Nvidia
Once considered a two-bit components manufacturer, Nvidia is set to produce one of the world's most powerful Central Processing Units (CPU), tailor-made for high-powered computing, artificial intelligence, and data processing.
Nvidia named its new CPU Grace after renowned computer scientist Grace Hopper, which it developed in conjunction with U.K. mobile chip behemoth Arm Ltd, a company that Nvidia is hoping to acquire.
Per Nvidia, Grace will boast ten times the performance of any data center chips on the market. According to the company, such high performance is necessary given the billions of parameters handled by today's AI models. Parameters which double every two and half months.
Nvidia's foray into the CPU market puts it on a collision course with Intel. Once the standard-bearer for US chip manufacturing, Intel has seen it's fortunes crumble of late. The company failed to build out its 7-nanometer chips last year, putting it squarely behind top-tier chip manufacturers. As a result of this and CEO Jensen Huang's remarkable tenure, Nvidia managed to overtake Intel last year as the largest chip maker by market value, according to the Wall Street Journal.
Nvidia's planned acquisition of the aforementioned Arm Ltd could give the company an even greater edge against Intel going forward. The $40 billion acquisition would be the largest such deal in the history of the industry. Arm's chips are the copied on earth, with more than $160 billion in components made according to the companies designs.
Such acquisitions are necessary in a chip market that demands greater and greater specialization. During the Intel era, the market wanted chips that were reasonably good at doing most things. But today, the paradigm is shifting toward more specialized chips.
Going forward, Intel won't be helped by its practice of designing and producing its chips in-house. The implication here is that if Intel wants to build more specialized chips, not only will it have to design them, it will also have to perfect the manufacturing process.
This model is the opposite of Nvidia's, which licenses out the manufacturing of its chips to third parties like Taiwan Semiconductor Manufacturing Company
Specialized chips like Nvidia's Grace, whose first two customers are the Los Alamos National Laboratory and the Swiss National Supercomputing Center. While Grace might initially power supercomputers, remember that yesterday's supercomputer is today's smartphone.
Intel's chips might be fine today, but the company still has a lot of ground to make up. Nevertheless, Nvidia also has a lot of ground to make up. In 2019 the company pulled in just $10.9 billion in sales versus Intel's $72 billion.
But those are 2019's figures, so who's to say whether the situation won't be reversed in 2029.
- 1.https://www.wsj.com/articles/nvidia-plans-new-chip-to-compete-with-intel-in-data-center-market-11618264071?mod=tech_lead_pos7
- 2.https://nvidianews.nvidia.com/news/nvidia-announces-cpu-for-giant-ai-and-high-performance-computing-workloads
- 3.https://www.wsj.com/articles/amd-nvidia-chip-away-at-intels-semiconductor-dominance-11603909252?mod=article_inline
- 4.https://en.wikipedia.org/wiki/Arm_Ltd.
- 5.https://www.anandtech.com/show/16610/nvidia-unveils-grace-a-highperformance-arm-server-cpu-for-use-in-ai-systems