CPUs on a tear, Helios near—Is AMD finally turning the corner?
AMD (AMD.O) released its Q1 2026 results (to Mar 2026) after the U.S. market close on May 6 Beijing time. Key points:
1) Overall results: $AMD(AMD.US) reported revenue of $10.25bn in Q1 2026, up 37.8% YoY, beating the Street (~$9.9bn). Growth was driven by Data Center and Client. GAAP GPM was 52.8% (+260bps YoY), helped by mix shift toward higher-margin Data Center.
2) Opex: R&D was $2.4bn (+38.7% YoY), and SG&A was $1.25bn (+41% YoY). Core opex growth tracked revenue, keeping the core opex ratio around 35%.
Net income was $1.38bn, largely affected by non-recurring items. On an operating basis, core OP was $1.77bn (+57% YoY) with a 17% core OPM.
3) Segment mix: Data Center and Client together accounted for over 80% of revenue.
1) Client: gaining share, with revenue of $2.89bn, up 26% YoY. Global PC shipments rose 3.8% YoY, yet AMD’s Client revenue grew much faster, reflecting continued share gains in PCs.
2) Data Center: server CPU in an upcycle; MI450 anticipated. Q1 revenue was $5.78bn, up 7% QoQ, driven mainly by tight supply-demand for server CPUs.
i) Server GPU: AI GPU revenue dipped slightly QoQ due to adjustments in China, including roughly $390mn from last quarter’s MI308 shipments to China.
Beyond near-term prints, the market is focused on the MI450 family ramping in 2H26, AMD’s first move from single-die delivery to rack-scale clusters. AMD has lined up collaborations with OpenAI and Meta, and MI450’s ramp will directly inform deliveries.
ii) Server CPU: stronger server CPU demand made this the largest incremental driver in Q1. Dolphin Research estimates server CPU revenue at around $3.3bn, up 15% QoQ. Third-party data show AMD’s server CPU unit share has exceeded 20%.
4) Guidance: Q2 2026 revenue of $10.9–11.5bn (above the Street at ~$10.5bn), midpoint $11.2bn implies +9% QoQ. Non-GAAP GPM guided to ~56% (vs. Street 55.3%).
Dolphin Research view: CPU upcycle in place; awaiting MI450 for dual-engine growth
Q1 topline beat was solid, primarily driven by server CPUs. MI355 AI GPU was softer QoQ on China-related adjustments, while investors are more focused on MI450 later this year. AMD has signed collaboration frameworks with OpenAI and Meta for MI450, which transitions delivery from single-die to rack-scale clusters.
Beyond Q1, guidance was strong. Q2 revenue is guided to $10.9–11.5bn (+6–12% QoQ), ahead of the ~$10.5bn consensus, and non-GAAP GPM around 56% (vs. 55.3% consensus). The upside next quarter is mainly server CPU-driven, and Dolphin Research expects server CPU and related revenue to reach about $3.8bn, up 15% QoQ.
Outside the quarter, the Street is focused on several items:
a) Hyperscaler capex: cloud majors remain the main chip buyers in this AI boom.
All four reported recently, with Meta and Microsoft explicitly raising capex. For 2026, Dolphin Research estimates combined capex for Meta, Google, Microsoft and Amazon could exceed $700bn, up 70%+ YoY, with a back-half weighted profile as NVIDIA’s Rubin and AMD’s MI450 ramp in 2H.
b) CPUs: moving to center stage in AI
With PCs up 3.8% YoY in Q1, Intel’s Client revenue grew just 1.3% YoY, while AMD grew 26%, showcasing rising CPU competitiveness.
Across total CPU (PC + server), AMD’s share has kept climbing in recent years and has overtaken Intel in desktop.
Training prioritized raw compute, leaving CPUs less central, while in inference, latency matters more. CPUs handle orchestration and data pre-processing, directly affecting throughput, latency and efficiency, lifting AI server CPU demand and pricing.
Meanwhile, AMD continues to chip away at Intel’s share. Third-party data indicate AMD’s server CPU unit share has surpassed 20%.
c) AI GPU: MI355 is the workhorse today; the focus is on MI455 in 2H.
