⚡ Earnings Preview AMD · NASDAQ Semiconductors / AI Infrastructure Published Apr 28, 2026 · Q1 2026 Print: May 5

Advanced Micro Devices
Q1 2026 Earnings Preview

AMD enters Q1 2026 earnings as the only credible challenger to NVIDIA's AI GPU monopoly — with MI300X/MI325X in hyperscaler deployments at scale and EPYC eroding Intel's server CPU stronghold. The question isn't whether AMD is winning. It's whether the Data Center GPU ramp is accelerating fast enough to justify a ~25x forward multiple.

Next Earnings
May 5, 2026
Current Price
~$105
Market Cap
~$170B
Data Center Rev (Q4 '25)
~$3.9B (+69% YoY)
🏛️
01 / 07
Company Snapshot

Advanced Micro Devices (NASDAQ: AMD) designs high-performance processors, graphics chips, and adaptive computing solutions. Founded in 1969 by Jerry Sanders as a second-source manufacturer of Intel logic chips, AMD spent its first four decades as a scrappy underdog in x86 CPUs and discrete graphics — perpetually behind Intel in CPUs and NVIDIA in GPUs but always alive enough to force competition. That dynamic inverted in 2017 when CEO Lisa Su — who took over in 2014 when AMD was losing $1 billion annually — launched the Ryzen CPU and EPYC server CPU lines on the Zen architecture. EPYC has grown AMD's server CPU market share from under 1% to roughly 25% by 2025, and the MI300X accelerator has established AMD as the only GPU company that hyperscalers deploy at scale alongside NVIDIA.

AMD operates four reportable segments: Data Center (EPYC CPUs + MI-series Instinct GPUs), Client (Ryzen desktop/laptop CPUs and Radeon iGPUs), Gaming (Radeon discrete GPUs + semi-custom chips for PlayStation 5 and Xbox Series X|S), and Embedded (Xilinx FPGA and adaptive computing, acquired 2022). The 2022 Xilinx acquisition ($49B) added FPGAs and adaptive SoCs critical for aerospace, defense, automotive, and 5G markets. Data Center now drives the growth narrative — the three other segments are either in secular decline (Gaming semi-custom cycle end) or recovery (Embedded inventory correction).

Founded
1969
HQ
Santa Clara, CA
Employees
~28,000
Exchange
NASDAQ: AMD
Business Model
Fabless semiconductor
FY2025 Revenue
~$33B (+28% YoY est)
Data Center (~47%) Client (~28%) Gaming (~8%) Embedded (~17%) MI300X / MI325X EPYC Turin Ryzen AI ROCm Xilinx FPGA
♟️
02 / 07
Recent Strategic Moves

AMD's strategic agenda over the last 18 months has been sharply focused on two objectives: winning hyperscaler AI GPU wallet share and integrating ZT Systems to sell full AI rack solutions — not just chips. The moves are designed to challenge NVIDIA's full-stack dominance, not just its silicon.

MI300X / MI325X ramp: hyperscaler deployments at scale
2024–2026
The MI300X (192GB HBM3, CDNA 3 architecture) launched December 2023 and became the first GPU competitor that major hyperscalers deployed at meaningful scale. Microsoft Azure, Meta AI, and Oracle Cloud all publicly confirmed MI300X deployments — Microsoft uses it for a significant portion of Azure AI inference. MI325X (256GB HBM3e) followed in Q4 2024 with improved memory bandwidth. AMD guided $5B+ in MI300-series revenue for 2024, and actual results significantly exceeded that figure. The 2025 MI325X ramp continued into major hyperscaler contracts with OpenAI and Google Cloud confirmed in operator briefings.
ZT Systems acquisition closed (~$4.9B) — AI rack solution play
Q1 2025 (close)
AMD closed the ZT Systems acquisition for ~$4.9B in early 2025. ZT is a hyperscale server solutions provider that designs and delivers complete AI rack infrastructure for data center customers. AMD retained ZT's system design capabilities (the strategic asset) and sold ZT's manufacturing arm to Sanmina Corporation. The strategic logic: AMD now sells complete AI clusters (compute + system design + software support), directly competing with NVIDIA's DGX and HGX full-rack solutions. This shifts AMD from a chip vendor to an AI infrastructure solutions provider.
EPYC Turin (Gen 5 Zen 5) launched — server CPU market share gains accelerate
September 2024
EPYC Turin launched in September 2024 with up to 192 cores (Zen 5 architecture), delivering 2.5x the performance of the outgoing Genoa in key HPC and AI-adjacent workloads. Turin directly addresses Intel's remaining server CPU strongholds: Oracle, AWS, and Google all announced Turin deployments. AMD's server CPU market share reached approximately 25% in Q4 2025, up from under 1% in 2018. Turin's performance-per-watt advantage over Intel Granite Rapids is documented across independent benchmarks. Intel's next credible response isn't until Clearwater Forest (Xeon 6 successor), targeted for late 2026.
ROCm 6.x advances — closing (not eliminating) the CUDA software gap
2024–2025
ROCm, AMD's open-source GPU software stack, released versions 6.1 and 6.2 through 2024-2025 with significant improvements to performance, framework compatibility, and model support. PyTorch and JAX both officially support ROCm. Meta AI runs large-scale inference on ROCm-backed MI300X in production. The remaining gap: CUDA-native code typically runs 10–25% faster in equivalent compute scenarios, and AMD's ecosystem of pre-optimized ML kernels is still smaller than NVIDIA's. The gap is narrowing — but it's not closed, and the asymmetry persists for fine-grained CUDA-dependent applications.
⚔️
03 / 07
Competitive Position

