In recent years, every industry outlet, meta-universe, new energy, automatic driving, industrial robots, artificial intelligence engines to AIGC, NVIDIA has not missed, each of them has participated and done well, just two days ago NVIDIA also announced that NVIDIA Research won the innovation award of the autonomous driving challenge on CVPR and was the first place.
Today we will not talk about NVIDIA's achievements in the field of autonomous driving, we still look at its layout in the field of AI.
On May 30, NVIDIA's stock price rose briefly to more than $400, successfully jumping into the ranks of companies with a global market value of more than trillion US dollars, ahead of NVIDIA are Apple, Microsoft, Saudi Aramco, Google's parent company Alphabeta and Amazon. But just a few hours later, the stock price fell back and fell back below trillion again.
Like a dark horse, Nvidia briefly knocked on the door of the trillion-dollar club.
Although involved in many fields, but since the establishment of the company 30 years ago, NVIDIA's core product is only one, that is, graphics chips, compared with general-purpose chips, graphics chips are good at processing parallel computing, early is a special chip to solve PC image display. The creation of NVIDIA originated from the exploration of the billion market at that time, although it was aimed at the "non-existent market" 30 years ago, but 30 years later it was a feverish market of up to trillions of dollars - especially under the impetus of AIGC, the demand for AI infrastructure is huge, NVIDIA holds its development lifeblood, graphics card chips represent computing power, machine learning. NVIDIA A100 chip and higher generation H100 chip and corresponding graphics card are even more difficult to find.
In this historic wave of generative AI, NVIDIA founder and CEO Jensen Huang announced a number of important NVIDIA announcements at the COMPUTEX2023 computer show in Taipei on May 29, including the NVIDIA DGX GH200 AI supercomputer, NVIDIA MGX, an open modular server for accelerated computing, the GH200 Grace Hopper superchip, Seven products, including the acceleration network platform Spectrum-X, are announced to further help revolutionize accelerated computing and AI.
The next day, NVIDIA's stock price rose rapidly, and the company's market value exceeded the trillion-dollar mark for the first time. It has to be said that the east wind of the stock price rise comes from AIGC.
In the past few months, ChatGPT's explosion has made AIGC the hottest track known to everyone on the earth, and various VC investment institutions have rushed in, and many AI startups have sprung up. According to public data, the total financing of the primary market of the global AIGC industry reached 2.642 billion yuan in April, and some institutions predict that the compound annual growth rate of AIGC will reach 60%.
Although NVIDIA is in the leading position in the AI era, strong competitors are still close behind. Before there is chip giant Intel, and then there is the well-known semiconductor chip company Chaowei Semiconductor AMD.
AMD will launch the MI300 against the NVIDIA GH200 supercore this year to focus on the artificial intelligence large model training market; Intel is also stepping up its own research and development of artificial intelligence chips, and it is reported that it has recently acquired a series of AI-accelerated chip companies. OpenAI, which has always worked closely and closely, has launched an open-source computing power programming model TRITON, which no longer relies solely on NVIDIA's CUDA (NVIDIA's general-purpose parallel computing architecture computing platform).
In such a competitive situation, NVIDIA chose to wait for the opportunity and began to look for a new moat.
In the past two months, NVIDIA announced that it has invested in 3 startups in the field of artificial intelligence, and invested in 3 unicorn companies in subdivided fields in one go, which shows its hard work in the field of AI artificial intelligence.
01. Deep Binding Partners
Further occupy the market segment
On April 21, CoreWeave, an artificial intelligence computing power provider that has successfully transformed from a blockchain mining company, announced a $221 million Series B financing, led by Magnetar Capital, with participation from NVIDIA, former GitHub CEO Nat Friedman and former Apple executive Daniel Gross. Magnetar contributed $111 million, with the rest split equally between NVIDIA, Friedman and Gross. An Nvidia spokesperson also said the investment represents a "deepening" of its partnership with CoreWeave.
Then, on June 1, CoreWeave announced a new round of funding of $200 million, quickly becoming a unicorn.
The boom in generative AI technology has accelerated the need for CoreWeave's specialized cloud infrastructure, which is designed to provide cloud services specifically for enterprise-grade GPU-accelerated workloads. On its website, CoreWeave claims to build cloud solutions for compute-intensive use cases (machine learning and artificial intelligence, visual effects and rendering, and pixel streaming) that are 35 times faster than large general-purpose public clouds, offering computing power "80 percent cheaper than traditional cloud service providers" while providing computing power for a wide range of products, including the Nvidia A100 GPU chip. And on its website, we also learned that CoreWeave offers cheaper NVIDIA A40 GPUs, mainly for visual computing, and A100 GPUs for artificial intelligence, data analysis and high-performance computing.
Nvidia's global director of partner business development claims that CoreWeave is a valued and loyal partner of NVIDIA and the first elite computing cloud solution provider in the NVIDIA partner network. CoreWeave invests in a wide range of high-end NVIDIA GPUs, including the largest A40 deployment in North America, to ensure customers have access to the best compute for the complexity of their workloads. In addition, CoreWeave uses the NVIDIA Quantum InfiniBand network ensemble and NVIDIA SHARP to build an A100 distributed training cluster with orbit optimization design to provide the highest possible distributed training performance.