The MI355-to-MI455 upgrade centers on compute, memory and clustering. Compute is roughly doubling (moving to 2nm), and memory upgrades (HBM3E to HBM4) boost bandwidth and capacity.
Beyond compute and memory, the key watch is AMD’s rack-scale solution. As workloads shift from training to inference, raw compute matters less at the margin. For example, MI355 sits in the same performance tier as NVIDIA’s B200, but AMD’s competitive gap lies in the lack of rack-scale cluster delivery.
For MI450, AMD plans to tightly pair it with the Helios rack platform, shifting delivery from single-die to rack-scale clusters to directly counter NVIDIA’s Rubin.
For large CSPs, the need is cluster-scale deployments rather than buying standalone boards. Rack-scale delivery for MI450 fits that need, and with volume only starting in 2H, AI GPU revenue in 2026 should be back-half weighted.
AMD has signed multi-GW agreements with OpenAI, Meta and Oracle, and the MI450 ramp will directly determine deliveries to these clients.
With a current market cap of $579.2bn, AMD trades at about 33x 2027E post-tax core earnings (assuming a 44% two-year revenue CAGR, 56.7% GPM and 13% tax rate). Versus AI peers, AMD’s PE is above NVIDIA and TSMC, reflecting the CPU upcycle and expectations for a larger AI share.
Near term, results hinge on CPUs and MI355 shipments, with tight server CPU supply fueling the CPU upcycle. Medium to long term, as server CPU demand surges, the market will watch MI450 for new customers, orders, and rack-scale volume delivery.
In this CPU recovery cycle, AMD’s core earnings base should rise with high visibility. Management lifted the total addressable CPU market to $120bn (from $60bn) and now assumes a 1:1 CPU/GPU ratio, implying CPUs will not be secondary to GPUs in AI, materially raising server CPU expectations.
On AI GPUs, AMD still trails NVIDIA on raw compute, but that does not preclude inference wins, with OpenAI, Meta and Oracle already on board.
The key watch into 2H is MI450’s rack-scale mass production and deliveries. Successful fulfillment for existing large customers could unlock more external demand, allowing AMD to benefit from a dual-engine CPU+GPU AI cycle.
Below is Dolphin Research’s detailed take on AMD’s results:
I. Overall results: topline beat, guidance stronger
1.1 Revenue
AMD delivered Q1 2026 revenue of $10.25bn, up 38% YoY and ahead of ~ $9.9bn consensus. Growth was led by Data Center and Client.
While AI GPU dipped QoQ, the server CPU upcycle lifted Data Center further. Dolphin Research estimates server CPU-related revenue at ~$3.3bn, up 15% QoQ.
1.2 Gross profit
Gross profit was $5.42bn (+45% YoY), with GPM at 52.8% (+260bps YoY), driven by higher Data Center mix.
The company guided Q2 non-GAAP GPM at ~56%, up ~60bps QoQ, mainly on server CPU growth. MI450’s 2H mass production could dilute margins, but long-term GPM is still guided at 55–58%.
1.3 Opex
Total opex was $3.65bn, up 40% YoY, with noticeable increases in both R&D and SG&A.
Breakdown: i) R&D was $2.4bn (+39% YoY), with continued investment focused on AI. ii) SG&A was $1.25bn (+41% YoY), tracking revenue growth and reflecting higher employee incentives.
1.4 Profit
The Xilinx acquisition creates sizable amortization headwinds, pressuring reported profit for some time. For underlying performance, ‘core OP’ is more indicative.
Core OP = GP − R&D − SG&A. Ex-acquisition effects, Dolphin Research estimates Q1 core OP at $1.77bn (+57% YoY), mainly on the server CPU recovery.
II. Segment detail: server CPU super-cycle; watch MI450 ramp
Data Center and Client are the main revenue drivers, together above 80%. Benefiting from MI-series AI GPU shipments and stronger server CPU demand, Data Center’s mix keeps rising.
2.1 Data Center
Q1 Data Center revenue was $5.78bn (+57% YoY), beating ~ $5.6bn consensus, driven mainly by the server CPU upcycle.
