CPU Price Per Core Index 2026: Xeon vs EPYC

The Non-Linear Reality of Server CPU Price Per Core in 2026

Infrastructure engineering has historically operated on assumptions of linear hardware pricing: doubling core density was expected to track a predictable multiple of platform spend. In mid-2026, empirical analysis of manufacturer 1,000-unit (1kU) list pricing confirms that this linear relationship has collapsed. Modern server architectures demonstrate significant divergence between entry-level silicon, mid-stack workhorses, and high-density scale-out bins.

Pricing behavior diverges sharply between the two silicon vendors. Intel structures its Xeon stack with steep density premiums on flagship parts, where buyers paying for maximum per-socket core counts face acquisition costs approaching twice the per-core expense of mid-stack alternatives. Conversely, AMD leverages dense 'Zen 5c' chiplet packaging to make its highest core-count processors, such as the 256-core EPYC 9996, its most economical enterprise silicon per core. To make rational hardware investments, engineering leads must normalise published tray pricing across architectural tiers before designing virtualisation clusters or scale-out nodes.

For UK enterprise buyers, this evaluation requires rigorous baseline modelling. Before factoring in local distribution margins, enterprise discounts, and Value Added Tax, published manufacturer price sheets provide an objective benchmark. By evaluating physical core counts against vendor list prices, procurement teams can identify the exact inflection points where silicon density ceases to be economically sound.

• •Vendor strategies diverge at the top bin: Intel charges steep density premiums on high-end Xeons, whereas AMD's densest EPYC chips offer its lowest per-core enterprise pricing.
• •Mid-stack processors frequently provide the most advantageous per-core economics across both major architectures.
• •Manufacturer 1,000-unit list prices serve as the baseline for normalising core costs before localised channel discounting.

Intel Xeon Analysis: The Impact of the July 2026 Repricing

In July 2026, Intel executed a significant price adjustment across its enterprise Xeon portfolio, altering recommended customer pricing (RCP) across modern and previous-generation architectures. The changes fundamentally shifted per-core economics for data centre platforms, reinforcing the cost disparity between general-purpose enterprise silicon and specialised high-density parts.

The flagship 128-core Xeon 6980P saw its list price increase from $12,460 to $13,955. Prior to this adjustment, as reported during earlier market tracking, the processor delivered 128 cores at approximately $97.00 per core. Following the July 2026 update, the unit stands at approximately $109.02 per core. Similarly, the Xeon 6978P experienced an upward revision in reported recommended customer pricing, moving from $11,025 to a new RCP of $12,348.

The 64‑core Xeon 8592+, widely deployed in dual‑socket enterprise platforms, was repriced from $11,600 up to $12,992. When divided across its 64 physical cores, the updated list price yields roughly $203.00 per core. This places the processor among the most expensive mainstream options on a raw per-core metric, illustrating the steep financial commitment required for mature socket designs compared to ultra-dense high-core architectures. For teams seeking further insights into EPYC Turin vs. Xeon 6, this price repositioning represents a critical turning point in total infrastructure cost modelling.

AMD EPYC Stack Evaluation: From EPYC 4004 to Dense 9006 SKUs

AMD's enterprise and entry-tier server catalogue spans distinct microarchitectural tiers, exhibiting dramatic price-per-core shifts across single-socket edge parts, high-density cloud-native chips, and flagship high-performance processors. Evaluating published 1kU tray prices exposes where AMD concentrates its aggressive silicon value.

At the entry tier, the single-socket AM5-based EPYC 4004 series delivers some of the lowest published cost per core levels in the current x86 server CPU market. The 12-core EPYC 4464P lists at $429, translating to a very low $35.75 per core by current published server CPU list prices. Moving to the 16-core EPYC 4564P at $699 raises the unit cost slightly to $43.69 per core. For remote offices, dedicated hosting, and lightweight network appliances, these entry-level parts undercut conventional dual-socket enterprise silicon by a massive margin.

Within the dense compute space, AMD's specialised 8005 and 9004/9005 series present a nuanced progression. The 16-core EPYC 8125P carries an official list price of $799, yielding $49.94 per core. Scaling to higher core counts, the 84-core EPYC 8635P lists at $5,799, resulting in $69.04 per core. Meanwhile, the previous-generation 128-core EPYC 9754 lists at $10,631, representing $83.05 per core.

