Nvidia’s Fiscal 2019 Q3 Earnings: Little Impact From Crypto
Why the crypto decline doesn't hurt Nvidia.
What AMD's and Intel's Q3 results portend for Nvidia.
What AMD's Next Horizon event portends for Nvidia.
This idea was discussed in more depth with members of my private investing community, Rethink Technology. Start your free trial today »
Rethink Technology business briefs for November 9, 2018.
Why the crypto decline doesn't hurt Nvidia
Yesterday, Morgan Stanley analyst Joseph Moore is reported to have lowered his price target for Nvidia (NVDA) in part because of declining crypto mining demand. I'm reminded of the Nvidia short thesis of SA contributor Akram's Razor, who posited a severe impact to Nvidia's Gaming segment revenues due to a drop in demand for graphics cards for crypto currency mining.
The primary problem with this thesis is the assumption that crypto mining sales ever constituted a significant portion of the Gaming segment. Nvidia and its add-in board partners did a very good job of market segmentation with respect to crypto by creating a line of GPU add-in boards tailored for crypto mining. These cards lacked video outputs, and so could not be repurposed as PC video cards. This made them less expensive, but also ensured that a subsequent glut of Nvidia mining cards on the used market would have no impact on gaming card sales.
The efficacy of this strategy was apparent in Nvidia's fiscal Q2 results. Nvidia books the mining specific graphics card sales in its OEM market segment. OEM suffered a 54% y/y decline and a 70% sequential decline, clearly showing the impact of the crypto slowdown. The revenue delta was $271 million q/q.
However, the Gaming segment was up 52% y/y to $1.805 billion. There may have been some crypto impact, but it's probably in the noise. The thesis that a crypto downturn would hurt gaming GPU revenue simply had no substance. Despite the so-called crypto headwinds, Nvidia's overall GPU revenue grew by 40% y/y as did its total revenue.
What AMD's and Intel's results portend for Nvidia
At this point, Nvidia has probably worked through whatever impact there was going to be due to crypto. However, the statement during AMD's (AMD) Q3 conference call by CEO Lisa Su regarding the impact of crypto seems to have reignited fears:
In graphics, the year-over-year revenue decrease was primarily driven by significantly lower channel GPU sales, partially offset by improved OEM and datacenter GPU sales. Channel GPU sales came in lower than expected, based on excess channel inventory levels, caused by the decline in blockchain-related demand that was so strong earlier in the year.
There seems to be an assumption that what hurt AMD must also hurt Nvidia, but I doubt that's the case. Nvidia seems to have pursued the crypto specific card approach more aggressively than AMD's board partners. In addition, the most popular Nvidia general purpose cards used for mining were low-end GTX 1050 and 1060 series cards. Nvidia's gaming business is mostly focused on the premium end of the price and capability spectrum.
Nvidia's release of its next generation Turing gaming cards should enhance this effect and provide additional protection. Turing architecture cards provide new capabilities in the form of real-time ray tracing and 4K upscaling that are not present in any other AMD or Nvidia cards.
Users are still waiting to exploit those features, but Nvidia's RTX series cards seem to be selling well. Third party add-in board prices are running above Nvidia's RTX Founders Edition cards.
Based on all these factors, my conclusion is that whatever caused AMD's add-in board decline, it was an AMD specific problem. If anything, I believe AMD's results highlight its lack of competitiveness with Nvidia in high end gaming, especially with the advent of Turing.
On the other hand, Intel's reported growth in the PC market for Q3 bodes well for continued strong demand for Nvidia's gaming oriented graphics cards. Intel pointed to both consumer and enterprise demand driving an enlarged PC TAM. On the consumer side, demand is driven almost exclusively by PC gaming, since there really isn't a reason to own a PC otherwise.
This was especially true for Intel this quarter as it deliberately focused on higher end PCs with its Core series processors. This probably maximized Intel's profits while leaving the low end to AMD. The strong growth that Intel saw in this segment indicates that Nvidia also will see continued strong growth in PC gaming demand.
Likewise, Intel's strong datacenter growth indicates that hyperscale data centers and cloud providers continue to spend strongly on capital equipment. This willingness to continue to invest in infrastructure also indicates that there will be continued strong demand for Nvidia's datacenter offerings in the long run.
In the short term, demand may fall off as Nvidia's key datacenter product, the Volta V100, reaches the end of its sales cycle. With AMD about to launch a 7 nm Vega GPU for the datacenter segment, and TSMC (TSM) ramping up its 7 nm process, it's reasonable to expect that Nvidia will field a 7 nm replacement for Volta in early 2019.
This will be of especial interest to hyperscale users for whom energy consumption is an important contributor to total cost of ownership. With a Volta successor possibly a few months away, I expect datacenters to curtail purchases of Volta. Nvidia's Datacenter growth, which had been so strong early in the year, may have slowed some more in fiscal Q3
I wouldn't be surprised if analysts and tech pundits declare this a disaster for Nvidia, focused as they are on the near term. However, I don't expect this to be a lasting effect. Nvidia's management have proven they are capable of very long range planning with Turing. Undoubtedly, the successor to Volta has been in the works for some time, and only awaits 7 nm costs to decline at TSMC before Volta's successor enters production.
