Originally published on February 10, 2016, this post was substantially updated on December 30, 2019.

Things have been heating up in the discrete GPU (graphics processing unit) market between Nvidia (NVDA) and Advanced Micro Devices (AMD). The latest contender is Intel (INTC), which is making a second attempt at producing its own discrete GPU. 

The GPU market has been attracting investor attention since 2016, when Nvidia introduced the concept of GPU computing. It’s leveraging GPUs in PC gaming, crypto mining, autonomous driving, 5G, AI and machine learning computing from data centers to the edge. 

Back in 2015, Nvidia took some major GPU market share away from AMD. However, AMD, in its Q4 of 2015 earnings, stated that it will launch new products that compete with Nvidia and win back its lost market share. And it did just that. 

As you can see in the above graph, AMD started to lose market share in the calendar third quarter of 2014. That’s when Nvidia gained market share. Due to reduced competition, Nvidia managed to boost revenues. However, a year later, in the third quarter of 2015, AMD gained some lost market share in the AIB (add-in-board) segment, while Nvidia lost. According to JPR (Jon Peddie Research), AMD’s market share grew from 18% in Q2 of 2015 to 18.8% in Q3 of 2015. Nvidia’s market share fell from 81.9% to 81.1% during the same period. 

From this point forward, AMD’s market share only rose. It peaked at 36.1% in the second quarter of 2018, according to this JPR report. Meanwhile, Nvidia’s market share fell. It bottomed at 63.9% during this timeframe. 

Both also companies launched new products in that period.

• In January 2016, both companies launched hotfixes for their graphics drivers: Nvidia’s 361.82 update and AMD’s Radeon Software Crimson Edition 16.1.1.
• Both companies are developing new architectures with HBM (high bandwidth memory) 2.0. Nvidia is developing Pascal, and AMD is developing Polaris.
• Both companies also plan to shift GPU production to 14/16-nm (nanometer) node process. They’ve partnered with Samsung and Global Foundries for these efforts.

AMD also launched the industry’s first hardware-virtualized GPU, the FirePro S-Series. However, the bulk of AMD’s market share gain goes to its 14nm Polaris GPU. It proved to be a big hit in the mass market.

Although AMD’s Polaris didn’t beat Nvidia’s 14nm Pascal in performance and power consumption, it offered decent performance at a lower price. And it’s important to note that JPR measures market share in terms of number of GPUs shipped.

As AMD targeted the masses instead of enthusiasts, its volume numbers increased faster than its revenue. But in Nvidia’s case, its premium-priced GPUs attracted gaming enthusiasts. So Nvidia saw higher revenue and earnings growth than AMD. 

Another major reason for AMD’s market share gain was the cryptocurrency boom in late 2017 and mid-2018. Crypto miners preferred AMD GPUs over Nvidia’s for mining currency since they offered a better price-to-performance ratio.

And when the crypto boom was at its peak in the second quarter of 2018, AMD’s market share reached 36.1%. This entire crypto-related rise fell in the second half of 2018. AMD’s market share bottomed to the pre-crypto level of 18.8% in Q4 2018. Note that we took all this market share data from JPR. 

After the crypto bubble burst in the third quarter of 2018, Nvidia started to reclaim its GPU market share in the gaming space. However, its next-generation 12nm Turing GPUs (the generation after the 14nm Pascal launched in 2016) faced several hurdles.

Nvidia introduced a ten-year advanced “real-time ray tracing” technology in the Turing GPU. As this technology is ten years ahead of its time, it lacked the right ecosystem to leverage the benefits of ray-tracing. For instance, the current AAA games didn’t support ray tracing. So this disadvantage hurt Nvidia’s GPU market share in the first half of 2019. 

With time, the adoption of Turing grew—and so did Nvidia’s market share. In the third quarter of 2019, its discrete GPU market share rose to 72.9% from 67.9% in the previous quarter, according to JPR. This market share is growing, thanks to ray-tracing. But it will once again face a major challenge in 2020. In the coming year, AMD plans to launch its second-generation RDNA2 (Radeon DNA)-based 7nm GPU that features ray tracing.

