AMD's data center revenue surpasses Intel for the first time

① Data center revenue reached $5.8 billion, surpassing Intel for the first time: up 57% year-over-year to $5.8 billion, with server CPU revenue growing over 50% to a record high. During the same period, Intel’s DCAI revenue was $5.1 billion, up 22%—AMD’s data center revenue scale has surpassed Intel’s for the first time.

② The server CPU TAM has doubled from $60 billion to $120 billion: The total addressable market (TAM) for server CPUs in 2030, estimated at approximately $60 billion during the analyst day in November last year, has doubled in just five months. This growth is driven by Agentic AI, which significantly increases demand for CPUs—the ratio of CPUs to GPUs is shifting from 1:4 or 1:8 toward 1:1 or even higher CPU proportions.

③ Meta’s 6GW+OpenAI Multi-Generation Contract Locks in Long-Term Visibility: Meta plans to deploy up to 6 gigawatts of AMD Instinct GPUs, including custom accelerators based on the MI450 architecture; collaboration with OpenAI continues to advance. Management stated that AI revenue from data centers will reach the "hundreds of billions of dollars" level by 2027.

④ GPU revenue declined quarter-over-quarter in Q1 due to the transition in China: Data center AI revenue saw a slight sequential decrease due to a significant drop in China-related revenue from Q4. True GPU volume ramp-up begins in Q3 (with Helios initial capacity), significantly escalates in Q4, and continues through Q1 2027.

⑤ Free cash flow tripled to a record $2.6 billion: FCF margin reached 25%, with operating cash flow rising from $940 million in the same period last year to $3 billion. However, the ramp-up of MI450 in the second half will dilute gross margin—this product is below the company’s average.

⑥ Consumer demand faces pressure in the second half: Memory price increases impact PCs and gaming—Management expects gaming revenue in the second half to decline by more than 20% compared to the first half, and PC shipments will also be affected by rising memory and component costs.

The most striking aspect of this earnings report is not the 38% revenue growth, but a structural inflection point: AMD’s data center quarterly revenue reached $5.8 billion, surpassing Intel’s DCAI revenue of $5.1 billion during the same period for the first time. From the debut of AMD’s Zen architecture in the server market in 2017 to today, AMD has taken nearly a decade to achieve this historic milestone.

But the other side of this earnings report is equally worth examining. The data center AI business (i.e., Instinct GPUs) actually saw a slight sequential decline in Q1—due to significantly higher revenue from China in Q4 and a sharp drop in Q1. This indicates that AMD’s “true breakout period” in the AI accelerator market has yet to arrive. Lisa Su has clearly outlined the production timeline for Helios (the full-rack solution combining MI450 and Venice) to begin in Q3 and scale significantly in Q4—investors will need to wait patiently for delivery proof in the second half of the year.

Meanwhile, the story around server CPUs is becoming increasingly compelling. The TAM doubling from $60 billion to $120 billion is a figure worth pausing to absorb. Management’s logic chain is clear: the adoption of agentic AI → explosion in inference demand → each agent requires a CPU to orchestrate and process data → the CPU-to-GPU ratio rises from 1:4 to 1:1 or higher → server CPU TAM doubles. Server CPU revenue guidance for Q2 shows year-over-year growth exceeding 70%, with the full-year growth trajectory accelerating further.

Here is a detailed analysis of the financial report.

AMD reported first-quarter revenue of $10.3 billion, a 38% year-over-year increase, exceeding the upper end of guidance. Non-GAAP gross margin was 55%, up 1 percentage point year-over-year; operating profit was $2.54 billion, up 43% year-over-year, with an operating margin of 25%. On a GAAP basis, operating profit was $1.5 billion (+83% YoY), net income was $1.4 billion (+95% YoY), and EPS was $0.84 (+91% YoY). The primary differences between GAAP and Non-GAAP results stem from amortization of acquired intangible assets ($551 million) and stock-based compensation expenses ($487 million).

