Applied Micro (NASDAQ:AMCC) has done an absolutely phenomenal job spinning its X-Gene story to investors and analysts alike for the last year or so. Despite the fact that this product is likely to be uncompetitive and is woefully behind schedule, and despite the grandiose revenue ramp promises (multiple years out, of course) the company's X-Gene microserver processor venture is unlikely to be particularly profitable, investors have continued to buy this company on the X-Gene "dream".
However, this quarter investor and analysts alike were served a very harsh dose of reality in the form of a lackluster forward guide and more vague promises and claims with respect to the company's upcoming X-Gene product line. The sell-off in shares of this name have likely only just begun, with further steep declines as the "hype balloon" has finally begun to deflate in the face of reality.
Q4 Top Line Miss, Analysts Withold The Kudos
A general rule of thumb for trying to determine the "success" of a given quarter/guidance is to gauge the analyst reactions. In particular, the sell-side crowd covering Applied Micro has largely been gushingly bullish despite the company's continued massive GAAP net losses ($0.45/share during the quarter alone) thanks to both the X-Gene rollout as well as increased carrier spend (the business in which Applied Micro actually sells real, non-vaporware products). As the shares continued to rise, the sell-side simply raised its targets ahead of the report (as the shares had exceeded most of the price targets in place).
While the company actually "beat" the top line consensus of ~$55M by $0.4M, the guidance range of $53M - $57M fell SIGNIFICANTLY short of the sell-side range of $55M - $61M. Curiously enough, the current revised range stands at $55M - $58M with consensus at $56M against a midpoint of $55M. This is probably fair given that Applied has been able to slightly beat the midpoint of its guide over the last several quarters without too much issue.
The most telling part of the call, however, was that the sell-side largely withheld the "kudos" that they've been prefacing their questions with. Indeed, we even had the analysts referring to some of Applied's potential competitors in the micro-server chip space such as Cavium Networks (NASDAQ:CAVM) and Advanced Micro Devices (NASDAQ:AMD) were "pretenders":
Ambrish Srivastava - BMO Capital Markets
Okay. And then maybe the bigger picture question Paramesh, there's a lot of what I would call a vaporware talk out there on the ARM side so I don't want to get into that because you guys, at least, to my humble opinion, are way ahead of the ARM, I'll call some of them pretenders.
But Intel, you seemed to have wakened up the big giant, not to assume that they were ever asleep. Can you please help us understand what you've seen from them in terms of whether it's PR or their product releases? And how does that impact or when you compare your roadmap with them, where do you see opportunities or threats for X-Gene? Thank you.
Is this fair? Cavium Networks, for example, has delivered year-after-year very high performance, highly integrated network processors. Advanced Micro Devices, too, has years of actually generating hundreds of millions in server MPU sales. Then, of course, we have Calxeda and Marvell (NASDAQ:MRVL) whom have, at least in limited quantities, commercially shipped ARM-based microserver/storage processors.
What has Applied Micro shipped again? It promised a very suspiciously similar product known as "Titan" in the mid-to-late 2000's that had suspiciously good performance/power characteristics (sound familiar?) even on an ancient process node. Well, that product never came to market and the only MPUs Applied has ever shipped were behind-the-curve single/dual core, stock IBM POWER core based processors.
Naturally, Mr. Srisvastava pointed out that the 800-lb server gorilla, Intel (NASDAQ:INTC), that has driven heavyweights like IBM (NYSE:IBM) and Oracle (NYSE:ORCL) into the very fringes of the datacenter, is making a serious run into this space. Shockingly, Dr. Gopi completely brushed off this threat with spin that would make an electron dizzy:
So I think, given the recent announcements that we've seen from Intel, it heartens us relative to being strategically on the right path to solve the problem for datacenters. And I believe that we have done the - we've broken new ground and it's heartening to see the x86 industry standard server component vendors starting to acknowledge the new ground that we've broken.
What's ironic about Dr. Gopi claiming that Applied was the one that "broke" the ground is that Intel's low power, highly integrated parts are shipping for revenue today while Applied continues to brag about how many customers have signed up for evalation samples of its 40 nanometer parts.
