AMD has just hired two executives to run the business responsible for its graphics chips and architecture. They face a steep challenge as they seek to catch rival graphics giant Nvidia, update the AMD graphics roadmap and develop products that will increase revenue and market share through 2020.
Without a direction for the growing artificial intelligence and machine-learning markets as well as the core gaming segment, the AMD
graphics division appears to be vulnerable.
Mike Rayfield will operate the business portion of the graphics business unit, coming from Micron with a previous stint at Nvidia. David Wang returns to lead the engineering and technology portion; he had helped with AMD graphics development for a decade before moving to Synaptics in 2012.
They replace Raja Koduri, who left for Intel to build a discrete graphics group from the ground up, while also managing the CPU design group. That puts added pressure on AMD to develop new products that could close the gap with Nvidia
can get up to speed.
The question for AMD’s Radeon Technologies Group is how quickly it can do that under the new leadership. Though business practices, partnerships, and policy can be adjusted, if the fundamental architecture that powers its graphics products is lacking and needs to be redesigned, that can be as much as a three-year task. If the chip design is solid and only minor changes to be made, it is still a full year of work before any input Wang might have can be implemented. There is no technical reason that I see for a full redesign of the architecture, so it seems more likely a minor cleanup is all that is needed.
As a first step, I would expect Mayfield and Wang to provide investors with a new roadmap within three months. At the company’s CES event for media and analysts earlier this month, the roadmap was vague and open-ended, leaving many gaps between now and 2020. The only announcement was for a revised graphics chip for machine learning that would be sampling in small quantities to customers in 2018, but no timeline for production or revenue was given.
The gap between AMD and Nvidia graphics technology remains complex. Nvidia has the dominant position in machine learning over not just AMD but every other competitor. AMD has the advantage of being the only other company with a robust graphics IP portfolio, but the work Nvidia has done with the development community and in software tools to make that development easier has paid off. AMD’s first goal should be addressing this space, both with updated graphics products and an increased investment in its software stack.
On the gaming side, Nvidia hasn’t been able to grab more market share from AMD despite arguably having the better products. Thanks in part to the consistent demand driven by cryptocurrency mining, AMD is able to sell every graphics card it can build. It’s basically a race between the two to produce more graphics cards. There is a need for balance and control on that side as overproduction during a crypto-crash could result in inventory without a market to sell to.
But cryptocurrencies might be the first and biggest opportunity for the AMD unit’s new leadership. Nvidia has continually shied away from being known as a cryptomining chip provider, despite all indications that consumers and investors are buying a large percentage of Nvidia graphics chips for exactly that. There is risk, but if AMD were to lean into the space and create a unique spin, something that combined the benefits of gaming and mining into a single product, it could result in a quick spike in revenue. Both AMD and Nvidia are trying to find ways to improve margins with the graphics card shortage (most of that additional income has gone to resellers and the channel).
AMD should also direct more energy and resources on its combined processor-graphics chips called APUs (accelerated processing unit). As the need for alternative solutions for PC gamers has grown, AMD is unique with its ability to offer products with both high-performance processor cores and high-performance graphics. Focusing on that product line would create a space for AMD to market to that neither Nvidia nor Intel have the ability to address.
The key points that AMD’s new graphics leadership team has to address are clear. Get the roadmap and architecture back on track to compete with Nvidia long-term. Provide partners with a reason to believe that the company can compete in the machine learning and artificial intelligence spaces that are providing extreme growth opportunities for the competition. Extend the software infrastructure to draw in developers to use the AMD graphics-chip-based solutions for these machine-learning applications. Develop a plan to take advantage of the near-term cryptocurrency mining demand while also buffering the company from the potential pitfalls of a bitcoin crash. Double down on the PC gaming space to grow the install base and build brand excitement.
It’s not an easy list, but with strong competitive options in every direction, AMD must produce on these points if it is to see long-term growth.