Data center revenue reached $5.8 billion, a 57% year-over-year increase and a 7% sequential increase, setting a new all-time high. Operating profit was $1.6 billion, with a margin of 28%, expanding by 3 percentage points year-over-year.

The server CPU business was the standout highlight—marking the fourth consecutive quarter of revenue records, with year-over-year growth exceeding 50%. Both cloud and enterprise customers grew by over 50%, and the number of EPYC-powered cloud instances increased nearly 50% year-over-year to over 1,600. Growth was driven by the ramp-up of fifth-generation EPYC Turin and continued shipments of fourth-generation Genoa. Lisa Su specifically noted that the growth was primarily driven by volume increases rather than price hikes, with ASP improvements stemming largely from product mix and higher core counts.

More importantly, the redefinition of TAM. At the analyst day last November, the estimated server CPU TAM for 2030 was approximately $60 billion (18% CAGR), but just five months later, management doubled this figure to over $120 billion (>35% CAGR). The rationale is that agentic AI is reshaping the demand equation for CPUs—previously, CPUs served merely as a "supporting role" (head node) to GPUs, with ratios of 1:4 or 1:8; now, agentic workflows require substantial CPU resources to handle orchestration, data processing, and parallel tasks, shifting the ratio toward 1:1 or even higher.

The data center AI (Instinct GPU) segment is in a "building momentum" state. Q1 saw a slight sequential decline, primarily due to higher China revenue in Q4 and a reduction in Q1. However, year-over-year growth remains in "significant double digits." The MI355X demonstrated comprehensive competitive performance in the latest MLPerf benchmarks. The MI450 series GPUs have begun sampling with leading customers, with H2 production timelines set for initial capacity in Q3, significant ramp-up in Q4, and further acceleration in Q1 2027. Management noted that demand forecasts from leading customers have exceeded initial plans, and the pipeline for large-scale deployments from new customers is expanding, including multiple "gigawatt-scale" opportunities.

The horizontal comparison highlights AMD’s momentum. Intel’s DCAI revenue during the same period was $5.1 billion, growing at 22%—marking the first time AMD’s data center revenue has surpassed Intel’s and representing a milestone victory for AMD in the server market since the launch of its Zen architecture in 2017. Intel’s counterpoints include the selection of Xeon 6 as the host CPU for NVIDIA DGX Rubin NVL8 and its multi-year partnership with Google. However, based on growth rates and market share trends, AMD’s advantage continues to expand, with management reaffirming its target of over 50% market share in server CPUs.

In the AI accelerator market, NVIDIA still dominates with approximately 75-80% market share, but AMD is positioning itself as the second-largest player through strategic contracts with Meta (6GW) and OpenAI.

Client and gaming revenue reached $3.6 billion, up 23% year-over-year and down 9% quarter-over-quarter (seasonal). Operating profit was $575 million, with a margin of 16%.

Client business revenue reached $2.9 billion, a 26% year-over-year growth driven by strong shipments of Ryzen processors and continued market share gains. Commercial PCs were the standout highlight of the quarter—sales of Ryzen PRO PCs increased by over 50% year-over-year, with Dell, HP, and Lenovo expanding their AMD product lines. New enterprise customer wins spanned the financial, healthcare, industrial, and digital infrastructure sectors.

Game business revenue reached $720 million, a 11% year-over-year increase, primarily driven by demand for the Radeon 9000 series GPUs, partially offset by a decline in semi-custom (console) revenue.

However, management issued a clear warning for the second half of the year: gaming revenue is expected to decline by more than 20% compared to the first half due to rising memory and component costs, and PC shipments will also be affected. This dynamic stems from surging demand for HBM and DDR5 driven by AI, which has elevated all memory prices and indirectly pressured consumer electronics. Nevertheless, management still expects client business to grow year-over-year for the full year and outperform the market.

