The world's largest chipmakers are all battling to win the AI computing prize. Who's taking home the gold medal?
August 6, 2021
Anyone following the semiconductor industry lately has noticed that it’s been an exciting year. The emerging presence of artificial intelligence and cloud computing has led to a global supply shortage, where only a handful of chipmakers are capable of producing the processors required for the most demanding of applications.
Artificial intelligence is extremely computationally-heavy, and it requires extremely high-performance chips to keep up with all of its necessary calculations. The demand for AI is also going nowhere but up.
Several of those companies are knocking in out of the park in their quarterly results. But not everyone is winning in this important race.
The best looking track star right now might be AMD (NASDAQ: AMD), whose first quarter results show that it’s really hitting its stride. Revenue grew 93% and earnings per share tripled in year-over-year comparisons.
The key driver of AMD’s results has been its obsessive focus on higher-performance opportunities in recent years. Revenue from its Enterprise, Embedded and Semi-Custom segment grew an incredible 286% year-over-year and has now eclipsed $1.3 billion in quarterly sales. Within that segment, cloud data center sales “more than doubled”, as its newest EPYC 7003 processors are being deployed into the data center servers of the world’s largest cloud service providers.
Serving deep-pocketed customers with your highest-priced products is a recipe for success. AMD’s operating profit grew 233% and operating margin is now 22% of overall revenue (on a non-GAAP basis). AMD is a mega-cap chipmaker who’s financial results look more like a small-cap software company. It’s succeeding in pulling off an amazing business transition.
NVIDIA (NASDAQ: NVDA) joins AMD at the head of the pack and also reported awesome quarterly earnings. Its 79% annual growth of its own data center segment demonstrate that its focus on providing AI accelerators has more than paid off. NVIDIA’s GPUs have some of the data center’s best power efficiency, making them a popular choice for providers who like their technical performance and surrounding software ecosystem.
While Intel (NYSE: INTC) has been a dominant force in computing for half a century, it’s finding itself in a much more challenging situation when it comes to competing for the data center. Intel uses its own fabs to manufacture chips, and it has stumbled in progressing process technologies to make semiconductor chips with nodes of less than 7 nanometers. This has caused it to fall behind its chief rival Taiwan Semiconductor (NASDAQ: TSM) in manufacturing to world’s smallest and highest-performance chips.
That’s important because Intel is also falling behind chip designing rivals as well. AMD and NVIDIA both worth with Taiwan Semi as their manufacturing partner, who they outsource the manufacturing to. This has helped the chip designers keep up with the world’s most demanding semiconductor customers.
Intel’s reliance on its own in-house production has it at a competitive disadvantage for the time being. But will that disadvantage endure?
Many believe that Intel might still be able to break late in this race. Its new CEO Pat Gelsinger is highly-regarded throughout the industry and even worked directly with Gordon Moore. Gelsinger is committed to reinvigorating Intel’s manufacturing, and is investing tens of billions of dollars in growing its American-based capacity. It’s also reportedly interested in acquiring Global Foundries, another US-based contract manufacturer (that interestingly used to be the manufacturing arm of AMD).
Why would Intel be loosening the purse strings in the middle of a competitive disadvantage? There are several potential theories.
One is that the demand for computing is shooting off the charts. Intel missed out on several opportunities to serve Apple (NASDAQ: AAPL) with cutting-edge smartphone chips and then had strike two when it missed the migration to the cloud. It certainly doesn’t want to become stuck with its feet in quicksand when future opportunities arise, such as for self-driving cars and the Internet of Things.
Another theory is that there are political factors heavily at play. To counteract supply tensions in Southeast Asia, the United States has vocally expressed its interest in having a domestic second-source semiconductor supplier. There could be billions of dollars of political funding earmarked for Intel’s new fabs, as a way to offset the rising political tensions in Taiwan.
But “second-source supplier” could mean that Intel would still be playing second fiddle. Being spec’d in isn’t the same as physically manufacturing chips for contracts and collecting revenue. And cold hard cash is what actually funds capacity expansion and process technology improvements.
As such, my money is on AMD and NVIDIA continuing to remain as the favorites to win gold in the global race for AI. AMD’s decision to decouple its manufacturing and NVIDIA’s focus on introducing GPUs into the data center have paid off, and I expect both of them to ultimately stand at the top of the podium for several years to come.
Already a 7investing member? Log in here.