AMD’s data-center surge contrasts with Intel’s increasingly shaky technology roadmap

Desktop dominance gives way to a data-center reversal
Competition reshaped across PCs, mobile, and AI
Intel enters a survival phase—its next move becomes pivotal

AMD posted record data-center revenue in the third quarter, drawing sharp contrast with Intel’s continued weakness. The result stems from AMD’s “full-stack strategy,” which combines hardware with its own software platform. During the same period, AMD also secured over 30 percent market share in desktops. Meanwhile, ARM and Qualcomm are expanding their presence with their respective product lineups, adding fractures to the competitive landscape. Coupled with major customers’ accelerated adoption of AI accelerators, this shift is spreading across the entire server market.

Growing preference for AMD as enterprises refresh servers

According to Reuters on the 23rd (local time), AMD’s data-center division posted revenue of 4.3 billion dollars in the third quarter, a 22 percent increase from a year earlier. Intel’s data-center division, by contrast, saw revenue fall 1 percent to 4.1 billion dollars during the same period. AMD’s Instinct MI300 accelerator series, which reached server-class revenue parity with Intel just six months after launch, drove the strong performance. The industry broadly views AMD as having solidified its growth footing as the AI accelerator market—once dominated by Nvidia—undergoes structural realignment.

AMD’s momentum stemmed from its full-stack strategy. While Intel maintained a “department-store-style” hardware lineup, AMD bundled its CPUs and GPUs with its ROCm (Radeon Open Compute) software platform, offering developers a unified ecosystem. This strategy aimed to lock customers into a long-term platform and led Microsoft, Meta, and other major tech companies to adopt Instinct accelerators.

In desktops, AMD’s gains are equally pronounced. Mercury Research data shows AMD reached a record 33.6 percent desktop market share in Q3, up 5.2 percentage points year-over-year. Explosive demand for the 9800X3D—touted as the “fastest gaming CPU available”—played a central role. Intel, however, failed to catch up with its poorly received Core Ultra 200 and continues to rely heavily on sales of its three-year-old Raptor Lake generation. Combined with the data-center reversal, these trends underscore the market’s center of gravity shifting toward AMD.

AMD also announced a joint venture with Cisco and HUMAIN, a Saudi Public Investment Fund (PIF) portfolio company offering full-stack AI solutions, to build up to 1 gigawatt of AI infrastructure by 2030. The company is already progressing with a 100-megawatt AI infrastructure project, laying groundwork to absorb future data-center demand within its own ecosystem. This aligns with AMD’s broader effort to transition customers approaching contract expiration with Intel into its platform by expanding from desktops into servers and AI infrastructure.

Intel pushed into a defensive posture on all fronts

Intel now faces simultaneous pressure: AMD in PCs, ARM in mobile and servers. According to Bank of America, Intel’s Q3 PC and server unit growth was just +2 percent and –1 percent quarter-over-quarter, while AMD recorded +10 percent and +1 percent, and ARM registered +7 percent and +16 percent. The disparity reflects how AMD and ARM expanded share through stronger product availability and expanded core configurations while Intel faced supply constraints across its process-based product lines.

ARM’s momentum was particularly strong in servers. ARM-based server shipments rose 16 percent quarter-over-quarter in Q3. Some analysts claimed Intel’s supply shortages boosted ASPs and offered temporary benefits, but overall growth continued to lag AMD and ARM. The combined impact of process delays, missteps in Intel’s foundry strategy, and product-line confusion intensified competitive pressure across PCs, mobile, and servers.

Qualcomm’s rise in AI PCs has further complicated Intel’s environment. Its Snapdragon X2 Elite series, launching early next year, features an Orion-based CPU architecture with up to 18 cores and GPU performance 2.3 times higher than the previous generation. With a power range spanning 10 watts to over 100 watts, the lineup covers everything from ultraportables to high-performance machines. Qualcomm’s efficiency-first approach and battery-sustained performance model directly threaten Intel’s traditional strongholds.

Roadmap disruption amid weakened R&D

A larger problem is Intel’s inability to present a clear path forward. The company recently halted development of the 8-channel version of its next-generation AI accelerator processor, Diamond Rapids. Intel effectively eliminated the successor to Granite Rapids (Xeon 6700P), retaining only the 16-channel high-bandwidth platform. Given the strong value proposition of the 8-channel Xeon 6700P, many interpret this move as Intel sacrificing its best-selling lineup in favor of uncertain future demand.

Intel’s leadership argues that a fragmented product stack slowed development speed. Consolidating everything onto a single 16-channel platform is intended to shorten engineering cycles and cover both high-end and mid-tier segments with one socket and one platform. An industry source noted that “data-center performance has shifted from CPU-core competition to memory-bandwidth competition,” adding that Intel is abandoning its old playbook to adapt to a high-bandwidth AI era.

Intel is also relying more heavily on external capital and government support. One notable example is Intel’s agreement to design and manufacture a custom Xeon x86 processor exclusively for Nvidia in exchange for 5 billion dollars. This places Intel CPUs in a subordinate role to Nvidia GPUs and reflects a shift toward prioritizing stable demand over market leadership.

The foundry division remains under pressure. Intel admitted that its earlier failure to secure external customers during new process development caused process optimization to skew toward internal products. To avoid repeating this mistake, the company plans to involve external customers earlier in the 14A process definition stage. But increased upfront investment has delayed the breakeven point to late 2027—nearly a year later than previously expected. Intel maintains that “a longer loss period is painful but reflects the reality of securing real customers and building actual production lines.”