Embedded revenue was $873 million, up 6% year-over-year but down 8% quarter-over-quarter. Operating profit was $338 million, with a margin of 39%. Growth was driven by improved demand in areas such as test and measurement, aerospace, and communications. Design wins increased by double digits year-over-year, worth billions of dollars, reflecting the expansion of the embedded business from a FPGA-centric portfolio to a broader mix of adaptive embedded x86 and semi-custom solutions.

Free cash flow of $2.6 billion, more than triple the year-ago period, with an FCF margin of 25%. Operating cash flow of $3.0 billion (vs. $940 million in the same period last year), driven by higher net income and improved working capital efficiency.

Capital expenditures amounted to $389 million (vs. $212 million year-over-year), representing an 83% increase, though the absolute amount remains relatively low—AMD’s fabless model means capacity investments are primarily made by supply chain partners. Research and development expenses totaled $2.4 billion, up 39% year-over-year and representing 23% of revenue, reflecting accelerated investment in the AI roadmap.

On the balance sheet, cash and short-term investments amounted to $12.3 billion, inventory was $8 billion (essentially flat), and long-term debt stood at $2.4 billion. In Q1, the company repurchased 1.1 million shares ($221 million), with $9.2 billion remaining under the authorized repurchase program.

One-time items to note: In Q4 '25, there was $280 million in long-term investment gains (boosting Q4 EPS), compared to only $66 million in Q1 '26, contributing to the Non-GAAP EPS decline from $1.53 (Q4) to $1.37 (Q1). Additionally, ZT Systems’ manufacturing business was divested in Q4, resulting in a $11 million gain from discontinued operations in Q1.

Su Ma outlined a clear growth path during the earnings call:

Short-term (Q2 2026): Revenue guidance of $11.2 billion (+46% YoY, +9% QoQ), with gross margin improving to 56%. Server CPUs grew more than 70% year-over-year, and both data center AI and servers posted double-digit quarter-over-quarter growth.

Medium-term (H2 2026–2027): Helios mass production is the key catalyst. The MI450 series GPUs begin shipping in Q3 and see significant volume ramp-up in Q4. Venice (6th-gen EPYC) is set to launch this year, with more customers than any prior generation in validation and ramp-up phases. Management has expressed "strong and growing confidence" that data center AI revenue will reach "tens of billions of dollars" by 2027.

Long-term: AI GPU CAGR target raised above 80% (exceeding the previous target). Long-term EPS target of $20+ (current annualized ~$5.5, implying nearly fourfold growth potential). Long-term gross margin target of 55%-58%.

Execution risks and timelines for Helios mass production. The MI450+Venice rack-scale solution represents AMD’s bet in the AI accelerator market, and progress in mass production during Q3–Q4 will directly determine the revenue trajectory for 2027. Contracts with Meta and OpenAI provide demand anchors, but significant execution risks remain between “sample delivery” and “scalable shipments.” Key indicators to watch: quarter-over-quarter change in DC AI revenue in Q3 (expected to rebound significantly) and any announcements of new large-scale customers. If Helios shipments in Q3/Q4 meet expectations, AMD could elevate its DC AI revenue from approximately $10 billion to a $20–30 billion range by 2027.

Where is the ceiling for CPU market share? A 50% market share target under a $120 billion TAM implies $60 billion in annual revenue—this would make server CPUs a larger business than AMD is today. Intel is countering with Xeon 6 and Granite Rapids, while the Arm camp (AWS Graviton, Ampere, and various in-house chips) is rising. Lisa Su’s response is that the market is large enough to accommodate different CPUs for different workloads, and AMD’s “full portfolio” approach (general-purpose, head node, agentic-optimized) offers a strategic advantage over Arm’s “single-point products.” Whether this argument holds will depend critically on customer validation progress with Venice.

The impact of rising storage prices on consumer-facing businesses remains a secondary concern requiring ongoing attention. A decline in gaming revenue of more than 20% in the second half of the year and pressure on PC shipments may dampen overall revenue growth. However, given that data centers now account for over 56% of total revenue and are growing significantly faster than the consumer segment, this drag is structurally diminishing.