Nvidia's market capitalization soared by $1.8 trillion, overtaking Google and Amazon!
New Zhiyuan reports
Editor: Momoko Run
On Monday, local time, Nvidia's market value exceeded $1.8 trillion, overtaking Google and Amazon. At the same time, the new GPU RTX 2000 Ada debuted, and the performance skyrocketed by 1.5 times compared to the previous generation.
The tide of AI has made Microsoft's market value exceed 3 trillion yuan, firmly occupying the top spot in the world.
Now, Nvidia, which dominates the world by selling computing power, is about to become the biggest potential stock after Microsoft.
On Monday, local time, Nvidia's market value soared by $1.8T, even surpassing Google and Amazon, ranking third in the United States, setting a new milestone.
The CEO of Perplexity AI commented, "Nvidia's market capitalization quickly caught up with and surpassed Apple. H100, Cuda, Tensor RT have bigger moats than iPhones and AppStores."
The CEO of Stability AI even bluntly said, "GPUs are the new oil"!
Just a few days ago, Sam Altman posted, "OpenAI now outputs about 100 billion words a day. All the people on the planet output about 100 trillion words a day."
At the same time, he said, "More GPUs are urgently needed"!
At present, the chip shortage faced by OpenAI is urgent. Recently, even Altman personally went out to raise $7 trillion to build a chip empire and break the Nvidia monopoly.
确实，GPUs Are All Your Need早已成为行业共识。
According to Omdia Research statistics, last year alone, the world's major technology companies shipped 650,000 H100 units, of which Meta and Microsoft each had 15,000 GPUs.
It can be seen that the demand for GPUs to accelerate the training, integration, and deployment of AI models has surged dramatically around the world.
Just today, NVIDIA debuted the new GPU, the RTX 2000 Ada, with 2,816 CUDA cores and 16GB of memory, designed for professionals in a variety of fields to handle workflows for multi-AI applications.
The most critical thing is that foreign media Tom's Hardware reported that the price is only 625 US dollars (4495 yuan).
Can the continuous introduction of new GPUs lead LinkedIn to the world's most valuable company?
Jensen Huang: It doesn't take $7 trillion to develop AI
In the future, if anyone is going to challenge Nvidia's GPU empire, it may not be current rivals AMD and Intel, but the new chip company that Sam Altman is building.
After it was revealed that Altman was seeking $7 trillion in financing, hoping to reshape the entire chip industry, netizens first fried the pot.
"Give me 7 trillion dollars, I'm not the villain in the cartoon. 」
Professor CS of Huashengda University:
"But what you do best seems to be making up dangers that don't exist and hindering the industry through them. 」
A few days after the news was fermented, the bigwigs also began to talk about Altman's almost crazy plan.
Boss Huang said at a recent World Government Summit in Dubai. He believes that the chip industry will reduce the cost of AI because these components are being manufactured "faster and faster".
"Buying more chips is not the only solution, because all the chips are going to get faster and faster, so the total amount needed will not be as much."
The implication is that in response to Altman's plan to "rebuild a chip industry", Boss Huang feels that the existing semiconductor industry will meet the demand for chips, and it is unnecessary to "rebuild the wheel".
And Boss Huang also made it clear that in the next 5 years, the world's capital spent on supercomputing centers will double.
The AI craze is hype? We're just getting started. You can buy more Nvidia stocks
We are at the beginning of this new era. Data center installations are about a trillion dollars. In the next four to five years, we will have $2 trillion worth of data centers powering software around the world.
And many bigwigs in the technology industry also have a negative attitude towards Altman's view that he wants to rebuild the chip industry.
After all, the scale of 7 trillion seems to be too big a space hole, and the "moon landing project" may not be enough to describe his difficulty, and the "sun landing project" is more appropriate.
Sam Lessin, a Silicon Valley investor and early Facebook executive, wrote an article titled "The Era of Absurdist Capitalism."
He called Altman's $7 trillion fundraising move just a "show" to "purify" his extreme fans.
You have to admire Sam Altman...... He knows how to leave his mark in the present society.
I'm not resentful of Sam's showmanship itself – he's just continuing Elon's game of "autopilot is coming soon" or "colonizing Mars in 2024...... Make promises, make obviously ridiculous but exciting statements, all in order to attract fans who follow him... Sam is just playing a one-man game...
Start with the AGI that creates panic and when it runs out--...... Let's come up with the biggest numbers we can think of!
Of course, on this occasion, the "AI professional troll" Macus will never be indispensable.
He wrote a lengthy post on his blog titled "Why the World Is Saying No to Sam Altman's $7 Trillion Plan." 」
Article address: https://garymarcus.substack.com/p/seven-reasons-why-the-world-should?utm_source=profile&utm_medium=reader2
He believes that the $7 trillion project will have a great negative impact on natural resources, the labor market, and other aspects, and it is also very likely that human beings will be exposed to unnecessary risks at the economic and environmental levels.
And most importantly, he is very skeptical about why humanity is betting its future on large language models.
And Altman himself, in response to the criticism of the bigwigs for his ambition to challenge Nvidia and rebuild the chip empire, simply sent a tweet in response:
I'm here trying my best for our common goal, and you're blogging about me and I'm going to fail.