And Nvidia’s troubles won’t end with AMD. Intel is making a second attempt at discrete GPU. And this time, it seems the chip giant will crack the code. 

Intel enters the gaming GPU space

Intel first announced its discrete GPU at the 2016 CES (Consumer Electronics Show). There, representatives stated that it’s developing a gaming NUC (Next Unit of Computing) that will compete with AMD’s and Nvidia’s GPUs. The NUC will be powered with a quad-core sixth-generation Skylake Core i7 and Iris Pro 580 iGPU (integrated graphics processing unit).

Fast-forward to October 2019. Intel’s chief architect, Raja Koduri, teased the launch of the company’s first Xe GPU in mid-2020. Many rumors have circulated about which form the first Xe GPU will take—mobile gaming GPU or desktop gaming GPU. Only time will tell if the second time’s a charm for Intel in the GPU space. 

Competition is part and parcel of business, and it will always affect Nvidia’s earnings—especially its gaming revenue. However, it doesn’t stop Nvidia from diversifying into other areas. Autonomous driving, AI from data center to edge, and 5G are all new frontiers for the company. So let’s explore the impact of growing competition, market trends, and technology transition on Nvidia’s earnings. 

2016 was a revolutionary year for both AMD and Nvidia. Even though Nvidia started losing GPU market share to AMD in 2016, the former’s revenue continued to grow. Nvidia’s earnings grew by strong double digits between 2016 and 2018. And NVDA outperformed peers AMD and Intel by a huge margin. Despite rising competition from AMD, market trends (like sports and crypto) and technology transitions (like AI and rendering) drove Nvidia’s earnings to new highs. 

Nvidia’s revenue and EPS hit a new high in fiscal 2019

Nvidia saw exponential growth between fiscal 2017 and 2019.

• Its revenue more than doubled from $5 billion to $11.7 billion. 
• Its non-GAAP EPS rose fourfold from $1.67 to $6.64.

As a seasonal trend, the company tends to see strong growth in the second half of a fiscal year versus the first half. And in fiscal 2016’s third quarter, when the Pascal launched, Nvidia reported revenue of $1.3 billion. This number was a record high at the time. And since then, the GPU giant has hit new highs.

Its revenue peaked in the third quarter of fiscal 2019 to $3.18 billion. In those three years, Nvidia earned its first-year growth from the eSports trend, its second-year growth from eSports and AI trends, and its third-year growth from crypto and AI trends. 

As for EPS, they grew from $0.46 in the third quarter of fiscal 2016 to $1.84 in the third quarter of fiscal 2019. Unlike revenue, Nvidia’s quarterly EPS peaked $2.05 in the first quarter of fiscal 2019. Its EPS are a function of a richer product mix and improved operating efficiency. As the company ruled the high-end GPU market, its EPS grew faster than its revenue. Moreover, it leveraged its one GPU architecture across the gaming, data center, automotive, and professional visualization spaces. 

Overall, fiscal 2020 has been a weak year for the semiconductor industry. The trade ban and tariffs in the US-China trade war directly affected semiconductor sales. And Nvidia was no exception. So the company refrained from sharing full-year guidance at the start of the year.

But as we near the end of 2019, Nvidia is guiding for fiscal 2020 revenue of $10.8 billion, down 8% year-over-year. It also expects its non-GAAP EPS to fall 26% to $4.9.

However, the worst is over for Nvidia. And the company is set for a strong fiscal 2021. Its key growth drivers will play a major role in driving fiscal 2021 revenue. 

Nvidia earns revenue through five segments—but not all its segments drive growth. Its fiscal 2016 earnings got a boost from the gaming, automotive, and data center applications. However, the company’s other segments—professional visualization, and PC, and Tegra OEM (other equipment manufacturers)—have been slowing.

So let’s look now at each of these five segments. We’ll unpack the growth drivers and deterrents for each one. 

Gaming remains Nvidia’s core and its biggest business. The segment has accounted for more than 55% of its revenue since fiscal 2016. And back in fiscal 2016, Gaming was its second-fastest growing segment with 37% year-over-year growth.

