Since the turn of 2005/2006, AMD’s share of the PC processor market has been declining. With small fluctuations until the beginning of 2017, when the Ryzens were released. The decline was quite slight at first, more pronounced at the turn of 2006/2007. Then the situation stabilized slightly and with some product lines the slide was temporarily reversed or slowed down. Another turning point came with the era of CEO Rory Read, under whose leadership the share only declined without any stagnation, let alone (albeit temporary) improvement.
AMD’s market share then did not grow steadily, but it was possible to record two more significant jumps. The first relates to the aforementioned release of the first Ryzens in early 2017. The second followed around the spring of 2019, when mobile Ryzens 3000 built on a 12nm APU entered the market. Picasso, followed by Intel ‘s problems with limited 14nm production capacity, and then the situation reached the 7nm release Zen 2 / Ryzen 3000 for desktop. Later, when Intel’s factories took a second breath, however, the decline in AMD’s share / Intel’s strengthening did not follow. AMD’s share continued to grow, so Intel’s lack of production capacity did not load AMD as much as some sources expected in 2019.
In the second quarter of this year, AMD achieved a share of PC processors of 16.9% (0.9 percentage points more than in the previous quarter), which is the highest since 2006. Although this number contrasts sharply with its share in the retail segment, where AMD already for several years it has reached a much higher share. But retail is not what it was in the second half of the 1990s and around the year 2000 – it has been shrinking for a long time, more and more people are buying assembled computers. It is in those (OEM) segment that Intel is able to maintain its position, which is given by relationships with manufacturers, but also (and a lot) production capacity. From the OEM’s point of view: If I have a production capacity of 100 computers a day and Intel supplies me with processors to assemble 100 computers a day, what’s the point for me to move to AMD and make better but only 20 computers a day (because AMD won’t supply me with more processors) ?
AMD thus reached about 17%, but does it have any room for any further (significant) growth? GlobalFoundries is fully utilized, with TSMC it has already become the largest customer of this process with 7nm generations, and although TSMC lines are expanding, the interest of other manufacturers (Intel, Nvidia) in these lines is growing, so it will not be just AMD.
So far, she has dealt with the situation by deciding to focus on the high-end instead of the low-end. So not to aim at any price for the number of processors sold and market share, but to produce a smaller number of chips from the available wafer, but so much more powerful that it can have higher margins on them. However, this step has already taken place, so it cannot be used again. It will be quite interesting to see how society copes with this situation. It may not be necessary if Intel manages to increase competitiveness and stop AMD’s market share from growing.
Finally, it is worth noting the simple fact that you also need higher and higher production capacity to maintain market share, because production volumes are still growing. If you make 17 processors out of 100 sold this year and 17 processors next year, it will no longer be 17%, but a little less, because the total number of processors sold will not be 100, but 115.
If we do not talk only about PC processors, but also include other segments (servers, etc.), then the share of AMD rose to 15.8% (expressed in percentage points: by 0.7% quarter-on-quarter and by 4.1% year-on-year).
According to MercuryResearch, AMD currently has 11.6% of the server market. For servers, it depends on what the author of the statistics considers 100%, ie the market. Some statistics operate with the entire market (regardless of architecture, such as this one), others work with x86 markets (according to which AMD has already exceeded 15% share), and may differ depending on what types of products they still consider / or no longer consider. for server (ie eg whether they count IoT, etc.).
–
