AMD Market Share Reversal: x86 Servers Dominance Hits 46% by Q1 2026

On 2026-05-29, Mercury Research published its Q1 2026 x86 CPU tracker. AMD’s server CPU revenue share landed at 46.2%, a record, up from 41.3% the prior quarter. In the same window, AMD’s own data center segment revenue, $5.8B for the quarter, climbed past Intel’s Data Center and AI revenue of roughly $5.1B for the first time in either company’s history. The AMD market share reversal in x86 servers dominance, a decade of percentage-point swings inside a roughly $22B-a-year market, has now flipped the revenue lead between the two incumbents.

The non-obvious part is that it happened the same quarter Arm crossed about half of hyperscaler CPU compute share, confirmed onstage at Computex 2026 the week of June 2.

Key facts at a glance

• Q1 2026 x86 server CPU revenue share: AMD 46.2%, Intel 53.8% (Mercury Research, via secondary press).
• Q1 2026 server CPU unit share: AMD 33.2%, Intel 66.8%. Intel ships more chips; AMD earns more per chip.
• AMD broke the 10% server share barrier in Q1 2021 after years in low single digits.
• AMD Q1 2026 data center segment revenue: $5.8B (+57% YoY), passing Intel DCAI’s ~$5.1B for the first time.
• Arm-based silicon crossed roughly 50% of hyperscaler CPU compute share in FY26, confirmed at Computex 2026 on 2026-06-02.

How x86 server share is actually measured

Two numbers matter, and they tell different stories.

Unit share counts chips shipped in a quarter. Revenue share counts dollars billed. In server CPUs the two routinely diverge because high-core-count parts can list at well over $10,000 a socket while low-end SKUs sell for a few hundred. A vendor can hold the smaller unit slice and still take the larger revenue slice if its mix skews to the expensive top-bin parts.

Mercury Research is the share publisher both AMD and Intel reference in their own investor materials. The methodology is private, so the same number gets re-reported across Mercury Research subscribers without an independent re-derivation.

A worked example: AMD’s climb from sub-5% to 46.2%

For most of the 2010s, AMD’s server share was a rounding error. Bulldozer-era parts could not compete on perf-per-watt with Intel Xeon, and the hyperscalers were standardized on Intel platforms.

The first inflection landed in 2017, when AMD shipped the first EPYC generation on Zen. Even then the share needle moved slowly. The Next Platform tracked the moment AMD finally crossed the 10% unit share barrier in Q1 2021, ending years stuck in low single digits (The Next Platform).

Four years later, TechPowerUp flagged AMD crossing 40% server revenue share for the first time. Q4 2025 came in at 41.3%. Q1 2026 jumped to 46.2%, reported by Tom’s Hardware and TweakTown. AMD’s own segment-level disclosure, framed at length in the Q4 2024 earnings supplement on EDGAR, walked through the data center segment as the company’s primary growth engine (AMD Q4 2024 earnings slides).

Period	AMD revenue share	AMD unit share	Note
2018-2019	low single digits	low single digits	Pre-EPYC ramp
Q1 2021	~10%	~10%	First time crossing 10%
Q4 2025	41.3%	not specified	Sustained EPYC adoption
Q1 2026	46.2%	33.2%	Record; revenue gap reflects high-core-count mix

The shape of the climb matters. AMD did not jump to the top of the chart in one product cycle. Five generations of EPYC, each landing a TSMC node and a core-count uplift ahead of Intel’s roadmap, compounded into a roughly 40-point share swing inside a $22B-a-year market.

Why this matters, and the part most coverage misses

The AMD-versus-Intel scoreboard is the easy story. The harder one is what x86 itself is doing inside the cloud.

At Computex 2026, Arm’s CFO Jason Child publicly confirmed Arm crossed roughly 50% of hyperscaler CPU compute share, with corroboration from Patrick Moorhead and Omdia (TechTimes from Computex 2026). AWS Graviton now powers the majority of new EC2 capacity. Google’s next-generation TPU pods replace the x86 host CPU with custom Arm-based Axion. Microsoft Cobalt and Meta MTIA follow the same pattern.

