AMD EPYC CPUs Reach Record Server Revenue Share of 46.2%
Posted by SirActionhaHAA@reddit | hardware | View on Reddit | 28 comments
Posted by SirActionhaHAA@reddit | hardware | View on Reddit | 28 comments
Turtvaiz@reddit
I'm surprised it's still so low. Intel seemed to have much worse value and performance for many years
Exist50@reddit
Might be buoyed by enterprise. AMD's been mostly focused on the big hyperscaler sockets, and less in the 8ch space where a lot of enterprise deployments are. Though it sounds like Intel won't have any new offering for those for a number of years.
IBM296@reddit
To be fair, it wasn't really widely known that Intel chips were that bad before 2022/2023. After that you can see that AMD's growth has really accelerated.
dabocx@reddit
Venice is going to be a very interesting launch if the hype for it holds up. And what the volume is.
Geddagod@reddit
I mean even just based on the on paper confirmed specs from AMD, Venice seems incredibly impressive.
Even if Venice takes a bit to ramp too, Intel isn't rumored to have a competitor out till next year, and Intel themselves don't think DMR will be actually competitive.
heylistenman@reddit
DMR has such a weird vibe around it. Lip-Bu Tan denouncing the move away from hyperthreading, the cancelation of 8-channel versions, the rumored delays, the acceleration of Coral Rapids.
At this point I wouldn’t be surprised if they gave DMR the Falcon Shores-treatment.
Exist50@reddit
Probably best if they don't. Hasn't been working for Falcon Shores either. Every time they cancel the next gen and try to "accelerate" the one after, no actual acceleration happens.
Coral Rapids seems to be in a difficult spot timing-wise as well. DMR will already launch a node behind. If COR doesn't get a healthy 14A (which is '28, probably late '28, at best), then they risk competing with Verano at a 1.5-2 node advantage. They'd be better off getting DMR out the door to at least displace GNR, even if it's not ideal.
jaxspider@reddit
It would be a higher number if they could actually produce more.
IBM296@reddit
Exactly a year ago, Intel's CEO said the company had 55% of the data centre market.
26H1 Intel unit share was 54.9% according to this article. Looks like his remarks were correct lol
SirActionhaHAA@reddit (OP)
That's including arm. Amd's x86 dc share was in the 39.4% at that time, so intel had 60.6% of the x86 dc share.
IBM296@reddit
53.8% is the revenue share. X86 unit share of Intel is still 73%.
SirActionhaHAA@reddit (OP)
Correct, amd has much lower "unit share" because their cpus contain more cores and are faster. Businesses don't buy sockets, they buy cores. Shrinking a 100 sockets system into 20 is literally the main point of buying epycs, it reduces tco
It's the revenue share that matters.
Minced-Juice@reddit
AMD' higher revenue share and lower unit share means large companies prefer AMD's higher priced SKUs because it consolidates rack space for a given number of cores.
Conversely, smaller companies prefer Intel because of better availability and pricing from vendors.
It has been like this for quite a while now. At least as far back as Rome (Zen 2).
Soft_Proof@reddit
Shh they do not understand. Fun fact is thay AMD has only 29%-32% of x86 market (server+consumer) and it stale from 2025. The revenue they get is from pricing and plateau phase. If Intel will win with new xeons AMD will decline fast.
SirActionhaHAA@reddit (OP)
Total cpu revenue from q1 2025 was 31.7%
Total cpu revenue from q1 2026 was 38.1%
Soft_Proof@reddit
Revenue grow from higher priced EPYC, not from volume. Also TPUs in mix, not from x86 by itself. Seems ya have no clue.
Geddagod@reddit
Mercury Research also tracks unit share, and their data claims unit share (volume) also grew.
BleaaelBa@reddit
IBM296@reddit
That's not going to happen cuz AMD's new EPYC's are also launching at the same time.
Exist50@reddit
Earlier, it sounds like.
SirActionhaHAA@reddit (OP)
Mobile share hit 28.9%, almost 1/3 of the x86 laptop share. Debunking again the reddit myth which claimed that amd hardly ships any laptop parts.
siazdghw@reddit
That's still quite bad when you look at the context and will definitely decrease in the future.
Up until Meteor Lake (essentially a 2024 product) AMD had clearly better mobile products. While MTL and LNL were big advances for Intel that flipped the tables they weren't really mass market products like PNL will be once it trickles down to most of the segments, against AMD who is offering uncompetitive rebrands currently.
Considering the lifespan/time between laptop purchases is around 3-4 years, we haven't even seen Intel's turn around hitting hard yet. There's basically a 'lag' in the industry where turn arounds or significant improvements don't show up in market share until people are ready to upgrade. Just like we saw when AMD did it's turn around. When the Phoenix laptop owners (a majority of AMDs mobile sales) decide to upgrade later this year or next, it will likely be to PNL and not the inferior Gorgon/Strix unless AMD cuts prices to try and stay competitive.
Next year AMD will probably be back to <25%
Earthborn92@reddit
No one who doesn't need a laptop this year wants to buy a laptop. Memory has made it difficult.
AMD is at a disadvantage in terms of tech to Intel in the one year when laptop sales would be very low. I expect it'll be much more competitive once the Zen 6 / RDNA 5 family rolls out.
chandleya@reddit
They’re still pushing a ton of 7800 series stuff on the market lol. We live in strange times. It’s still acceptable to see an i5-13000 in a brand new laptop.
InflammableAccount@reddit
To be fair, there were some SKUs that were as rare as a unicorn. But who knows why? Poor adoption from OEMs, better than expected binning, poorer than expected binning, etc.
The one time a release cycle was genuinely scarce was Hawk Point. That mofo was practically vaporware compared to other releases.
Geddagod@reddit
The more interesting fact, IMO, is the unit share. AMD gained 2.3% unit market share QoQ... in a supply shortage where Intel sold older server chips that would not have sold at all were it not for the shortage.
Paraphrasing Stacy Rasgon in Intel's Q4 2025 earnings call, how is Intel losing market share and not having enough supply... when they have their own fabs?
SlamedCards@reddit
Because they took down Intel 7 and weren't aggressive on adding Intel 3 until later this year
Geddagod@reddit
Taking down Intel 7 was understandable, but Pat's planning on Intel 3 volume really screwed them over unfortunately.