The new RTX 2000 Ada debuts
The RTX lineup is a new addition to the RTX 2000 Ada, a GPU designed to meet the needs of AI-accelerated design and visualization.
This entry-level GPU is made up of 19.9 billion transistors, comes with 16GB of RAM, and 2,816 CUDA cores.
Generative AI has become a cross-industry driver, allowing businesses to carefully select the hardware that meets their workflow needs.
Nvidia's launch of the RTX 2000 Ada is a response to this need, offering AI, graphics, and computing technologies.
The new GPUs shine in professional workflows, with 1.5x the performance of the previous-generation RTX A2000 12GB for tasks such as 3D environments, design reviews, and industrial design.
16GB of memory meets the needs of modern multi-app workflows, ensuring professionals can seamlessly work with AI-powered tools, multi-monitor setups, and high-resolution content.
Powered by NVIDIA RTX technology, the RTX 2000 Ada also introduces NVIDIA Deep Learning Super Sampling (DLSS) technology, which delivers ultra-high-quality, photorealistic ray-traced images that are more than 3x faster than before.
What does this mean?
Not only does this have a significant impact on graphics processing, but it also enhances enterprise virtual reality workflows, especially in areas such as product design and engineering review.
Architects and urban planners can leverage GPUs to accelerate visualization workflows, while product designers and engineers can leverage GPUs' rendering capabilities to quickly iterate and generate designs.
Content creators can seamlessly edit high-resolution videos and images, integrating AI for realistic visuals and the help of content creation.
In embedded applications and edge computing, the RTX 2000 Ada enables real-time data processing for medical devices, optimizing manufacturing processes through predictive maintenance, and integrating AI-driven intelligence in retail environments.
At present, some industry leaders have already used the power of the RTX 2000 Ada.
其中就包括，Dassault Systèmes的Solidworks应用程序，Rob Wolkers Design and Engineering，还有WSP。
Professionals who use GPUs say "its compact design, excellent performance, and game-changing 16 GB of RAM."
Highlights summary version
The RTX 2000 Ada features the latest technology in the NVIDIA Ada Lovelace graphics processor architecture:
- 3rd Gen RT Cores: Up to 1.7x faster ray tracing performance for high-fidelity, true-to-life rendering.
- Fourth-generation Tensor cores: Up to 1.8x AI throughput over the previous generation, with structured sparsity and FP8 precision, delivering higher inference performance for AI-accelerated tools and applications.
- CUDA cores: FP32 delivers up to 1.5x the throughput of the previous generation, significantly improving performance for graphics and compute workloads.
- Power efficiency: Up to 2X faster performance for professional graphics, rendering, AI, and compute workloads at 70W of power consumption as the previous generation.
- Immersive Workflows: Up to 3x the performance of virtual reality workflows compared to the previous generation.
- 16GB GPU Memory: Enables users to handle larger projects while supporting error-correcting code memory for greater compute accuracy and reliability for mission-critical applications.
- DLSS 3: A breakthrough in AI-powered graphics, dramatically improving performance by generating more high-quality frames.
- AV1 Encoder: The 8th Gen NVIDIA Encoder, a.k.a. NVENC, supports AV1 and is 40% more efficient than H.264, opening up new possibilities for broadcasters, streamers, and video callers.
Nvidia said the new GPUs will be sold through global distribution partners and will be offered by major manufacturers such as Dell, HP and Lenovo starting in April.
Create custom AI chips and share a share of the $30 billion market
Nvidia's ambitions go far beyond that.
Alan Weckel of 650 Group, a research firm, estimates that the market for custom chips for data centers will grow to $10 billion this year and double by 2025.
According to Needham analyst Charles Shih, the broader custom chip market will be worth about $30 billion in 2023, accounting for about 5% of global chip annual sales.
A few days ago, Reuters exclusively reported that Nvidia was setting up a new business unit focused on designing custom chips, including advanced AI processors, for cloud computing companies and other companies.
At present, Nvidia H100 and A100 have monopolized 80% of the high-end AI chip market.
But some tech companies like Meta, Microsoft, and OpenAI are starting to develop in-house chips to meet specific needs. Doing so helps reduce energy consumption and has the potential to reduce design costs and time.
Nvidia is trying to play a role in helping these companies develop custom AI chips.
According to people familiar with the matter, Nvidia has been in discussions with Amazon, Microsoft, Google, and OpenAI to produce custom chips for them. In addition to data center chips, it is also seeking customers in telecommunications, automotive, and video games.
Resources:
https://twitter.com/search?q=Nvidia&src=trend_click&vertical=trend
https://www.bloomberg.com/news/articles/2024-02-12/developing-ai-shouldn-t-cost-7-trillion-nvidia-ceo-jensen-huang-says
https://www.tomshardware.com/pc-components/gpus/nvidias-new-pro-gpu-costs-less-than-dollar650-rtx-2000-ada-generation-arrives-with-2816-cuda-cores-and-16gb-ecc-vram
https://blogs.nvidia.com/blog/rtx-2000-ada/