At the highest tier of scale-out compute, AMD's 9005 and 9006 series balance extreme density against overall socket price. The 96‑core EPYC 9655 has an official AMD 1k‑unit list price of $8,999 (about $93.74 per core). The 192‑core EPYC 9965 is officially listed on AMD’s product page at a 1k‑unit price of $11,988, which yields $62.44 per core, while earlier Tom’s Hardware coverage in 2025 had reported a higher historical list price of $14,813. Reaching the pinnacle of socket density, the 256-core EPYC 9996 from the 9006 series lists at $14,904. Despite its substantial absolute price tag, distributing that cost across 256 physical cores yields $58.22 per core—substantially lower than the flagship Xeon 6980P at $109.02 per core.

Cross-Stack Comparison: Exposing the Mid-Stack Sweet Spots

When tabulating the entire competitive landscape across Intel and AMD, data proves that core costs do not follow a simple ascending or descending slope. Instead, the market forms distinct pricing plateaus governed by target workload and physical packaging. Enterprise buyers looking to determine how many CPU cores your server needs must look past top-line marketing specifications to evaluate physical compute density against capital outlay.

The absolute lowest raw cost per core resides in dedicated entry-level silicon such as the EPYC 4464P ($35.75) and EPYC 4564P ($43.69), making this tier optimal for edge appliances, branch office (ROBO) workloads, and dedicated hosting where single-socket AM5 constraints fit the brief. For scale-out cloud platforms and consolidated virtualisation hosts requiring maximum socket density, ultra-dense parts like the 256-core EPYC 9996 ($58.22 per core) and 192-core EPYC 9965 ($62.44 per core) deliver unmatched multi-tenant efficiency. Meanwhile, mainstream general-purpose silicon like the 96-core EPYC 9655 ($93.74 per core) is targeted at balanced enterprise virtualisation with higher per-core memory bandwidth.

On the Intel side, the post-July 2026 pricing establishes an effective floor near $109.02 per core for the 128-core Xeon 6980P, well-suited to dense AI inferencing and HPC clusters requiring Intel AMX engines, while mid-density legacy sockets like the 64-core Xeon 8592+ ($203.00 per core) cater primarily to established dual-socket database and ERP refreshes where re-qualifying platforms is prohibitive. However, procurement teams must recognize that raw price-per-core is not a complete buying signal: it ignores IPC, clock speed variations, and per-core throughput differences between Intel P-cores, standard Zen 5, and dense Zen 5c architectures. For balanced architectural planning, reviewing a detailed comparison of Xeon 6 vs EPYC Turin provides the essential performance-normalised context to evaluate whether cheaper per-core silicon meets workload SLAs.

Total Cost of Ownership: Power, Sockets, and Software Licensing

Procurement calculations that stop at raw hardware list prices fail to capture total cost of ownership (TCO). In mid-2026 data centre operations, server CPU capital expenditure represents only the starting point. Core counts directly dictate enterprise hypervisor and database licensing expenses, where commercial software models charge per physical core.

Platforms licensed on a per-core basis, such as standard hypervisors and relational database management systems, penalise high-core deployment strategies if workloads do not fully utilise that compute capacity. Deploying a 256-core EPYC 9996 or a 128-core Xeon 6980P triggers licensing fees that dwarf the initial hardware purchase. In these environments, higher per-core processor costs on lower-core SKUs can ultimately yield lower total operational expenditure.

Furthermore, thermal design power and socket density heavily impact physical facility overhead. Consolidating workloads onto fewer, denser sockets reduces chassis count, rack units, power distribution hardware, and ongoing facility cooling. When organisations learn how to spec a server effectively, they must weigh the $58 to $109 per-core silicon price against multi-year datacentre power contracts and facility infrastructure.

The UK Reality Check: Currency Conversion, Channel Margins, and VAT

All official manufacturer price lists from Intel and AMD are published in US dollars based on 1,000-unit tray quantities. For UK enterprise organisations, translating these US‑dollar list figures into realistic hardware budgets typically requires adjusting for foreign‑exchange fluctuations, distributor import margins, and local taxation.

