Is Mellanox The Next Nvidia?

| About: Mellanox Technologies, (MLNX)
This article is now exclusive for PRO subscribers.

Mellanox (NASDAQ:MLNX) is a fabless semiconductor company that designs interconnect products for the datacenter. In particular, the company's products focus on both InfiniBand and Ethernet adapter and switch silicon. Now, while Mellanox's Ethernet business is growing nicely as the Web 2.0 infrastructure/cloud builds out (ethernet is the interconnect of choice here), the firm's real crown jewel is the high performance computing space in which it can deploy its InfiniBand products.

Why is InfiniBand so important?

InfiniBand, for those of you unfamiliar with the technology, is a very high speed network connectivity standard. It connects up systems as Ethernet does, but it is generally much faster at doing so. For many datacenter applications, particularly those that are cost-sensitive, Ethernet is usually the better choice because it provides rather robust speeds for comparatively little cost.

However, for those applications that want no-compromises speed, InfiniBand is generally the way to go. The trade-off here is on the cost side of things. In the land of high performance computing, where there is a real emphasis on the performance part of the equation, the folks building the datacenters are more than willing to pony up the extra cash to deploy InfiniBand.

Mellanox has a leading position in InfiniBand products today. A couple of years ago, Mellanox's chief competitor in this space was QLogic (NASDAQ:QLGC), but Mellanox pulled far ahead on the technology side of things. Eventually QLogic gave up trying to compete with Mellanox and sold the business to Intel (NASDAQ:INTC) for $125 million. While QLogic was a pushover in this space, the Intel competitive threat has loomed for quite some time.

The Intel Threat: It's More Subtle Than Originally Thought

With QLogic's InfiniBand business now under the banner of Intel's highly successful datacenter group, the competitive threat to Mellanox's market share is dramatically heightened. Intel's datacenter group sports an exceptionally high operating margin and, as a result, the company can afford to aggressively invest in new technologies. Further, given that Intel holds a dominant share of the server CPU market, it has been increasingly able to bundle complementary technologies with its platforms.

For Mellanox, it would seem that on the surface, Intel's entry into this space would shake up the competitive dynamic to a meaningful degree but not a fatal degree. After all, while Intel probably upped the resources it was allocating to the former QLogic business, it is important to not lose sight of the fact that this is Mellanox's core competency and that designing a superior product may be rather difficult (fellow author and friend Stephen Simpson seems to believe that Intel's advances in this space are likely to be all bark but no bite).

However, Intel may hold a proverbial trump card. The company has expressed its plans to integrate InfiniBand connectivity straight onto processors designed for high performance computing. While in a prior article I had pointed out that integration of InfiniBand onto the company's mainstream Xeon processors would not happen for at least the next couple of years, it does seem that Intel is accelerating its integration efforts in its Xeon Phi CPUs targeted solely at the high performance space.

Indeed, a recently leaked slide from VR-Zone confirms that Intel does indeed plan to integrate two 100 Gb/s InfiniBand ports into a special vairant of its Knights Landing processor. Interestingly enough, Mellanox's 100Gb/s InfiniBand silicon will likely be made available in the 2014 timeframe (to coincide with the launch and subsequent ramp of Intel's next generation "Grantley" platform) - outrunning Intel's integrated solution (and likely its discrete solution) by about a year.

Is Mellanox The Next NVIDIA?

While integration is the new buzzword, I believe that the long-term threat to Mellanox's business could be along the lines of how Intel "threatens" NVIDIA (NASDAQ:NVDA) in the PC graphics space. While there will surely be some share erosion from Intel's discrete parts and, more markedly, its integrated parts, as long as Mellanox can offer higher performance discrete solutions ahead of when Intel offers integrated solutions, then Mellanox's InfiniBand business should be in good shape going forward. This does increase the risk profile of the investment, but at this point the pessimism appears to be baked in.

Longer term, I do suspect that Mellanox will be looking to decrease its dependence on Intel platforms. This is why we are seeing Mellanox and NVIDIA becoming so chummy, and this is why Mellanox and NVIDIA are promoting IBM's (NYSE:IBM) POWER 8 architecture; success for IBM's design in HPC means that both of these players can lessen their dependence (and thus risk of share loss to Intel's in-house solutions) on Intel.

At Any Rate, This Won't Be The Focus In 2014

Despite the challenges, I do plan to initiate a long position in Mellanox. The play here is that Intel's upcoming "Grantley" platform launch will drive a real uptick in HPC demand/buildout which should bode rather well for sentiment around Mellanox. I suspect there will be very little share erosion from Intel's own InfiniBand parts (and these will be the discrete ones), and I further expect that Grantley should ramp at least as steeply as Romley did (if not more aggressively as the Haswell-EP CPUs will sport ~14 cores and have twice the SIMD throughput via AVX2) in HPC, as well as in other segments.

Mellanox's InfiniBand business should do well in 2014 as should the rest of its datacenter connectivity products. This certainly isn't a "buy and hold" type of investment, but I do believe that the market does not yet fully appreciate that Mellanox is at a cyclical bottom and that next year the company is likely to see rather dramatic growth that will fuel very similar optimism to what was seen in 2012. The company probably won't return to its $100+/share bubble highs, but a run into the low $50's seems more than reasonable at this point (more detailed justification of this can be found in my initial thesis here).

Disclosure: I am long INTC, NVDA. 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 initiate a long position in MLNX.