Nvidia sells GPUs to both gamers and PC OEMs (original equipment manufacturers). And it reports these sales in its Gaming segment. Back in fiscal 2016, its revenue from PC OEMs was slowing as overall PC demand was falling. According to Gartner’s preliminary results, worldwide PC shipments fell 8.3% year-over-year in the calendar fourth quarter of 2015. This period falls under Nvidia’s fourth-quarter of fiscal 2016, ending in January of 2016. 

In fiscal 2017, which ended January 2017, AMD won on the volume front. But Nvidia won on the ASP (average selling price) front. Gamers widely adopted Nvidia’s high-end Pascal-based GeForce GPUs as the trend of eSports picked up. eSports have become so popular that today, official eSports tournaments take place. And they attract millions of viewers.

It was in fiscal 2017 that Nvidia entered the game console market. Its Tegra processors powered the Nintendo Switch, which turned out to be a huge success. And the eSports trend boosted demand for gaming notebooks and desktops, driving Nvidia’s revenue from PC OEMs. 

eSports and the Nintendo Switch game console’s popularity drove Nvidia’s fiscal 2018 revenue as well. But the year also saw a significant growth driver—the crypto trend. The cryptocurrency Ethereum’s prices were rising significantly, which encouraged many people to switch to mining Ethereum. For that purpose, miners needed GPU. And although AMD’s GPUs were preferable, a supply shortage pulled that demand to Nvidia GPUs as well. The whole wave of crypto mining drove revenue for all GPU players riding the tide. 

However, the growth rate for the company’s gaming revenue slowed to 13% year-over-year in fiscal 2019. The causes were the end of the crypto trend, the slow uptake of Turing, and the onset of the US–China trade war. Over the years, Nvidia increased its revenue from China to 24% in fiscal 2019 from 16% in fiscal 2016. So the 2019 trade disputes between the US and China hurt Nvidia significantly. 

The company’s fiscal 2020 earnings will see Gaming revenue bottom because of these headwinds. I expect Nvidia’s gaming revenue to fall 7% year-over-year to $5.8 billion in fiscal 2020, slightly above the pre-crypto level of $5.5 billion. 

However, I think the growth will pick up again in fiscal 2021 as the trade war eases, the crypto effect vanishes, and Nvidia launches its 7nm Ampere GPUs to compete with AMD’s 7nm Navi GPUs. Nvidia’s next big bet is on cloud-based gaming streaming through GeForce Now. This initiative is still in beta, and it could start generating revenue next year. 

Also, AMD has an advantage over Nvidia next year. AMD Navi GPUs will power Sony’s and Microsoft’s next-generation game consoles. And unlike Nvidia, AMD is a market leader in console chips. 2020 will be the first year of next-generation consoles, so they will sell well.

Moreover, AMD is launching new products to regain its lost market share. This effort could slow revenue growth for Nvidia in the gaming space. Alarmed by the increasing competition from AMD, Nvidia has been diversifying into other high-growth areas, such as data center and automotive.

While Gaming remains the biggest revenue generator for Nvidia, Datacenter has become its fastest-growing segment. From contributing just 6.7% toward Nvidia’s revenue in fiscal 2016, Datacenter rose to contribute 25% in fiscal 2019. This jump came as Datacenter revenue rose 145%, 133%, and 50% in fiscal 2017, 2018, and 2019. The driver of this exponential growth is Nvidia’s three generations of Tesla data center GPUs: the Pascal, Volta, and Turing. 

In fiscal 2017, Nvidia’s data center revenue rose as major cloud companies adopted Nvidia’s Pascal-based Tesla P100 GPUs to bring deep learning capabilities. The company also introduced the TensorRT deep learning framework. This launch took place when the AI revolution just began, and Nvidia was the sole provider of GPU accelerators for data centers. Every type of data center—from supercomputers, hyperscalers, and cloud companies to enterprises across various industries—has adopted deep learning and inferencing. 

Looking at the growing computing needs of its data center customers, Nvidia launched its Volta GPU architecture dedicated to data center. Inside Volta, Nvidia for the first time introduced Tensor cores dedicated to deep learning. Also, it leveraged the Volta architecture by launching a complete suite of AI products. These launches included Tesla V100 GPUs for DLT (deep learning training), the DGX-1 and DGX-2 Supercomputer, the consumer-grade Titan V card for PCs, Nvidia GPU Cloud, and GRID for GPU on cloud. 