AMD’s near-monopoly inside x86, sitting at 46.2% revenue and climbing, is a near-monopoly of a category whose four largest customers (AWS, Google, Microsoft, Meta) have published roadmaps to substitute merchant CPUs with their own Arm-based parts. The durable threat to AMD’s lead is not an Intel comeback. It is the hyperscalers turning into competitors and shrinking the addressable merchant x86 TAM that AMD just conquered. The Register covered the same structural shift after the prior Mercury print.

AMD itself read the room early. The November 2025 strategy briefing reframed the company’s addressable market as a $1T compute opportunity that explicitly includes GPUs, networking, and accelerators (AMD November 2025 strategy), not just x86 sockets.

Common misconceptions

• “AMD has 99.9% of x86 server revenue.” This figure has circulated in podcast commentary in early June 2026. No public Mercury Research print approaches that number. The verified Q1 2026 figure is 46.2% revenue share.
• “Intel ever owned 100% of x86 servers.” Mercury Research’s historical series shows Intel topped in the high 90s. AMD always retained a low-single-digit slice, even at its 2018 low.
• “Unit and revenue share are the same.” In Q1 2026 they differed by 13 percentage points (33.2% vs 46.2%). Mix matters.
• “AMD overtaking Intel in data center means Intel is finished.” It does not. Intel still holds the larger unit share in servers and roughly 70% of consumer PC CPU share. The competitive picture is asymmetric across segments.

What this does not tell you

• Mercury Research’s sampling methodology is private. The Q1 2026 numbers are the industry standard but cannot be independently re-derived from public data. [NEEDS RESEARCH: cross-check Omdia, IDC, and Counterpoint server CPU trackers once their Q1 2026 prints land.]
• Forward durability of AMD’s lead is inferential, not measured. Intel’s Diamond Rapids has slipped to 2027 while Clearwater Forest (288 efficiency cores on 18A) is now shipping. Either roadmap event could move the next print materially.
• Custom Arm silicon adoption inside hyperscalers is not visible in Mercury’s x86 series because those parts are not x86. The x86 TAM compression from in-house parts at AWS, Google, Microsoft, and Meta is a known direction with no clean public quantification. [NEEDS RESEARCH: a public estimate of merchant x86 server TAM compression from top-four hyperscaler in-house parts.]
• AMD’s mid-2010s low point in server share was in the low single digits, not zero.

FAQ

What does “x86 server CPU revenue share” actually measure?

The dollar value of server-class x86 processors sold in a quarter, split by vendor. Mercury Research is the canonical source. Unit share is the count of chips shipped, which is a different number because server CPUs vary widely in price.

Why did AMD overtake on revenue before it overtook on units?

AMD’s EPYC line skews toward high-core-count parts with average selling prices several times higher than the mainstream Xeon mix. A 192-core Turin processor lists at multiples of an 8-core Xeon. AMD captured more revenue per chip even while shipping fewer total chips.

Is Intel finished in servers?

No. Intel still holds 53.8% of x86 server CPU revenue and a larger share of units in Q1 2026. The unit lead means Intel still dominates lower-end and edge server tiers. Diamond Rapids slipping to 2027 hurts, but Clearwater Forest with 288 E-cores on 18A is now shipping.

Could Arm and custom silicon really displace x86 in servers?

The trend is already visible. Arm’s CFO confirmed roughly 50% hyperscaler CPU compute share at Computex 2026. AWS reports the majority of new EC2 capacity runs on Graviton. Whether that becomes a permanent ceiling for x86 or a slow decline depends on hyperscaler buying behavior, which is not externally visible quarter to quarter.

Where can I read AMD’s own disclosure?

AMD files segment revenue in its 10-Q and quarterly earnings supplements. The Q4 2024 earnings slides on EDGAR walk through the data center segment as AMD reports it (AMD Q4 2024 earnings slides).