Published 1kU figures represent tray pricing offered to high-volume server OEMs in US dollars and exclude local channel costs. Translating these figures into realistic UK hardware budgets requires applying a spot foreign exchange rate, an estimated distributor margin (typically 6% to 10% for specialised enterprise silicon), and standard 20% UK VAT. Assuming an indicative baseline exchange rate of £1 = $1.30 (approx. £0.77 per USD) and an 8% channel markup, the landed UK cost shifts significantly from raw US price cards.

For Intel's 128-core Xeon 6980P ($13,955 list), converting at £1 = $1.30 yields a base of ~£10,735. Applying an 8% distributor markup brings pre-tax cost to ~£11,594 (£90.58/core ex-VAT), and adding 20% VAT lands the processor at approximately £13,913, or ~£108.70 per core delivered. By comparison, AMD's 256-core EPYC 9996 ($14,904 list) converts to ~£11,465 base, reaches ~£12,382 ex-VAT with the 8% markup (£48.37/core), and totals roughly £14,858 inc-VAT—landing at an effective £58.04 per core. While enterprise rebates from tier-one OEMs alter net pricing, this worked arithmetic gives UK infrastructure leads a grounded benchmark for rack budgeting.

Methodology

This dataset and price-per-core index was compiled in mid-2026 by aggregating publicly available manufacturer price lists and documented enterprise customer pricing updates. AMD figures were gathered directly from official AMD product technical specifications and published 1,000-unit tray pricing sheets across the EPYC 4004, 8005, 9004, 9005, and 9006 processor families.

Intel Xeon data reflects tracking of recommended customer pricing (RCP) following the documented July 2026 price revisions, using figures reported across enterprise hardware analysts—including adjustments captured by The Register, Server Configurator, and historical baseline reporting from Tom’s Hardware. Price-per-core calculations derive exclusively from dividing the verified manufacturer 1kU list price in US dollars by the physical core count of each respective processor.

All calculated figures exclude value-added tax, channel distribution markups, local freight, system chassis integration, memory, and software licensing. Figures represent physical core metrics and intentionally omit hyperthreading and virtualisation thread counts to maintain objective hardware comparisons.

Strategic Hardware Sourcing and TCO Summary

The mid-2026 server processor market presents distinct value sweet spots for informed enterprise architects. Scale-out cloud infrastructure and multi-tenant virtualization clusters achieve peak capital efficiency with AMD's ultra-dense Zen 5c parts, led by the 256-core EPYC 9996 and 192-core EPYC 9965, while branch offices and edge appliances benefit from entry-level AM5 silicon. Conversely, enterprises committed to Intel ecosystems must absorb the July 2026 repricing on Xeon 6 flagships and factor in significant legacy platform premiums on parts like the Xeon 8592+. To maximise capital allocation, procurement teams must balance these raw silicon benchmarks against core-based software licensing, rack thermal limits, and real-world per-thread performance before finalising server orders.

Conversely, enterprises committed to Intel ecosystems must account for the July 2026 repricing, which elevated flagship parts like the Xeon 6980P to $13,955 ($109.02/core) and previous-generation enterprise workhorses like the Xeon 8592+ to $12,992 ($203.00/core). To maximise capital allocation, buyers should balance raw price-per-core metrics against enterprise application licensing models, rack thermal limits, and validated vendor supply chains before finalising server procurement.

Sources

Every figure in this article traces to the sources below.

• •AMD — EPYC 9996 Technical Specifications and 1kU List Pricing
• •AMD — EPYC 9965 Product Details and Tray Pricing
• •AMD — EPYC 9655 Specifications and 1k-Unit Pricing
• •AMD — EPYC 8635P and 8125P Server Processor Catalogue
• •AMD — EPYC 4004 Series Pricing and Core Counts
• •The Register / Kad8 — Reporting on Intel July 2026 Xeon Repricing Updates
• •Server Configurator — 2026 Server CPU Price Per Core Index Statistics
• •Tom's Hardware — Enterprise CPU Market Analysis and Historical 1kU Baseline Tracking