Big image: The great challenge in analyzing a rapidly growing company such as NVIDIA is an understanding of all the different companies in which they participate, many of the products they announce, and the comprehensive strategy they follow. After the keyword speech by CEO Jensen Huang at the annual company GTC Conference This year, the task was particularly arduous. As usual, Huang covered a huge set of topics on a lengthy presentation, frankly, leaving more than a few people scratching their heads.
However, during a useful questions and answers session with industry analysts a few days later, Huang shared many ideas that suddenly made all the products and partnerships that you covered, as well as thinking behind them, Crystal Clear.
In essence, he said that NVIDIA is now an Amnesty International Infrastructure, as it creates a platform for devices and programs that large cloud computing providers, technological sellers and information technology departments in institutions can use to develop autonomous applications.
It goes without saying, this is far from his role as a grooming chips for computer games, or even from her efforts to help pay algorithms. However, it apparently unites many different ads and provides a clear indication of the place where the company is heading.
https://www.youtube.com/watch?
NVIDIA goes beyond its assets and reputation as a semiconductor design house in the decisive floor of the empowerment factor for the infrastructure of the future world of abilities that work with artificial intelligence materials, or Huang described it, “Intelligence Factory”.
In his main book GTC, Huang discussed NVIDIA’s efforts to enable an effective generation of symbols in modern basis models, and to link these symbols to the intelligence that organizations will benefit from to generate future revenues. He described these initiatives as building a factory for Amnesty International, related to a wide range of industries.
Although it is a great vision, the signs of the emerging economy-based economy-and the competencies brought by artificial intelligence to traditional manufacturing began to become clear. One of the companies that were built only on the services of artificial intelligence (Think Chatgpt) through manufacturing and automatic distribution of traditional commodities, there is no doubt that we move to an exciting new economic era.
In this context, Huang extensively identified how the latest NVIDIA shows create a faster and more efficient symbolic symbol. He initially dealt with the conclusion of artificial intelligence, which is the simplest of artificial intelligence training that initially brought NVIDIA to a prominent position.
However, Huang argued that inference, especially when using it with amazing new thinking models such as Deepseek R1 and Openai’s O1, will require about 100 times more computing resources than current inference methods. In other words, there is no reason to worry that the most efficient large language models will reduce the demand for computing infrastructure and we are still in the early stages of the AI’s infrastructure.
One of Huang, the least understanding ads was a new software tool called Nvidia DynamoDesigned to enhance the inference of advanced models.
Dynamo, a version that was promoted from the NVIDIA TRIONEN’s TRTON inference program, is dynamic in a GPU resource in the various stages of reasoning, such as Premill and Decode, each with distinct computing requirements. It also creates a cache of dynamic information, and is efficiently managing data across different types of memory.
Dynamo works similar to the coordination of containers in cloud computing, and it is intelligent with the intelligence of the resources and data needed to generate the distinctive symbol in the artificial intelligence factory environments. Nvidia called Dynamo “OS for AI factories.” Practically, Dynamo enables institutions to deal with up to 30 times of inference requests with the same resources of devices.
Of course, GTC will not be if NVIDIA does not also have chip and devices ads and there was a lot this time. HUANG presented a road map for GPU in the future, including update to the current Blackwell series called Blackwell Ultra (Series GB300), providing improved HBM memory on improved performance.
The new Vra Rubin structure, which includes the new arm -based central treatment unit called Vera and GPU of the next generation called Rubin, each includes more cores and advanced capabilities. Huang has hinted to a generation behind it – which was named after mathematics, Richard Fainman – the display of NVIDIA Road Map to 2028 and beyond.
During a later question and answers session, Huang explained that the disclosure of future products provided it is very important to the ecological system partners, allowing them to prepare adequately for the upcoming technological transformations.
https://www.youtube.com/watch?
Huang also emphasized many partnerships announced in GTC for this year. The great presence of other technological sellers showed their passion for participating in this growing ecosystem. On the side of the account, Huang explained that the infrastructure of the artificial intelligence agency to the maximum requires developments in all traditional computing areas, including networks and storage.
To this end, NVIDIA has unveiled the new silicon technology of optical networks between GPU and discussed a partnership with CISCO. The Cisco Cisco Silicon partnership in routers and keys designed to integrate the AI factories that suffer from GPU into institutions environments, along with a joint program management layer.
For storage, NVIDIA cooperated with leading devices service providers and data platforms companies, ensuring that its solutions may benefit from the acceleration of GPU, thus expanding the market impact on NVIDIA.
Finally, based on the diversification strategy, Huang presented more works by the company for self -government vehicles (especially a deal with General Motors) and robots, both described as part of the next big stage in developing artificial intelligence: AI.
NVIDIA knows that being an infrastructure and an ecological system means that it can benefit directly and indirectly with the high tide of the computing of artificial intelligence
NVIDIA offers components for automobile companies for many years so far, and likewise, they had robots for several years as well. However, the matter is different now that they are linked to Amnesty International’s infrastructure that can be used to train the models that will be published on these devices better, as well as providing inference data in the actual time needed to operate them in the real world.
While this tie to the infrastructure can be said that it represents a relatively modest progress, in the greatest context of the infrastructure strategy of the company’s comprehensive company, it is more logical and helps to link many of the company’s initiatives in a coherent whole.
Understanding all the different elements that HUANG and NVIDIA in GTC this year are not simple, especially because of the extinguishing nature of all different ads and much broader access to the company’s ambitions. Once the pieces are collected, the NVIDIA strategy becomes clear: the company plays a much larger role ever and is in a good position to achieve its ambitious goals.
At the end of the day, NVIDIA knows that it is an infrastructure and an ecosystem that means that it can benefit directly and indirectly with the high tide of artificial intelligence computing, even as their direct competition rises. It is a smart strategy and a space that can lead to greater growth for the future.
Bob Odonil is the founder and chief analyst Research Technalysis, LLC Consulting technology company provides strategic advisory services and market research for the technology industry and professional financial community. You can follow it on Twitter Bobodtech