It seems like AMD (NASDAQ:AMD) has been caught in the crossfire in the US-China trade war this time around. A couple of weeks ago, US regulators imposed a ban on AMD and NVIDIA (NVDA), restricting their exports of certain artificial intelligence chips to China. This has fueled all kinds of speculation about the extent of financial and operational impact on AMD. While some bears believe this is a catastrophic hit to the chipmaker, bulls remain unfazed and view this as an insignificant event. So, in this article, I’ll discuss both sides of the story and attempt to explain why AMD won’t be affected much by this ban.
Let me start by saying that US regulators did not impose a blanket ban encompassing all AMD and NVIDIA-branded SKUs. Rather, it’s a very specific ban that restricts China-bound exports for AMD’s MI250 chip, and NVIDIA's A100 and H100 chips. This appears to be a part of the ongoing long-term effort by the US government to limit China’s burgeoning artificial intelligence industry and its potential misuse by the Chinese military.
NVIDIA quickly confirmed in its 8-K filing that this ban would mean the loss of $400 million worth of sales in its third quarter alone. Although the company did not disclose the annual impact, it’s only fair to say that this financial hit is recurring in nature and will be felt in subsequent quarters too.
AMD, on the other hand, disclosed it doesn’t expect material impact on its business. Given the negligible financial hit, investors are now wondering if the chipmaker’s chips are competitive enough in the first place to be sold in high-performance artificial intelligence workloads. The rationale is that, perhaps, nobody wants to buy its datacenter GPUs and that’s why it isn’t severely hit by these export bans. While that’s a legitimate concern, I contend that there are 2 entirely different reasons why AMD’s exposure to Chinese datacenters is limited.
First, the export ban covers only 1 chip (MI250) for AMD, as opposed to 2 for NVIDIA (A100 and H100), which cuts AMD’s exposure relative to NVIDIA. That’s admittedly a simplistic argument, but it’s a tangible contributor, nonetheless. Secondly, and more importantly, AMD is still in the process of becoming competitive in datacenter workloads. For instance, its MI250 Instinct accelerator was released less than a year ago, while NVIDIA's A100 has been available since May 2020.
Enterprises usually take several months and even quarters to qualify new hardware, especially if it’s a new market entrant, so I contend that AMD’s MI250 hasn’t existed in the market long enough to be qualified by enterprises and eventually absorbed in meaningful quantities. Otherwise, if the chip had lackluster performance and was downright incompetent, it wouldn’t be getting adopted in major cloud platforms (like here).
Having said that, a bigger concern at hand is that we don’t know if this is the end of the trade war or if US regulators will be imposing export restrictions on more semiconductor SKUs going forward.
The point that I’m trying to make here is that we just don’t know if the scope of such restrictions will be expanded going forward, or if this is it. This essentially presents a systemic risk factor for all US-based semiconductor manufacturers that have meaningful sales in China, be it NVIDIA, Intel (INTC), or AMD.
More to the point, the sudden ban of AMD and NVIDIA AI chips would surely be a big setback for Chinese datacenters in the short term. However, this newly-created vacuum would also encourage Chinese firms to ramp up investments and create indigenous GPUs for datacenters. This dynamic is already underway (as seen here and here), and it's bound to intensify market competition for AMD and NVIDIA over the longer run.
Having discussed the negatives, let’s now shift attention to some of the positive prospects of AMD’s datacenter division.
See, AMD announced its first datacenter GPU less than 2 years ago and it’s still in the process of becoming competitive in the space. This essentially means the chipmaker doesn’t have much to lose by the recently-imposed export ban in the first place. On the other hand, it has lots to gain in other markets across the globe since its market presence is currently miniscule at best. So, I view this export ban as a non-event for AMD as far as its datacenter prospects go.
Secondly, it’s important to note that AMD has launched several datacenter-focused GPUs within a span of just 2 years and many of its SKUs have turned out to be quite competitive against NVIDIA's finest. The Next Platform benchmarked different datacenter GPUs based on their price-to-performance metrics (lower is better) a few months ago. Note how some of AMD’s latest GPUs stack up well against NVIDIA's well-established A100 card, which is indicative of strong competitive positioning.
To further up the ante, AMD is expected to announce its next-generation MI300 accelerator sometime in 2023. This is expected to be fabricated using Taiwan Semiconductor’s (TSM) latest 5nm process which is bound to result in sizable performance gains over other legacy products that are built using dated processes. So, I contend that AMD’s datacenter push will intensify and continue bearing fruit in the coming quarters.
Having said that, AMD is undertaking a slew of measures to quickly grow its presence in the datacenter space. It completed the acquisition of Xilinx earlier in February, and then bought datacenter accelerator manufacturer, Pensando, later in April. These buyouts have twofold benefits for the chipmaker. For starters, they result in inorganic growth for AMD by expanding its product portfolio and its revenue streams. Secondly, the IP and technological ecosystems that these acquisitions bring along open up new possibilities for integrated product platform launches that eventually stand to expand the total addressable market for AMD and its newly-acquired businesses.
The chipmaker hasn’t disclosed a breakdown of its datacenter revenue coming specifically from exports to China. However, it does report its datacenter revenue overall, which, for the record, amounted to nearly $1.49 billion or 22.7% of the company’s total sales last quarter. This particular revenue stream has grown sequentially in all 5 of the last 5 quarters and I expect it to continue growing rapidly in the foreseeable future as well, due to the above-mentioned growth catalysts.
There’s no denying that the export ban on NVIDIA and AMD stands to shrink their total addressable market in the short term. But as I explained above in the article, this is a non-event for AMD as far as financials are concerned. The chipmaker has already undertaken a slew of measures - spanning from strategic acquisitions and aggressive product rollouts - that’ll more than offset this loss of business opportunity in China over the longer run. So, I remain bullish on AMD and its growth prospects in the datacenter space.
As far as valuations go, AMD’s shares are trading around 5.7 times its trailing twelve-month sales. This is quite low when compared to many of the other rapidly growing semiconductor stocks and suggests that it's undervalued at the time of this writing. Hence, investors with a multi-year time horizon may want to capitalize on the fear that’s currently prevalent on the Street and accumulate AMD's shares on potential price corrections. Good Luck!
This article was written by
Disclosure: I/we have no stock, option or similar derivative position in any of the companies mentioned, and no plans to initiate any such positions within the next 72 hours. 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.