NVIDIA’s discrete GPU market share
Intel’s (INTC) partnership with Advanced Micro Devices (AMD) would likely help the former tap into the mobile PC (personal computer) gaming market. But NVIDIA (NVDA) still doesn’t think of Intel as strong competition in the gaming space.
That said, Trefis warned that Intel could threaten NVIDIA if the former takes up more discrete GPU (graphics processing unit) market share. As Intel controls nearly 80% of the PC processor market, the addition of a discrete GPU inside its microprocessors would likely help the company become a top player in the discrete GPU market.
If NVIDIA loses market share in the discrete GPU space, it would lose its power to command higher prices and would be forced to cut its prices in order to stay competitive. A reduction in ASP (average selling price) would significantly reduce NVIDIA’s gaming revenue and overall earnings because it earns 60% of its revenue from this segment.
However, triple-digit growth in the data center segment would likely help NVIDIA partially offset declines in gaming.
Intel in the AI market
Intel’s venture into the discrete GPU market could threaten NVIDIA in the data-center space in the long term. Data center customers are adopting NVIDIA’s GPUs for their ML (machine learning) tasks, thereby reducing the demand for Intel’s server CPUs (central processing units).
This has encouraged Intel to develop AI-optimized (artificial intelligence) processors, but none of its products have achieved the success of NVIDIA’s GPUs. Now, Intel has created a group to develop high-performance graphics chips.
The company plans to leverage the experience of former AMD GPU head Raja Koduri to develop GPU solutions for AI. As Intel commands 99% of the server CPU market, it can integrate this data center GPU solution inside its server CPU and benefit from its large-scale operations.
Intel could also expand GPU solutions in autonomous driving and other AI applications, giving strong competition to NVIDIA.
In its fiscal 3Q18 earnings call, NVIDIA’s chief executive, Jen-Hsun Huang, responded to Intel’s entry in the discrete GPU space. Huang explained that NVIDIA’s modern GPUs are very complex and are backed by the CUDA API (application programming interface), which is a result of more than ten years of software engineering.
Huang stressed that NVIDIA has a seven-year head start in GPU computing, and CUDA is what enables its GPUs to accelerate applications. Intel would take three years to build GPU architecture.
That said, even if Intel succeeds in developing the most advanced GPU with the help of its high R&D (research and development) resources, it lacks strong software support, which makes GPUs useful to consumers.