There was no looking back from this point. Nvidia became a monopoly in AI hardware, and it attracted competition from Intel and AMD. In fiscal 2019, Nvidia’s Datacenter revenue growth slowed to 52%. The onset of the US–China trade war slowed purchases from data center customers. 

I expect Nvidia’s Datacenter segment to bottom in fiscal 2020. Datacenter revenue should fall 8% year-over-year to $2.7 billion. Also, I believe growth will likely revive next year. But it may not reach the triple-digit level while AMD and Intel make their way into deep learning and AI.

I think AMD will tap AI by fully optimizing its EYPC CPU and Radeon Instinct GPU for supercomputing. Intel has a broader range of accelerators, like Xeon CPUs, Nervana neural network processor, Altera programmable solutions, and many more. 

Automotive

While Gaming is Nvidia’s largest segment and Datacenter is its fastest-growth segment, Automotive is the company’s future. Back in fiscal 2016, Automotive was NVDA’s fastest-growing segment with 80% year-over-year revenue growth. It accounted for only 6% of the company’s revenue. This growth came as the overall automotive market performed well. Other semiconductor companies, such as NXP Semiconductors and Maxim Integrated, reported strong growth in their automotive segments. 

And back in fiscal 2016, Nvidia’s automotive revenue largely came from in-car infotainment. The trend of autonomous vehicles was just starting. Over the years, the company developed its DRIVE AGX processor, shifting its focus from in-car infotainment to AV. Today, it has the leading AV technology with end-to-end AV solutions. Moreover, Nvidia has branched out to trucks and commercial vehicles as well. So a slowdown in the car market didn’t put Nvidia’s Automotive revenue growth in the red. In fiscal 2019, its Automotive revenue grew 15%. And in fiscal 2020, it expects this number to grow 7.8% on a year-over-year basis. 

So Nvidia’s automotive revenue has doubled from $320 million in fiscal 2016 to $640 million in fiscal 2019. But it still accounts for 5% to 6% of the company’s revenue because the autonomous vehicle trend hasn’t picked up yet. Fiscal 2020 saw the advent of autonomous driving. Many third-party research reports expect autonomous vehicles to hit the market in calendar 2020 and 2021, after 5G infrastructure is in place. Autonomous vehicles need the 5G network to connect to data centers with low latency. 

The autonomous vehicle opportunity

Automotive could be Nvidia’s next revenue catalyst. However, Nvidia won’t be the sole company enjoying this opportunity. Realizing its AI mistake, Intel jumped early in the AV market by acquiring Mobileye in 2018. Intel is still behind Nvidia in terms of technology—but not as far behind as it is with deep learning. 

Nvidia believes autonomous vehicles could be a $10 trillion industry, including self-driving cars, trucks, and mobility services like robo-taxis and delivery vans. In this $10 trillion industry, Nvidia estimates its TAM (total addressable market) to reach $60 billion by 2035. Meanwhile, Intel expects its TAM to reach $100 billion by 2030.

Professional Visualization

As we’ve seen, then, Gaming and Automotive maintained their revenue contribution above 55% and around 5%–6% between fiscal 2016 and 2019. Datacenter exploded, increasing its revenue contribution from 6.7% to 25% of Nvidia’s revenue during that period. But one segment that showed a reverse trend was Professional Visualization or ProV, which reduced its revenue contribution from 15% to 9.6%. This drop was due to slow growth in the workstation market. 

This trend doesn’t mean that the ProV segment didn’t grow. It grew 11% year-over-year in fiscal 2017 and 2018. But the real growth came in fiscal 2019, when the segment’s revenue grew 21% as Nvidia launched ray-tracing technology. Nvidia’s first ray-tracing Turing product was a Quadro GPU for the workstation. The ProV segment has growth potential. But this potential isn’t as high as the automotive or datacenter segments. 

