Even after some chip fabrication stumbles and falling behind some of its peers technologically, Intel (NASDAQ:INTC) remains by far the world’s largest semiconductor company as measured by revenue. It hauled in a massive $77.9 billion in sales in 2020, an 8% year-over-year increase. With a global semiconductor shortage expected to last for the foreseeable future, Intel has said it will double down on its manufacturing and build a couple of new factories in Arizona.
A sea change is starting in the industry, though. NVIDIA (NASDAQ:NVDA) is flexing its muscles and announced at its investor day last week that it’s prepping a new data-center processor aimed at one of Intel’s tentpole businesses. As computing needs quickly evolve, NVIDIA is on a path to industry domination.
New processors for a new era of AI
NVIDIA has picked up serious momentum in recent years designing GPUs (graphics processing units) into data centers as computing accelerators. A GPU’s strength is in its ability to break down large and complex tasks into smaller parts and compute them in parallel. This makes them ideal for high-performance computing like training artificial intelligence (AI) algorithms, which require that massive amounts of data be worked through.
As a result, NVIDIA has been picking off chunks of data center market share as its GPUs get added to the general-purpose CPUs (central processing units) that dominate the space currently.
To be sure, CPUs will remain an essential part of data centers, which are massive and complex computing units, and CPUs will handle basic tasks like pulling information from memory banks and coordinating the movement of data. This is the realm Intel has dominated for decades. Its data center group (DCG) reported sales of $26.1 billion last year, accounting for one-third of total revenue. All by itself, Intel’s DCG dwarfs NVIDIA’s total revenue of $16.7 billion last year.
But NVIDIA isn’t content to just add its GPUs to data centers. It’s coming for Intel’s bread-and-butter CPU market share. It indicated as much with its pending acquisition of chip architecture designer ARM Holdings. And at its investor day, NVIDIA announced it’s directly entering the data center CPU market with Grace, a high-performance CPU based on ARM chip designs that will be available in 2023. Interestingly, NVIDIA is also planning to sell Grace on a single circuit board with its GPUs, connecting the two with networking tech it picked up from its Mellanox acquisition last year.
Intel is gearing up for expansion of its in-house manufacturing, but NVIDIA could put a damper on those efforts by seriously raising the bar in chip performance.
A more modern operating model
Granted, NVIDIA doesn’t fabricate its own chips. It simply designs them and outsources manufacturing to the likes of Taiwan Semiconductor Manufacturing (NYSE:TSM) — which incidentally surpassed Intel’s chip fab size a couple of years back and currently boasts 57% of global chip-manufacturing market share. Given this operating model, NVIDIA’s design house isn’t going to surpass Intel’s integrated design and fabrication business in terms of sales anytime soon.
Nevertheless, NVIDIA Grace could be just the first shot across Intel’s bow as it seeks to scoop up more chunks of the modern data center’s construction. Computing needs are rapidly evolving, and cloud computing is putting the power of entire AI systems at the fingertips of users working from PCs, laptops, and mobile devices. NVIDIA is purpose-building its suite of hardware for this growing demand, helping it leapfrog Intel technologically and confirming its spot at the table in a few years.
Again, Grace won’t start selling until 2023 (NVIDIA said it will use these chips internally for now in its own private servers), but Intel has nevertheless been put on notice. When it comes time to upgrade CPUs, customers will soon have a new option that integrates seamlessly with the GPUs they’ve already started installing in recent years.
NVIDIA’s biggest end-market remains video games, but its second-largest data center vertical is only just getting started. Intel has the most to lose, and NVIDIA’s tech is already the name to beat. The winds of change are blowing strongly in the latter’s favor.
This article represents the opinion of the writer, who may disagree with the “official” recommendation position of a Motley Fool premium advisory service. We’re motley! Questioning an investing thesis — even one of our own — helps us all think critically about investing and make decisions that help us become smarter, happier, and richer.
View more information: https://www.fool.com/investing/2021/04/19/why-intel-investors-worry-nvidia–data-center-chip/