Based on company and market views: AI GPU revenue was about $2.4bn (slightly down QoQ), while server CPU and related revenue was about $3.3bn (+15% QoQ). With MI450 ramping in 2H, AI GPU revenue should be back-half weighted.
In detail:
a) Data Center CPU: with combined ‘CPU + GPU’ positioning, AMD’s Data Center CPU share has surpassed 20%. While Intel’s x86 CPUs are collaborating with NVIDIA and Google, that largely provides customers an x86 option.
Amid the CPU upcycle and ongoing Zen iterations, DC CPU revenue should see sustained growth. Management guided server CPU revenue to grow ~70% YoY in Q2 and expects momentum to continue into 2H, underpinning strong full-year growth.
b) AI GPU: revenue currently comes mainly from the MI350 ramped in 2H25. Because MI350 is still delivered at the chip level, not aligned with mainstream rack-scale demand, AI GPU performance is relatively weaker.
Given capex hikes at players like Meta, Dolphin Research expects combined 2026 capex for Meta, Google, Microsoft and Amazon to exceed $700bn (+70% YoY), supporting high AI chip growth in 2026.
In this fast-growing market, AMD’s order wins hinge on product capability. Beyond the server CPU tailwind, the focus in AI GPUs is MI450’s 2H mass-production progress.
For MI450, AMD plans deep integration with the Helios rack platform to enable rack-scale delivery, directly countering NVIDIA’s Rubin and aligning with CSPs’ mainstream needs.
AMD has secured OpenAI, Meta and Oracle, and MI450’s ramp will directly affect deliveries to these customers.
2.2 Client
Q1 Client revenue was $2.89bn (+26% YoY), above the ~$2.7bn Street, driven by share gains vs. Intel in PCs.
Industry data show global PC shipments of 65.6mn units in Q1 2026 (+4% YoY), while AMD’s Client revenue rose 26% YoY. By contrast, Intel’s Client revenue grew only 1.3% YoY.
Even with a sluggish PC market, Dolphin Research expects AMD to outgrow the market on product competitiveness.
2.3 Others
1) Gaming: Q1 2026 revenue was $720mn (+11% YoY).
Details: i) Gaming GPU: Radeon RX9000 grew YoY. ii) Semi-custom: expected to decline YoY per console cycle dynamics, as 2026 is year 7 of the cycle and a transition period. With higher costs for memory and other components, Gaming could see a notable downturn in 2H.
2) Embedded: Q1 2026 revenue was $870mn (+6% YoY), driven by test & measurement and emulation, aerospace & defense, communications, and x86 embedded products.
Customized solutions continue to expand in DC and communications, extending beyond FPGA-centric offerings into adaptive embedded x86 and semi-custom, significantly enlarging the overall TAM.
<End>
Dolphin Research on AMD
Feb 4, 2026 call notes ‘AMD (Trans): Long-term view intact; MI500 to launch in 2027’
Feb 4, 2026 earnings take ‘Big words, small beats — when will AMD deliver?’
Nov 5, 2025 call notes ‘AMD (Trans): No new orders disclosed; MI450 focuses on rack-scale’
Nov 5, 2025 earnings take ‘AMD: Partnering with OpenAI? Not a mere backup’
Aug 6, 2025 call notes ‘AMD (Trans): Even excluding MI308, next-quarter AI GPU to grow YoY’
Aug 6, 2025 earnings take ‘AMD: Eating into Intel on CPUs; when can AI GPUs challenge NVIDIA?’
May 7, 2025 call notes ‘AMD (Trans): AI GPU to deliver double-digit growth for the year’
May 7, 2025 earnings take ‘AMD: Has Intel on the ropes — turning up the heat on NVIDIA?’
Feb 5, 2025 call notes ‘AMD (Trans): No sequential growth in DC in 1H25’
Feb 5, 2025 earnings take ‘AMD: DeepSeek stirs the pot; GPU backup thesis hit by ASICs’
Risk disclosure & disclaimer: Dolphin Research disclaimer and general disclosure
The copyright of this article belongs to the original author/organization.
The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.