As more developers and designers adopt VR/AR (virtual/augmented reality) and ray tracing technology, the demand for Nvidia’s Quadro will rise. However, AMD will continue to offer stiff competition in the workstation GPU market by providing better value for customers’ money. 

Let’s now look at the segments that are slowing down Nvidia’s revenue growth. 

Over the years, Nvidia broadened its GPU offering to different platforms of gaming, workstations, data center, AI, and cloud computing. However, not all segments drove growth for the company. The Tegra microprocessor for smartphones was the low-hanging fruit on Nvidia’s product tree. And the Tegra processor business has been facing falling revenues due to reduced demand from smartphones and tablets. Fierce competition from Qualcomm (QCOM) also reduced Tegra smartphone processor sales. So the GPU giant moved away from markets where its returns weren’t satisfactory.

Why did Nvidia exit the smartphone market?

It all started when AI was just a theory. The early 2010 decade saw the evolution of the smartphone that replaced the PC for most people connecting to the Internet. Qualcomm (QCOM) was the leader in the smartphone chip market. As the smartphone also involved computing, the two computing giants, Intel and Nvidia, jumped into the smartphone arena.

Also, smartphones need low-power computing and connectivity chips. Nvidia acquired baseband chipmaker Icera in 2011 and combined it with its Tegra processor. However, Nvidia’s Tegra smartphone processor failed to compete with Qualcomm’s ARM-based Snapdragon processor. 

2015 was a challenging year for the smartphone industry. Even Intel and Qualcomm took a hit as smartphone demand slowed in 2015, especially in China. So, in 2015, Nvidia exited the smartphone market. Intel followed Nvidia and exited the smartphone processor market as its Atom chips failed to beat Qualcomm chips in power efficiency. Later, in 2019, Intel exited the 5G smartphone modem market as well, thanks to Qualcomm. 

According to a June 2016 Ubergizmo article, Nvidia’s CEO Jen-Hsun Huang said at Computex 2016, “We are no longer interested in that market.” Explaining Nvidia’s decision, he stated that smartphones are primarily communications devices where compute performance plays a “supportive role.” Nvidia, on the other hand, is best at computing. And it can grow in areas where compute performance is a challenge.

Moreover, Nvidia identified far better opportunities in eSports and AI, where its mastery lies. So exiting smartphones was a wise decision for Nvidia—even though smartphone processors were a huge market. 

So on the one hand, Nvidia has exited the smartphone market. And on the other hand, AMD is entering the smartphone market. However, AMD’s strategy is way different than Nvidia’s. This year, it signed a licensing deal with Samsung for GPU technology in smartphones.

Nvidia broadens Tegra’s reach with new platforms

With Tegra smartphone processors gone, Nvidia looked for alternative platforms where there’s a need for computing at the edge. In fiscal 2016, Tegra only had two platforms: SHIELD for smart home entertainment and DRIVE for smart cars.

Over the years, it added custom Tegra processors for Nintendo game consoles, Jetson AGX for robotics, embedded devices like smart cameras, and Clara AGX for smart medical instruments. Jetson and Clara were added to Tegra as early as fiscal 2019. Game consoles were added in fiscal 2017.

Nvidia’s Tegra revenue rose almost threefold from $559 million in fiscal 2016 to $1.54 billion in fiscal 2019. The majority of this growth came from the Nintendo Switch, as Tegra revenue rose 47% and 86% on a year-over-year basis in fiscal 2017 and 2018. Although Nvidia tried to keep itself away from the low-margin game console market, the segment did contribute a significant amount to its revenue. A June 2018 Motley Fool article did the math and concluded that game consoles contributed 18% to Nvidia’s revenue in fiscal 2018. 

Most technology firms spend a large sum on R&D to build a rich IP portfolio. They either use the IP to develop products themselves or license the IP to others in return for royalties. ARM Holdings only earns through licensing. Qualcomm earns from both selling chips and licensing IP. Nvidia and AMD are mainly involved in developing GPUs. However, they continuously search for licensing opportunities since they generate stable cash flows. 

Nvidia has been making GPUs for more than 20 years. And it also holds 7,000 patents. As the use of GPUs increases in various industries, the company is looking to license its technology in return for a fee.

