x86 server CPUs

What the evidence shows

Who supplies it

A small number of additional x86-compatible vendors operate in niche or restricted markets. Zhaoxin, a Chinese fabless firm, develops x86-compatible server processors as a continuation of VIA Technologies' x86 line under a joint venture, while VIA Technologies itself has historically co-owned that venture. Hygon licensed AMD's first-generation Zen design in 2016 and merged with Sugon on May 26, 2025 by swapping stock, forming a combined Chinese server and supercomputer supplier. RDC Semiconductors has also designed x86-compatible RISC cores for embedded use, though these are not competitive in mainstream server sockets.[3][5][9][6]

Nvidia is positioning itself as a third major CPU supplier for AI servers through vertical integration, although its CPUs are not x86. Nvidia's chief financial officer Colette Cress stated that sales of the company's Grace and Vera CPUs for data centers are expected to hit $20 billion this fiscal year, which would outsell both AMD and Intel and make Nvidia the world's largest processor supplier by revenue. SoftBank acquired Ampere Computing in 2025 for $6.5 billion, adding ARM-based server CPUs with up to 192 cores in the current AmpereOne generation and a planned 512-core Aurora successor. None of these alternative-architecture suppliers appear in x86 shipment tallies because their instruction set is different, which is why the x86 duopoly number can coexist with a much more crowded overall server CPU market.[4][7][10]

What it depends on, and what depends on it

AI data center demand is currently the dominant driver of server CPU volume. Server CPU unit shipments rose more than 10 percent year-on-year in Q1 2026, driven almost entirely by AI data center demand, while Intel's server shipments stayed flat both sequentially and year-on-year over the same period. Inside the installed base, high core counts in the 65-to-128-core range controlled 32 percent of the server CPU market in 2025, fueled by AI inference, cloud-native applications, virtualization, and high-performance computing. Cloud and virtualization workloads together represented 35 percent of global server CPU demand that year.[1][8][11]

Where it sits in the stack

What would break it

Substitution risk is real but constrained by software. ARM-based server CPUs from Ampere and Nvidia, RISC-V designs, and custom ASICs can deliver better energy efficiency and sustainability than x86 when paired with the right workloads and software optimization, but they require recompilation or porting of the x86-targeted enterprise stack. The result is that ARM servers are most visible inside hyperscaler-designed racks such as Nvidia's GB200 and GB300 platforms, while mainstream enterprise AI deployments continue to use x86 hosts. Any acceleration of x86-to-ARM or x86-to-RISC-V migration inside the largest cloud providers would erode the x86 duopoly's volume base, but the migration rate is currently limited by the cost of qualifying new platforms across multi-year hardware refresh cycles.[2][10][7]

Related nodes

Sources

1. tweaktown.com · 2026-06-05T23:15:05
2. datacenterknowledge.com · 2026-03-25T09:00:00
3. grokipedia.com · 2026-01-17T04:27:23
4. tomshardware.com · 2026-05-22T14:30:00
5. en.wikipedia.org · 2026-06-21T22:32:05
6. tomshardware.com · 2026-05-14T10:40:00
7. newsletter.semianalysis.com · 2025
8. theregister.com · Q1 2026
9. theregister.com · 2025-05-27
10. dell.com
11. acumenresearchandconsulting.com · 2025