The close cooperation between NVIDIA and CoreWeave shows that NVIDIA is not only developing itself, but also seeking a number of external assistance to further occupy the segment of the AI computing power service segment.
02.Second only to OpenAI and Anthropic
At this point, NVIDIA has Cohere
In early June, AI startup Cohere announced the completion of a $270 million Series C funding round with a valuation of more than $2.1 billion, becoming another AI super unicorn. So far, Cohere has raised a total of $445 million, second only to OpenAI and Anthropic. Behind this financing is still Nvidia.
Unlike OpenAI's ChatGPT, which focuses on C-end consumers, Cohere focuses on the To B enterprise market. Cohere proposed that the product is not limited by any cloud platform, ensuring the private security of enterprise data, and its products also have two characteristics: model customization and customer support. Cohere provides enterprise customers with custom model training, i.e. fine-tuning, to adapt to different enterprise industry segments.
Cohere's model can create customized AI based on company-specific data, provide data security deployment options in the company's existing cloud environment, and provide multiple cloud hosting options (private cloud, AWS, Google, Cohere hosted cloud, etc.), allowing enterprises to have full control over data security and privacy, and protect the absolute security of data privacy.
Technically, according to the Stanford HELM language model introduced on its website, Cohere's latest model Command has 52 billion parameters, which is the largest version of Cohere's model, and its accuracy performance is higher than many well-known models that are 3 times larger, including GPT-3 with 175 billion parameters.
At the same time, Martin Kon, the current president and chief operating officer of Cohere, said in a previous interview that "search and retrieval will become the next core growth area for Cohere", Cohere is committed to enhancing the interpretability of AI in its artificial intelligence models, that is, the searchability of AI-generated content, when AI generates content, it will have indexed content sources, increase explainable output, so that users can verify and judge the generated content, without blindly trusting a model.
"We are at the beginning of a new era driven by accelerated computing and generative AI, and Cohere's team has made fundamental contributions to generative AI, and their services will help enterprises around the world leverage these capabilities to automate and accelerate," said Jensen Huang said in this round of investment. ”
So far, Microsoft invested in OpenAI, Google invested in Anthropic, and Nvidia had Cohere.
03. Within half a month,
Invest unicorns again
On June 14, Synthesia, an AI generation tool in the field of video production, announced the completion of a $90 million Series C financing, with a valuation of $1 billion, officially promoted to the ranks of unicorns, this round of investment was led by Accel, and NVIDIA (its fund NVentures participated) still appeared in the list of investors.
Synthesia is a British company that uses AI-generated technology to synthesize virtual digital human videos, with the concept of "replacing cameras with code, so that everyone is creative". Its main creative team are from well-known universities such as University College London, Stanford University, Technical University of Munich and University of Cambridge. Currently, its AI video generation platform supports text-to-speech functionality in more than 120 languages, allowing users to choose from more than 140 different AI avatars without the need for equipment or video editing skills. At the same time, in order to prevent malicious production of virtual character videos, Synthesia also has built-in multiple layers of security and authentication, and the generated videos are audited.
In fact, this company exploded in the AI video virtual human track a few years ago, because it occupied the two AI subdivisions of AI-generated video and virtual digital human at the same time.
Now Synthesia provides an AI-driven platform for enterprises, including digital assets such as character library, template library, language library and supports customer-defined images, which greatly saves the time and cost of video production compared with traditional video production methods, and has a high degree of video standardization. Its platform offers highly customizable products that are suitable for the needs of most businesses. According to data released by Synthesia, more than 50,000 enterprise customers have used Synthesia's AI video products, including large companies such as Amazon, Johnson & Johnson, Accenture, Tiffany and other large enterprises, of which almost 35% of the Fortune 100 list has partnered with it.
At present, Synthesia is already the number one AI video creation platform in the field.
NVIDIA is still everywhere.
04. Write at the end
Throughout the history of NVIDIA's development, all its involvement and investment actions revolve around its technical core - chip architecture, and at the same time, it continues to innovate and iteratively upgrade products around this core, resulting in its market share has been far ahead. Its CUDA parallel computing platform integrates the entire NVIDIA system, deeply binds partners and developers, and builds a strong ecological barrier for the development of NVIDIA. While improving R&D efficiency to defeat opponents, NVIDIA has also been constantly enriching its product matrix to meet the needs of customers in different industries, and its industrial layout has shown diversified development. Under the explosive wave of AIGC, the AI market has become a new target for NVIDIA to reconstruct business growth, focusing on the "core" to lead artificial intelligence computing.
END
Welcome to "Futuristic Universe" to explore AIGC together!
Notice: This article is the original article of "Futuristic Universe", please contact the public account background for reprinting, thank you for your support.
Image credit: NVIDIA, CoreWeave, Cohere, Synthesia
References: https://www.nvidia.cn/events/computex/