It will be an understatement to say that Intel (INTC +4.67%) is in a state of disarray. Its core markets are being attacked from several angles, its latest chip architecture has been delayed (yet again), and the company has been left rudderless after the resignation of Brian Krzanich. But there's still a lot that could go wrong for Intel in addition to all of these problems.
The company recently revealed that it plans to launch its first discrete graphics card by 2020. But even if Intel hits the promised timeline and actually manages to get a discrete GPU (graphics processing unit) on the market, I believe that it won't be able to make a big dent in this space. Its foray into discrete GPUs could, at best, turn out to be an exercise in futility.
Image Source: Intel.
Intel may have already missed the bus
Graphics card specialists NVIDIA (NVDA +0.99%) and Advanced Micro Devices (AMD +2.00%) have been plying their trade in the GPU market for a very long time now. Both of these companies have poured research and development (R&D) dollars into their GPU architectures and are currently preparing for another generational leap.
For instance, NVIDIA's Turing GPUs are expected to succeed the existing Pascal architecture this year. Now, there's a lot of speculation about the process node on which the Turing architecture will be based. The general consensus among industry watchers suggests that it could probably be based on a 12-nanometer process node as compared to the current generation's 16 nm and 14 nm nodes.
A smaller process node allows the chipmaker to pack more processing power into a smaller footprint, leading to lower production costs and more power efficiency. Now, the jump to 12 nm won't be a huge generational leap over the current architecture, but it sets NVIDIA up to eventually move to a more efficient 7 nm process node from Taiwan Semiconductor Manufacturing Company.
However, there are rumors that suggest NVIDIA is skipping the 12 nm process altogether and moving to 7 nm. But even if NVIDIA sticks to 12 nm first, we can assume that the graphics specialist will be able to roll out cards based on the more efficient 7 nm manufacturing platform by 2020. This is because NVIDIA usually releases new products every two years. Meanwhile, AMD recently showcased a 7 nm Vega GPU for use in servers and powerful workstations.
AMD will eventually bring this architecture to the consumer GPU market as well sometime next year. Intel, however, is probably waiting for its 10 nm process node to materialize so that it can compete against these upcoming chips. But the problem is that Intel has run into consistent delays with this technology.
Provided Intel manages to hit its timeline and launches 10 nm parts by 2020, it still won't be offering anything new over what NVIDIA and AMD would be at that time. This will make it difficult for Intel to score customers for its discrete GPUs given the dominance that its rivals have already established in this space.
Knocking NVIDIA off its perch won't be easy
NVIDIA currently commands nearly two-thirds of the GPU market, as per Jon Peddie Research. It won't be losing its stronghold anytime soon thanks to its solid presence in both emerging and developed markets.
NVIDIA's emerging market base has increased at a very impressive annual pace of 40% over the past five years, while developed markets have also clocked an envious 28% growth rate. This has helped the company establish a solid installed base of users who will likely upgrade to its subsequent GPU generations thanks to technology advancements such as ray tracing.
In fact, NVIDIA claims that just 30% of its installed base is currently on the latest Pascal platform, while the remaining uses older-generation GPUs. So the company has a big opportunity to sell more GPUs as its installed base upgrades to new platforms that come out in the next couple of years. Given that an average GPU is expected to provide service inside a PC for an average of three years, NVIDIA looks all set to lap up most of the addressable market as it launches its more advanced GPU technology before Intel does.
Intel is wasting time
It will be extremely difficult for Intel to crack the dominance of its rivals in the discrete GPU market given the massive technology and time lead that they enjoy. The company is better off focusing on other areas such as field-programmable gate arrays and server chips, two markets where it has a realistic chance of making it big but faces severe competition from the likes of Xilinx and AMD.
In all, Intel could be making a big mistake by turning its attention toward a market that's already notoriously competitive, as this could cause it to lose sight of the more important opportunities that it's sitting on.
