- Intel and Nvidia have lately been embroiled in a war of words regarding who is the AI leader.
- Intel claims that it's Xeon processors have been deployed in 97% of AI servers.
- Who will become the leader in this nascent field?
There is little doubt that artificial intelligence, or simply AI, is likely to become the next big opportunity for chip makers like Intel Corporation (NSDQ:INTC), the CPU leader, and NVIDIA Corporation (NSDQ:NVDA), the GPU king. Large Internet companies are already using AI to roll out a host of online services that understand speech and images. Deep learning chips are now being integrated into driverless cars, drones, and variety of other products in the much-ballyhooed Internet of things as I had discussed in an earlier article.
It's no wonder then that the two companies are already seeking to be recognized as the true leaders in this nascent field. Intel recently made bold claims that its Xeon Phi processor is 2.3x faster than GPUs in training, and provides 38% better scaling across nodes by scaling to 128 nodes while GPUs do not. Nvidia came out strongly by disputing Intel's claims saying:
''it’s understandable that newcomers to the field may not be aware of all the developments that have been taking place in both hardware and software."
"It’s great that Intel is now working on deep learning. This is the most important computing revolution with the era of AI upon us and deep learning is too big to ignore. But they should get their facts straight."
Intel's Xeon Phi Processor
Source: PC World
Nvidia says it has made big inroads in AI by designing graphic chips to process artificial intelligence applications including deep learning neural networks. But Intel has countered by saying only 3% of servers deployed for machine learning last year used GPUs. In contrast, Intel's proprietary Xeon chips have been deployed in 97% of all AI servers.
So who is the real king of AI?
The king of AI
Both Intel and Nvidia were [in]famously left trailing in the wake of the smartphone revolution. Both companies tried to later play catch up but found themselves struggling against better established competitors including Qualcomm (NSDQ:QCOM), Apple Inc. (NSDQ:AAPL), and Samsung Electronics (OTC:SSNLF). They finally decided to cut their losses with Nvidia being the first to take a bow when it put its Icera Baseband Modem manufacturing business up for sale in 2015. Intel soon followed suit when it announced in March that it was abandoning production of its Atom mobile processors.
We are now seeing the first innings of the AI revolution, and both companies are keen to establish early leadership positions in this emerging industry. As far as having a large installed base of AI server chips goes, Intel wins over Nvidia. But that's mainly because companies have traditionally used CPUs in deep learning tasks and have only recently started experimenting more heavily with GPUs. Internet companies have realized how well GPUs can handle AI tasks. GPUs are designed to handle vector and matrix operations in parallel compared to a single core CPU that handles such matrix operations in serial form processing, a single element at a time. Deep learning involves a lot of vector and matrix operations which gives GPUs an edge over CPUs. Although Intel has lately been saying that its integrated GPUs can match the performance of discreet cards, it will take time for these products to gain widespread adoption comparable to GPUs.
Intel has only recently started focusing more deeply on AI whereas Nvidia has been at it for years. Nvidia has built pretty comprehensive programming tools that customers need with AI chips, including its cuDNN deep learning library. Nvidia recently launched a 15-billion transistor to be used in AI applications and Intel responded by saying it will launch its AI chip codenamed Knights Mill early next year. Further, recently Intel bought AI startup Nervana to beef up its AI offerings.
AI driving Nvidia revenues
Although Intel's Xeon Phi processors have helped the company claim a bigger AI data centre share of the market, Nvidia's deeper focus in the field and the fact that cloud and Internet companies are beginning to deploy GPUs in their machine learning servers might help Nvidia gain an upperhand in the coming years. On the other hand, Intel's deeper financial wherewithal gives it a good opportunity to pursue opportunities in the field through R&D as well as through mergers and acquisitions.
But ultimately what matters most for investors is how AI will impact each company's growth. Neither Intel nor Nvidia breaks out its AI/deep learning revenues in its earnings reports. But it's becoming increasingly clear that AI is already having a tangible impact on Nvidia's top line. During the last earnings call, Nvidia said that it's already working with more than 3,500 AI customers across diverse industries, and its sales to cloud customers have been growing at more than 50%. According to a report by MarketsandMarkets, the artificial intelligence market is expected to grow from $419.7M to $5.05B in 2020 and to $10.4B in 2024. So this is a field that's still in its infancy.
Nvidia has ~10% of Intel's revenue base, and AI/deep learning is already beginning to have a tangible impact on the company's growth. Nvidia posted revenue growth of 21% during the last quarter, about double its growth rate a year ago. The company credited the rapidly growing deep learning industry for its renewed growth. So AI is likely to have a much bigger impact on Nvidia's growth over the next five years than it will on Intel's.
Investors should keep a close eye on how these two stalwarts square it off in the AI field in the coming years.