Don't worry; it gets worse.
Applied Micro Now Claims That Low Power Atom SoCs Are Not The Competition, But Xeons Are
When I began following Applied Micro two years ago, the whole point of X-Gene was that it would take a bunch of low-power processor cores, tie them together onto a single system-on-chip, and thanks to the high levels of integration and focus on more - but low power - cores, it could offer a performance/watt advantage.
Of course, when Intel responded with a set of products that did precisely this ("Avoton" for micro-servers, "Rangeley" for communications infrastructure), and did so at very low power levels (Intel's 8 core, 2.6GHz C2750 part has a max Thermal Design Power ("TDP") of 20W), Applied Micro began to backtrack.
Indeed, see this excerpt from the Q&A session of the most recent call:
Patrick Wang - Evercore Partners
Okay, great, that's all right. So it's fair to say that you feel pretty good about the ramp next year. Okay, and then the other question I want to ask was, well going back to X-Gene, there is lots of thought out there in terms of what X-Gene is capable of, what the competitors are at the timing, this and that. You know, there is a part out there today that you guys will be competing with and it's Intel's part. Have you - do you have a sense of kind of how X-Gene compares on a performance side of things versus Avoton, and I know that you guys are using E5 as your kind of competitor, but given the fact that Avoton is their Atom base part, can you help us understand how you expect to outperform them and by how much?
Paramesh Gopi - President & CEO
Patrick, it's a really good question and I want to take this opportunity to say two things. I think the industry is confused with this whole notion of micro server. There is no such thing as the micro server. All of the servers shipping today in the world are cloud servers because cloud is characterized by Linux, Apache, MYSQL, PHP, 64-bit X86. I would say almost 99% of those are all E3 and E5 based. What I want - the other characteristic in the cloud is large addressable memory for big sparse data, for big data which means two to four channels of very high performance DDR3 or DDR4 and many 10-gig pipes. So the struggle that we have and we talked to people and making them - to help them understand the market is, Avoton doesn't have the 10-gig or the memory support of what is required for the mainstream cloud, which is why we keep talking about Xeon, Xeon and Xeon that is the only existing platform that we need that we are in the process of basically either replacing or amending with our ARM portfolio. So I want to stop there because anything else that people talk about in the micro server space is the term that was coined in a, I think, with no basis. Its cloud servers, it worships today and fundamentally you have to talk about 10-gig, multiple 10-gig, multiple two to four channels of high speed DDR, a lot of the integrated storage and lots and lots of PCIE Gene three. We are talking about between 12 and 24 lanes, you can't find that on an Avoton part today. So it's very hard for us to draw that comparison.
Oh. So there's no such thing as a "microserver" then? I find that interesting given that in a press release dated Jan. 16, Applied Micro proclaimed it "contributed the first ARM-based micro-server specification" to Facebook's (NASDAQ:FB) OpenCompute project.
More interestingly, Dr. Gopi was claiming that there was "no difference" between a Xeon E3/E5 and an X-Gene in one particular virtualization workload (i.e. one that is conducive to many smaller cores rather than single threaded performance). The question, then, is which model of Xeon E3/E5? What generation? What bin? These chips range from 15W, dual core chips to 130W 12 core/24 thread monsters.
Of course, when the Raymond James analyst asked for any "proof" of these benchmarks, Dr. Gopi offered a pretty weak reply,
Hans Mosesmann - Raymond James
Okay that's helpful. And so is that information published some place or is there a white paperwork we find that information with. Is it more anecdotal at this point?
Paramesh Gopi - President & CEO
We actually have hard metrics, but our customers have provided us so that's why we said this on the call today. We wanted to make sure that there is at least some color that we can provide in terms of the class of processor, the class of platforms that we have been able to build and pilot ship, thanks.
