Using Nvidia’s Blackwell superchip to power a new era of artificial intelligence - Cafeqa

Using Nvidia’s Blackwell superchip to power a new era of artificial intelligence


The new superchip from chipmaker Nvidia will supposedly accelerate AI, the company claims. Is a new industrial revolution about to begin that will thirstily consume power?

Since Taylor Swift was conspicuously absent, it could not have been a concert. Even yet, on March 18, hundreds of fans flocked to a San Jose, California arena to watch and celebrate. The event’s main attraction was Jensen Huang, who showcased a new chip that will be released later this year. Following his two-hour performance, almost 27 million viewers had seen it on YouTube.

At Nvidia’s annual developer conference, Huang—CEO and co-founder of the company—presented while wearing his signature black leather jacket. Nvidia is still mostly unknown outside of the IT industry, but it has just made headlines as the third-most valuable US listed business, after Apple and Microsoft, thanks to a market value that surpassed $2 trillion (€1.84 trillion).


The company’s chips, known as GPUs, are involved in all of this. As a chip designer, Nvidia contracts with specialized manufacturers to produce its chips. After discovering new markets for its hardware in areas such as bitcoin mining, 3D modeling, and autonomous cars, the business expanded beyond its original focus on video games.

Crucially, they shifted their focus to incorporating their chips into GAI systems, which are self-learning AIs that can produce various forms of media such as text, graphics, and music.


Artificial intelligence (AI) technology alone may not seem like much, but ever since ChatGPT was introduced in November 2022, the world has been abuzz with the potential around this technology. Cloud computing behemoths and AI model builders are now Nvidia’s most valuable clients.

The latest superchip from Blackwell

Nvidia has an opportunity to fuel this revolutionary technology with its expertise. Approximately 80% of the world’s demand for these AI processors is now supplied by it.

Its name is Blackwell, and it was unveiled in California. It is an improvement over the company’s previous H100 chip, which Huang said was the most sophisticated GPU being manufactured at the time, and it has 208 billion transistors. Some activities are completed 30 times faster by the next-generation processor compared to its predecessor.

The corporation reportedly invested around $10 billion to create Blackwell, as stated by Huang. Costing between $30,000 and $40,000 per chip. The latest offering from the business is an AI chip, and the makers are hoping it will improve their market share.

The technology’s operation is described

A component of an advanced system, the Blackwell chip is said by the business to be suitable “for trillion-parameter scale generative AI.” Tasks are broken down into smaller bits by the chips. Faster computations are achievable thanks to this parallel processing.

According to Bibhu Datta Sahoo, a professor at the University at Buffalo’s Center for Advanced Semiconductor Technologies, the new chip has many characteristics that decrease energy consumption and delay.

The Blackwell chip allows for the connection of several GPUs, among other advantages, allowing for the training of big AI models with a minimal impact on the environment. On top of that, it supports switching data processing across various kinds of chips by including rapid decompression of the majority of data formats.

When asked whether the chip may have a profound impact, Sahoo told DW that it’s hard to tell since there are so many groups developing technologies that could transform the way AI models are trained. Even so, “the Blackwell chip is a very good step in the right direction.”

Energy savings, increased power

The time for change has arrived, according to Huang, who has pointed out that accelerated computing has hit a tipping point and general-purpose computing has lost pace. In San Jose, he said that a new industrial revolution was beginning. We need to be creative to find methods to scale up while keeping prices low. Then we can “consume more and more computing while being sustainable.”

Data centers must expand and upgrade in order to meet these demands. But there are those who worry that AI processors, which use a lot of electricity, may further strain power infrastructures. Even though their new processor is more powerful, Nvidia claims it uses less electricity overall.

No one disagrees amongst experts. In comparison to earlier generations of GPUs, the Blackwell chip may reportedly cut power usage during the training of massive AI models by a factor of three to four, according to Sahoo.

The significance of this energy efficiency cannot be overstated “given that the anticipated increase in the power consumption of data centers in the US from 17 GW in 2022 to 35 GW by 2030

Some are sceptical and worry about a financial bubble as investors rush in, despite the fact that there have been significant advances in developing strong processors for the next generation of artificial intelligence.

The most notable increase has been seen among AI hardware manufacturers. Since software cannot be utilized without the supporting infrastructure, this is to be expected. The expansion of technology is possible now that the necessary infrastructure is being put in place.

For the purpose of connecting and managing its superchip technology, Nvidia is increasing its spending in networking and software.

But there will be plenty of additional obstacles in the years to come. There may be pressure on supply networks throughout the world due to the increasing demand for semiconductors. To make matters worse, a large portion of the world’s chip production takes place in Taiwan.

The other players aren’t sitting on their hands, either. Intel, AMD, Cerebras, and Groq are just a few of the major competitors developing their own CPUs. Amazon, Google, and Microsoft—three of Nvidia’s most important clients—are all venturing into the semiconductor design industry.

It will be a costly struggle to maintain dominance in a market where scale counts and where new technologies are obsolete in a flash.

Lastest News