AMD competes in two distinct battlefields: datacenter AI GPUs (vs. NVIDIA), and server CPUs (vs. Intel). It is winning on both fronts — but from asymmetric positions. In GPUs, AMD is a credible challenger to a monopolist. In server CPUs, AMD is an ascendant market-share taker from a declining incumbent. The strategic risk: winning 20% of GPU share is worth far more than winning 50% of server CPU share, because GPU ASPs are 5–10x higher.

NVIDIA (GPUs) 🔴
Dominant incumbent. NVIDIA holds ~80% AI accelerator market share with the CUDA software ecosystem as its principal moat — 20+ years of developer investment that AMD cannot replicate on a relevant timeline. Blackwell GB200/GB300 NVL72 racks deliver ~1.1 exaFLOP/rack; AMD's comparable MI325X cluster is within striking distance on raw compute but still lags on interconnect efficiency (NVLink vs. Infinity Fabric) and software tooling. The key metric: AMD's MI300X wins on memory bandwidth (HBM3 quantity per chip is higher) and total memory capacity — which makes it superior for large-model inference and long-context inference workloads. NVIDIA wins on overall training throughput for CUDA-native models. AMD's data center GPU revenue is approximately 4–5% of NVIDIA's run rate — a real number, but a vastly smaller scale.
Intel (Server CPUs) 🟡
Weakening incumbent. Intel's Xeon Granite Rapids (Gen 6) is competitive in some workloads but generally trails EPYC Turin on performance-per-watt and performance-per-dollar. Intel is also distracted by its foundry strategy transition and ongoing management instability (Pat Gelsinger departed Dec 2024; interim leadership under David Zinsner). AMD's 25% server CPU market share is a structural gain — not a cyclical one — and the trajectory points toward 30%+ by 2027. Intel's Clearwater Forest (Xeon 6 successor) won't arrive until late 2026 at the earliest. In the meantime, AMD is winning data center refresh contracts at Oracle, AWS, and Google.
Hyperscaler Custom ASICs 🟡
Structural headwind, not near-term threat. Google TPU v7 Ironwood (4,614 TFLOPS per chip), AWS Trainium3, Microsoft Maia, and Meta MTIA are all growing at 44%+ annually. These ASICs offer superior TCO for specific, well-understood workloads and reduce AMD and NVIDIA dependency. The critical nuance: all hyperscalers also continue buying MI300X/MI325X and NVIDIA GPUs — custom silicon supplements rather than replaces merchant GPUs for the broad, diverse range of AI workloads. AMD is less exposed than NVIDIA to this risk because AMD is not yet the dominant provider; losing 5% of hyperscaler GPU spend to ASICs matters more to NVIDIA ($12B) than to AMD ($600M).
Qualcomm / ARM (Client CPUs) 🟢
Emerging long-term risk. ARM-based laptop chips (Apple M-series, Qualcomm Snapdragon X) are taking share in premium notebooks and demonstrating that x86's performance-per-watt isn't unbeatable. AMD's Ryzen AI 300 series with integrated NPU is the answer — it ships in mid-range through premium Windows AI PCs. Near-term client CPU risk is manageable; the structural shift to ARM notebooks is a 3–5 year threat to AMD's PC TAM, not a 2026 problem.
Bottom line on competitive position: AMD has achieved what no competitor has in 15 years — it's putting real pressure on both Intel (server CPUs) and NVIDIA (AI GPUs) simultaneously. It isn't winning either battle outright. But it doesn't need to. Being the preferred alternative at 15–20% GPU share and 25–30% server CPU share is a multi-billion dollar, high-growth franchise. The risk is that NVIDIA's Vera Rubin (H2 2026) re-opens the hardware gap just as AMD's MI350 ramps — and AMD has to execute flawlessly on software to hold its hyperscaler wins.
📊
04 / 07
Financial Trajectory

AMD's revenue trajectory over the last four quarters reflects a company in the middle of a major product cycle inflection: Data Center segment revenue has more than tripled in two years on the back of MI300X deployments, while Gaming continues to decline as the semi-custom console cycle matures and Embedded recovers from a prolonged inventory correction. The gross margin story is the other key thread — AMD is pushing toward 54%+ non-GAAP gross margins as Data Center (higher ASP, higher margin) becomes a larger share of the mix.