I don't suppose that Applied Micro would actually like to share these metrics in order for the processor analyst community to actually figure out what's going on, would they? Doesn't anybody else find it suspicious that Applied won't give so much as a SPECint or SPECfp value (these are industry standard performance benchmarks)? Of course, they had no compunctions providing "projections" for their next generation 28nm X-Gene (back before the first generation on 40nm even taped out):
The problem is that these numbers, even if true, damn the X-Gene project from the start. If the 28nm, 4 core/4 thread, 3GHz part truly only ends up scoring ~52-53 in SPECint 2006 at a single socket processor power level of, oh, 14-15W, then this is actually pretty uncompetitive as Intel claims that an Avoton part at 19W total rack power (suggesting that it is a 1.7GHz, 8 core/8 thread 12W single socket TDP part) scores ~69 in SPECint 2006.
Now, do note that the Applied Micro projections are for the X-Gene version 2 which will not even begin SAMPLING until Spring 2014 as per the call, so the current X-Gene (at 2.4GHz and doesn't get the micro-architectural improvements of the second gen part) will likely perform significantly worse on a performance/watt basis than Intel's Avoton. Maybe Applied is targeting Xeon power levels since it is two process nodes behind?
Dr. Gopi was right, this really is Apples versus Oranges: the Intel part is simply in a league of its own with respect to performance per watt. Let's also not forget that AMD will be sampling its own 28nm parts in 1H 2014 for launch in 2H 2014 (beating Applied's second generation X-Gene to market). Qualcomm (NASDAQ:QCOM) and Samsung (OTC:SSNLF) likely have products in tow for next year, too.
More Very Nonsensical Statements From Applied's Manangement
Take a look at the following charts and tell me if you notice a trend:
In the semiconductor space, chips become tougher to build as new process geometries necessitate more sophisticated design methodologies and man-power. This means more R&D as time goes on, and companies that cannot fuel further R&D expenses typically fall out of the running.
Applied Micro's CEO made the most curious statement on the call:
I can tell you that when we look at modeling this, we look at pretty much doubling our revenue in three years that's what we talk about and keeping expenses relatively flat.
Now, let's suppose Applied pulls this off. TTM revenues for the company were $208M. Let's suppose that in the next couple of years, the company simply doubles this revenue base to $420M. Let's also suppose that gross margins remain intact at about 60% (I'm being generous on a GAAP basis), and that opex stays roughly flat over those several years (unlikely, but let's play along). This suggests a $235M yearly run rate ($187M R&D, $48M SG&A). Does this math work out? No.
- Net Revenue: $420M
- Gross Profit: $252M
- Opex: $235M
- Operating Income: $17M or $0.23/share on a GAAP basis (let's assume for a moment that the years of NOL's they'e racked up reduce their effective tax rate to effectively zilch for a while)
Of course, we need to double revenues to get to this point (i.e. take a significant chunk of the micro-server market). Am I the only one who sees a real problem with this picture?
Note also that Applied STILL has ~$70M left to pay in consideration for the Veloce spin-in. Of course, the company will probably want to do as much of it in stock as possible (since the shares are highly inflated), but this will take a further toll on the company's roughly $75M in cash. If the company is REALLY smart, you will see them announcing a secondary very shortly. It'd probably coincide with some major "milestone" with respect to the X-Gene project so that investors are happy to buy up the excess shares that get dumped onto the open market.
It's tough to justify today's $12 price. With cash burn fueled primarily by a venture that will face significant competitive pressures in light of bigger, more experienced, and better capitalized players, shares look EXTREMELY overpriced here. While shorting a "story stock" may not be right for some (although if you do want to, wait until the bump that the shares will likely get at ARM Tech Con this week), the key takeaway is to not buy into this hype. This is a $4-6 stock in due time as the X-Gene fanfare fades and as the cash burn continues. If you want to bet on a small, high quality processor vendor that could actually strike it rich in this space, look at Cavium Networks (although it's priced for perfection, its management and product track record speaks very loudly). But even so, it's more likely that the "big dogs" with scale end up owning this space. Think Intel, Samsung, and Qualcomm if it wants to get in on the action.
Disclosure: I am long INTC, AMD. 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.
Additional disclosure: I plan to short AMCC on any spike following ARM Tech Con.