Back in fiscal 2016, Nvidia earned 5% of its revenue—or $264 million—from licensing. This revenue was from a five-year licensing deal Nvidia had with Intel. However, this deal came to an end in the first quarter of fiscal 2018. Rumors suggested that Intel might renew the deal or sign another GPU licensing deal with AMD. Neither of the two scenarios happened. Instead, Intel poached AMD GPU head Raja Koduri in November 2017 to build its own discrete GPU. And this GPU is set to launch next year. 

As for Nvidia, it lost its licensing revenue. And it also entered into a costly legal battle with Samsung and Qualcomm, which it lost. Currently, Nvidia’s licensing business is dead. Nvidia management is also not focusing much on licensing. Instead, it’s exploring new avenues in AI and AV. 

Weak semiconductor market ends Nvidia stock’s two-year-strong rally

So far, we saw that Nvidia reported exponential growth in fiscal 2017 and 2018 and strong growth in fiscal 2019. Nvidia’s fiscal year ends in January of that year. So its fiscal 2017 is close to calendar 2016. And as you can see from the above graph, Nvidia’s growth outperformed the global semiconductor market’s revenue growth between 2016 and 2018. Moreover, it’s expected to outperform the market, even in 2019. 

The WSTS (World Semiconductor Trade Statistics) expects global semiconductor revenue to fall 12.8% in 2019. Meanwhile, Nvidia expects its fiscal 2020 revenue to fall 8.2%. On the other hand, rivals Intel and AMD expect full-year 2019 revenue to rise 5% and 0.3%, respectively. This difference shows that peers performed better than Nvidia this year. And this same trend reflects in its stock price. 

Nvidia had a market capitalization of $14.7 billion on January 4, 2016. This market cap rose almost tenfold to $144.96 billion on December 27, 2019. In the first two years, Nvidia stock rose 225% and 81%, respectively, on the back of strong revenue. On the other hand, AMD stock rose 295% and -9% while Intel stock rose 8% and 30%. 

However, Nvidia stock fell 31% in 2018. And this entire fall came in the last quarter as the crypto trend vanished and Nvidia guided its first quarterly revenue decline. After peaking at $281 in September 2018, the stock fell more 50% between October to December 24, 2018. This was the time that the entire stock market collapsed, and tech stocks took the biggest hit because of the onset of the trade war. Nvidia was among the top decliners, falling 31%. AMD and Intel rose 80% and 4%, respectively. 

The Nvidia’s fall was so deep that the stock faced technical weakness until August 2019. In the second half, Nvidia stock rallied at a decent pace and pulled itself from the technical weakness. It rose 74% year-to-date on the back of demand recovery. This growth is lower than AMD’s rally of 145% but higher than Intel’s rally of 27.6%. 

Nvidia stock is currently trading at $236.87, which is near a 52-week high of $241.8. AMD is trading at an all-time high of over $46, and Intel at a 19-year high of over $60. Next year, I believe Nvidia stock can breach its all-time high of $281, which represents a 19% rise from the current trading price. 

All three stocks are ending the year near their 52-week highs. The general rule for stocks is to buy on the dip and sell on the rise. And Wall Street analysts have a “hold” recommendation for AMD and Intel, which are trading near their historic highs. So it’s not a good time to buy these stocks at the moment since it’s difficult to say how much they’ll rally. The stocks are already trading at their median price target. AMD and Intel have bullish price targets of $53 and $75, representing upside of 15% and 25%. 

This outlook isn’t the case for Nvidia, as it’s far from its all-time high. Wall Street analysts have a “buy” recommendation with a bullish target of $275 for NVDA stock, which is still lower than its all-time high of $281. As Nvidia stock has a higher possibility of growth, it’s a “buy” even at its 52-week high. 

Nvidia, AMD, and Intel are all growing stocks. Of course, one way to benefit from their rally is by investing in each of these stocks. But another way is to get exposure to these stocks through the VanEck Vectors Semiconductor ETF (SMH), which has risen 62% year-to-date.

These three stocks are among the top ten holdings of SMH. The fund invests 11.8% of its portfolio in Intel, 5.8% in Nvidia, and 4.8% in AMD.