Amazon (AMZN 0.87%) recently joined other tech giants such as Apple (AAPL 2.21%) and Alphabet (GOOG 0.58%) (GOOGL 0.42%) in designing its own custom chips. According to a report by The Information, Amazon has begun designing custom artificial intelligence (AI) chips for its Alexa-enabled home speakers.
The report also said that Amazon may be looking to extend the chipmaking venture into products for its Amazon Web Services (AWS) data centers. If that happens, it could potentially spell trouble for high-flying AI chipmakers NVIDIA (NVDA 1.58%) and Intel (INTC +1.15%). Here's how Amazon's ambitions could affect the AI race.
Could Amazon's new artificial intelligence chip threaten Nvidia's data center growth? Image source: Getty Images.
First to Alexa
Alexa has been the main focus for Amazon this year, as Alexa adoption has surprised even optimistic company executives with its growth. With Apple's recent entry into home speakers with the HomePod, however, Amazon likely feels the need to up Alexa's game in both sound and speed. A custom chip could boost Alexa's inference capabilities -- the ability to process external inputs like speech and make sense of them at the device level. Doing more processing at the device level means less information needs to be sent back to the cloud, which means faster response times.
Amazon has developed these chipmaking capabilities over the past few years through both internal hires as well as acquisitions, buying Annapurna Labs, an Israeli-based chipmaker, in 2015 for $350 million, and Blink, a maker of low-power security cameras, for $90 million late last year. Many believe Amazon was really after Blink's power-management chipmaking capabilities, rather than just its camera products, as Amazon already has its own Cloud Cam product that it rolled out last year.
Then to the data center?
The report was mostly centered on Alexa, but an AWS data center chip was also mentioned as a possibility. While Amazon has a big lead in cloud computing, other rivals, especially Alphabet, are investing huge sums to catch up in this all-important race.
Alphabet has been investing in custom hardware for years, and recently unveiled its second-generation tensor processing unit (TPU), designed specifically for AI neural networks. According to The New York Times, the TPU development not only puts Alphabet in a race for technological superiority with makers of graphics processing units (GPUs) like NVIDIA, but may also afford Alphabet negotiating leverage over GPU producers as well. (Google still buys NVIDIA chips, according to The Times. GPUs, originally designed for graphics and gaming, are the leading machine-learning chips today.) In addition, Alphabet has started renting out its TPU computing power to developers using Google's Cloud Platform.
The combination of technological innovation and the ability to lower costs is likely what caused Amazon to start developing its own chips, as it looks to maintain its cloud computing lead against these well-heeled competitors.
Trouble for NVIDIA?
While Amazon and Alphabet's development of custom AI chips is something to monitor, NVIDIA investors shouldn't panic just yet. The company recently reported a blowout quarter and making new chips -- especially when it's not your main focus -- can be costly and time-consuming. In addition, NVIDIA still gets the majority of its revenue from the gaming division, which accounted for 60% of sales last quarter; the data center market makes up only 21% of sales. Moreover, specialty custom chips are still a small part of data center chip content, and NVIDIA's GPUs are still taking share from traditional central processing units.
When questioned about Amazon and Alphabet, NVIDIA management said that while there will always be custom AI chips for specific tasks, NVIDIA's chips should continue to be the go-to choice for a wide array of industries and functions requiring GPUs.
Still, the data center market has been NVIDIA's highest-growth segment over the past two years, with quarterly revenue up an astounding 323% compared to two years ago. Therefore, any slowdown in this segment could potentially halt NVIDIA stock's rapid ascent. That means NVIDIA investors should definitely monitor the adoption of Google TPUs and Amazon's new chipmaking venture in the year ahead.
The race for AI supremacy is on, and there are many big tech companies with lots of talent gunning to make the best artificial brain on the market. Stay tuned.
