- Intel recently launched its latest laser chips.
- The move could boost Intel's profitability, ASPs and revenues.
- It could also potentially curb the growth of ARM-based offerings in the server space.
It appears that Intel's (NSDQ:INTC) dedicated efforts towards the development of silicon photonics have finally come to fruition. The chipzilla announced earlier this week that its chips with laser-based interconnects are ready and commercially available; the new technology being used here, goes by the name of silicon photonics. But what is this technology really? And how significant is it to Intel’s business? Let’s take a look.
What it's all about
Well, enterprises that have their own data centers, usually have several servers that work in tandem with each other. Server farm operators generally have very few systems that are isolated from their server clusters, and most of their network is setup to work in parallel with each other, so that the load on the entire network is evenly borne by all systems. This might have been an oversimplification of things, but that’s generally how hyperscale data centers operate.
Now, the problem that these data center operators face is that each of their stacked servers need to be connected to other servers on the network, with interconnects using PCI-e or networking cables, that reportedly deliver a peak bandwidth of 8Gbps and 40Gbps respectively. Since PCI-e cables are at least 10-times faster than their Ethernet counterparts, they are currently the preferred standard for interconnects. (Also Read: Intel stock: A Great Buying Opportunity)
This is where Intel is hoping to disrupt the market. The chipmaker’s latest tech deploys optical-fiber based interconnects. Granted that optical fibers aren’t exactly new, and have been in the market for decades, with several data center operators already moving to use more of such cables on their farms. But what’s different in Intel’s case is that it has released miniaturized on-die versions of optical transceivers to establish speedy interconnections between FPGAs and its Xeon chips.
These reportedly have a significantly lower total cost of ownership, a much lower latency and sport much higher bandwidth ~100Gbps, compared to the bulky and expensive copper-based PCI-e cables. This is something new and groundbreaking; it’ll cater to the server-segment for now. So why is Intel pursuing it?
Advantages for Intel
Well, Intel’s Xeon and FPGA chips will have an enhanced appeal after this release. Data center operators can expect about 100-times lower latency (uncorroborated figure) that could fuel a range of real-time applications such as credit card processing or e-commerce transactions. In such mission critical areas, I doubt enterprises such as Apple or Google would go with other server architectures just because they are cheaper. This could become the industry’s gold standard and in turn, bolster Intel’s brand recognition, market position and pricing power in the server segment.
Plus, we have seen numerous reports over the past year ahout how ARM and its partners (such as AMD, Qualcomm and Cavium) are making moves to grab a slice of the lucrative server segment. But with the introduction of silicon photonics, it will become harder for these companies to nab market share from Intel. So this could very well curb the growth of ARM in the server space.
Moreover, there are several large companies, including Google, Microsoft and Facebook, that design their own data centers and server clusters. A few companies of this scale go with PowerPC, and a few others have expressed their desire to explore ARM for their hyperscale server needs. But with Intel servers offering significantly faster performance with this release, I believe that such firms might think twice before moving away from the chipzilla’s offerings. So this would not only defend Intel’s market share from any potential erosion, but would also boost its revenues.
Intel will now be able to charge a premium for this new tech, considering that it's new and being hailed as “miraculous” and “mind-blowing.” This can, in turn, will boost Intel’s ASPs and gross profit-per-chip.
We have very little information available on costs and performance improvement variables associated with this tech as of now, so its very difficult to comment on how much of a TCO (Total Cost of Ownership) advantage there is. But it is of vital importance that the additional cost of setting these embedded chips in server stacks be minimal, otherwise it might not be able to compete with ARM in terms of performance-per-dollar metrics. So this aspect could be a potential spoiler. (Also Read: Will Intel Corporation (INTC) Disrupt The Drone Industry?)
Further, Intel isn’t the only one working on this technology. In fact, IBM, Luxtera, Fujitsu and a few other startups are working on this technology as well. So Intel would stand to lose its pricing power as this technology gets standardized over the coming years. The production yields could be low considering the fact that there’s an additional on-die optical transceiver embedded on Intel chips now. So, the chipzilla would have to work in this area and pay close attention to the yields, in case the market for these chips opens up fast.
Putting it all together
Silicon photonics definitely looks like a disruptive new technology. It can encourage enterprise clients to upgrade their setups, and that could potentially reinvigorate the demand growth of Intel’s server chips. It stands to also boost Intel’s revenue, ASPs and market dominance. So this comes as another move by the chipzilla to bolster its established business. Looking for tech stocks? Check out Amigobulls' top technology stocks, which have beaten the NASDAQ by well over 100%.