Last week, Jon Peddie Research delivered its Q2 2018 update for total GPU shipments. The data set shows the cryptocurrency boom well and truly over, with desktop GPU shipments falling 27.96 percent compared with Q1 2018. AMD shipped substantially fewer units, with a decrease of 12.28 percent, compared with Nvidia’s decline of 7.49 percent. Intel, in contrast, picked up 3 percent of market share. During the cryptocurrency boom, AMD was supposedly the preferred solution provider over Nvidia and its GPUs suffered higher amounts of price inflation, so the larger decline in AMD shipments makes sense for this reason alone.
Let’s break it down by company. As you read this, keep in mind that consumer market share, total market volume, and average selling prices all play a part in company earnings, and market share alone doesn’t tell us where each company held sway or how much they were earning off their GPU businesses.
AMD’s best years are Q2 2010 to Q2 2012. The HD 5000, 6000, and 7000 families launched in this era in the discrete market, with the HD 5000 family having a particular advantage over Nvidia’s products at the time. AMD doesn’t receive the same bump in market share from launching APUs that Intel did, though there are various reasons for that. Relevant intervening variables include the company’s weaker CPU performance at the time, the fact that AMD often shipped dGPUs in mobile SKUs already, and the possibility that AMD could have simultaneously gained market share at the low end from its APU business while losing market share at the high end. We’d need to see market share broken down by price bands to distinguish these effects.
But the large-scale trend is clear: AMD’s market share starts to slip badly after the first cryptocurrency boom, when its GPU prices — but not Nvidia’s — were driven into the stratosphere. We predicted this could happen and the share data proves it. Despite competing well against Nvidia in terms of MSRP and overall performance, the one-sided price skewing (or skewering) made it impossible for gamers to afford AMD cards. AMD’s market share starts to drop in Q3 2013 and it doesn’t stop until the company bottomed out in Q2 2015, having lost more than half its graphics market in less than two years.
From there, things improve, albeit modestly. Polaris helped the company retake some market share, but we don’t see any kind of clearly discernible impact from launches like Ryzen Mobile. The recent cryptocurrency boom, however, could hide the impact of any potential increases in that category, and it may be some additional quarters before we can tell if increased Ryzen Mobile shipments have resulted in net improvements to overall market share. In Q1 2018, AMD hit a four-year record for market share, at 14.9 percent. But with its current market share in Q2 2018 set at 13 percent according to JPR, it’s clear those gains were temporary. So far, AMD appears to be focused more on the semicustom and console markets for 2019 and beyond than it is in consumer graphics.
Finally, we come to Nvidia. Of the three companies, none has transformed its core business more than Nvidia since 2009 — which isn’t particularly surprising, considering how many articles were being written in 2009 – 2011 about how CPUs with integrated graphics could spell the doom of Team Green. Tesla and Tegra were both originally a bid to differentiate Nvidia’s business away from strict graphics, and the company now competes in realms like HPC and self-driving cars where they had only a small presence in 2009. Of the three companies, this graph does the poorest job of capturing Nvidia’s “story” — and we can actually prove it, via reference to another graph.
Look at the tiny bar of red just above the “50” line for Q1 2017. Now, eyeball the same line in Q2 2018. In 2014, Nvidia reported revenue of $4.13B. In its fiscal year 2018 (not the same as the calendar year), it reported revenues of $9.71B. That’s a 2.35x increase at the same time that the company’s share of the GPU market has changed from 15 percent of the market to 18.4 percent. It’s not that JPR’s data is wrong — I’m confident in the reporting — but that measuring just the consumer GPU market can imply (or fail to imply) incorrect shifts in earnings.
What’s striking about each firm is the degree to which they’ve carved out their own sustainable and profitable niches in graphics. Intel has its own plans for discrete GPUs arriving in 2020 and has seized significant amounts of market share compared with 2009. And its integrated graphics are also much better than they were, even if its onboard solution doesn’t compete particularly well against equivalent integrated solutions from AMD.
AMD’s console business isn’t captured in these graphs, but its earned billions for the firm since 2013 and its graphics business was central to those designs. Nvidia’s GPU market share hasn’t really changed much in recent years, but the company has taken its consumer graphics business and transformed it into everything from an HPC play to an automotive solution. Each firm has taken a different route to success and the degree of success they’ve enjoyed has been different, but all three companies have a thriving graphics business.