What AMD's Next Horizon event portends for Nvidia
Prior to AMD's Next Horizon event on November 6, there had been considerable speculation that AMD would unveil new 7 nm consumer CPUs and GPUs, but the event mostly focused on the key 7 nm products that we already knew were on the way, the 7 nm port of the Vega GPU and the new 7 nm EPYC 2.
CTO Mark Papermaster did provide some deep background into Zen 2 architecture that's relevant to both consumer and server CPUs, but that was the extent of it for the consumer side. Most of the rest of the day was focused on the 7 nm datacenter products.
In terms of specific availability, we heard nothing about specific 7 nm consumer CPUs or GPUs. Instead we were told that 7nm Vega would be available in Q4 of this year, and EPYC 2 would be available some time in 2019.
I'm sure that wasn't an oversight, and Su plans on unveiling the consumer products during her CES keynote in January. However, even this announcement schedule is revealing. AMD isn't shy about previewing its products. Mark Papermaster previewed the Zen CPU architecture at AMD's May 2015 Financial Analyst Day, and what he showed at the Next Horizon event was about equivalent to the May 2015 presentation.
Does that mean that Zen 2 is more than a year away from consumer availability? Could be. And this is really not AMD's fault, but rather the result of Global Foundries' broken promises for 7 nm.
The reader may recall that in August 2016, GloFo announced that it was going to “skip” the 10 nm node and proceed directly to 7 nm. This decision always was fraught with risk. If any of the major manufacturers were going to skip a node, GlobalFoundries was the least likely to succeed. It was by far the weakest in process technology in 2016, having only recently achieved 14 nm FinFET production with the help of licensed technology from Samsung (OTC:SSNLF), after giving up on its internally developed 14 nm process.
The decision would prove fateful for GloFo and sow the seeds of its disastrous decision to abandon its 7 nm efforts (and efforts for any succeeding nodes), announced last September. Naturally, everyone, including AMD, tried to put the best possible face on the decision, but the consequences of GloFo's failure are still being felt by AMD. This is apparent in the announced timelines for AMD's datacenter products.
It's apparent that Vega 7 nm always was in the works at TSMC, which is why it will be available by the end of the year. AMD probably felt that it had to get something more competitive with Volta to datacenter users by the end of the year and couldn't wait for GloFo to be ready.
Probably, AMD was still assuming that EPYC 2 would go to GloFo. Recall that Global Foundries CTO Gary Patton claimed as recently as May 31 that the 7 nm process was “on track” for high volume manufacturing for the second half of this year.
That AMD was probably working toward GloFo production of EPYC 2 is the reason why we don't have a firm timeline for EPYC 2 availability. AMD is starting from scratch on the tape out for TSMC, and tape outs take a lot of time.
The situation is probably even worse in the consumer GPU space. Navi also was probably going to GloFo, so once again, AMD has to start from scratch on a tape out for TSMC.
Bottom line is that because of Global Foundries, AMD's 7 nm consumer products will not appear until late in 2019, if then. This will allow Nvidia to reign supreme in the gaming GPU space with Turing. And by the time that AMD does manage to get a 7 nm consumer GPU out the door, there will undoubtedly be a 7 nm successor to Turing.
And what of AMD's 7 nm Vega? AMD claims that it offers the highest performance of any PCIE GPU accelerator card in double precision floating point math. This is true. The Instinct MI60 offers 7.4 TFLOPs (10^12 floating point operations/second) of performance compared to the PCIE version of the Volta based Tesla V100 at 7 TFLOPs. However, the Tesla is still more energy efficient at 28 GFLOPs (10^9 FLOPS) per watt than the Instinct MI60 at 24.7 GFLOPs per watt. Despite the port of Vega to 7 nm, Nvidia continues to have an architectural advantage in energy efficiency, which is the important metric as far as the datacenters are concerned.
Despite the general negativity about semiconductor stocks this earnings season, I remain very bullish about Nvidia. Mainly this is due to the very weak competition offered by AMD. For the past year, AMD has been unable to unveil a new GPU architecture or substantially new GPU products. Nvidia has been able to introduce a revolutionary new GPU architecture and has brought that to the consumer and datacenter markets.
2019 is shaping up to be not much different. AMD has thrown up a Hail Mary with 7 nm Vega that I doubt will impress any hyperscale users. It lacks Nvidia's Tensor Core hardware acceleration of machine learning tasks. Navi is unlikely to appear before the end of the year.
Even if AMD pulls off a miracle by getting Navi to market sooner, Nvidia won't be standing still. Nvidia has the money and resources to counter any AMD threat, either in the datacenter or in gaming PCs. Nvidia has probably already finished its tape outs for its next generation of 7 nm GPUs and is only waiting for the most opportune time to unleash them.
I remain long Nvidia and have taken advantage of the dip to add to my position. I continue to rate Nvidia an exceptional buy.
Rethink Technology is gaining a reputation as the best technology company analysis available. Consider joining today.
Disclosure: I am/we are long NVDA, TSM.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Editor's Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.