management experience and understand the balance between market demand and technological innovation.

DeepFlow Intelligence is Huang Kecheng's second attempt at entrepreneurship. He attaches more importance to market demand and tries to use technological trends to fill market gaps. Today, DeepFlow Intelligence is developing more and more stably and is working step by step on the further development of its products and commercial implementation.

Lenovo, a global technology conglomerate, is an ideal partner for DeepFlow Intelligence. Huang Kecheng is very eager to join the Lenovo ecosystem and conduct more in - depth cooperation with Lenovo.

"We are more of a research - oriented company and have deficiencies in understanding the industry. We need the support of companies like Lenovo. Also in the field of investment, we are willing to communicate with Lenovo Capital and even cooperate directly," said Huang Kecheng.

For young entrepreneurs, Lenovo, a strong, open and patient company, is like a "big brother" on the entrepreneurial path, always standing by. When a company encounters difficulties, Lenovo can always provide timely guidance or support to help the company overcome the difficulties.

In the next step, DeepFlow Intelligence will continue to improve its technological capabilities and at the same time accelerate the process of commercial implementation so that the company can move forward steadily. Huang Kecheng is an optimist and has full confidence in Lenovo and the Lenovo ecosystem. "This comprehensive ecosystem support gives us more courage to move forward."