Q1 2025 Revenue
$7.44B (+36% YoY)
Q2 2025 Revenue
$7.69B (+9% YoY)
Q3 2025 Revenue
$8.41B (+18% YoY)
Q4 2025 Revenue
$7.66B (+25% YoY)
Data Center Q4 2025
~$3.9B (+69% YoY)
Non-GAAP Gross Margin
~53.5% (Q4 2025)
Client Q4 2025
~$2.3B (+29% YoY)
Gaming Q4 2025
~$0.6B (-45% YoY)

The segment mix shift is the defining financial story: Data Center was ~29% of revenue in FY2023; it's ~47% in FY2025. That mix expansion drives gross margin toward the mid-50s — Data Center GPU and EPYC CPU carry structurally higher margins than Client CPUs or Gaming. The key gross margin watchpoint for Q1 2026: MI350 qualification costs and any Data Center pricing pressure as NVIDIA Blackwell supply improves. Gaming will continue declining through mid-2026 as PS5/Xbox semi-custom winds down ahead of next-gen console design wins. Embedded is recovering — FPGA inventory correction peaked in mid-2024 and AMD has guided sequential improvement through 2025–2026.

⚠️
05 / 07
Key Risks
📬 Post-Print Update — AMD prints May 5 AMC

Get the post-print update. We'll email you within 24h of the print with actuals, thesis verdict, and an updated bull/bear range.

AMD's risk profile is meaningfully different from NVIDIA's: NVIDIA has a near-monopoly facing structural disruption risk; AMD has a challenger growing into a dominant market from below, facing execution risk. For AMD, most risks are about whether it can execute fast enough to hold and expand market share — not whether the market exists.

High
NVIDIA CUDA Moat — Software Switching Costs
The single largest structural risk for AMD's AI GPU ambitions isn't hardware — AMD's MI350 (CDNA 4) matches or beats Blackwell B200 on several raw compute metrics. The risk is CUDA. Twenty years of CUDA-optimized model code, kernels, and tooling represents switching costs that no hardware win eliminates. ROCm 6.x has improved PyTorch/JAX compatibility, but real-world training on AMD GPUs still requires manual optimization that NVIDIA users get for free. If ROCm fails to close the software gap by the time NVIDIA's Vera Rubin ships (H2 2026), AMD's hardware competitiveness could be neutralized — and hyperscalers that tried MI300X may reduce allocation in favor of proven CUDA-native Blackwell or Vera Rubin infrastructure.
High
China Export Controls on MI-Series GPUs
U.S. export control rules restrict export of advanced AI accelerators to China, including AMD's MI300-series. AMD has historically generated ~15–20% of revenue from China (including client, gaming, and data center). The BIS October 2023 and October 2024 rules effectively prohibit export of MI300X/MI325X to Chinese customers without a license (rarely granted). AMD has less China data center GPU revenue to lose than NVIDIA (since NVIDIA was dominant there), but the restriction removes a multi-billion dollar TAM expansion opportunity and pressures AMD's ability to scale production economics faster. Any further tightening — including potential restrictions on less-capable chips — adds downside risk.
Med
Client PC Cyclicality
AMD's Client segment (~28% of revenue) sells Ryzen CPUs for desktops and laptops. PC market unit volumes have declined from the 2021 pandemic peak and are recovering slowly. ARM-based alternatives (Apple M-series, Qualcomm Snapdragon X Elite) are capturing premium notebook share. AMD's AI PC push — Ryzen AI 300 with integrated NPU — addresses some of this threat in the Windows ecosystem, but the structural TAM for x86 client CPUs is unlikely to return to prior peaks. A broader macro slowdown that defers PC refresh cycles would hit AMD's second-largest segment.
Med
MI350 Execution Risk — Timing and Yield
MI350 (CDNA 4 architecture, targeting H2 2026) is AMD's most important product cycle in years — it's the chip that must maintain competitive parity with NVIDIA's Vera Rubin. Any timing slip (to H1 2027) or yield issues with TSMC's 3nm node would cede GPU market share to NVIDIA at the worst moment. AMD has historically executed well on product roadmaps under Lisa Su, but CDNA 4 is AMD's first major GPU architecture refresh in two years, and the competitive stakes are higher than ever. A 6-month slip could cost AMD 2+ quarters of hyperscaler GPU revenue as customers default to Vera Rubin.
Low
Embedded Segment Recovery Pace
The Embedded segment (Xilinx FPGAs and adaptive SoCs) is still recovering from a severe inventory correction that began in mid-2023. Industrial and communications customers over-ordered during the supply shortage years, then drew down inventory through 2024. AMD guided sequential recovery through 2025, and Embedded revenues are recovering — but a slower macro recovery in industrial, automotive, or telecom could extend the correction. This is a lower-stakes risk: Embedded represents ~17% of revenue and the correction is acknowledged and largely priced in.
🚀
06 / 07
Catalysts Ahead

AMD's catalyst calendar through 2026 is defined by two product launches that bookend the year: MI350 (CDNA 4) and the potential initial reveal of MI400 roadmap positioning. Both are existential for AMD's GPU market share trajectory. Secondary catalysts — EPYC continued share gains, ZT Systems rack solution ramp, and Embedded recovery — provide support.

MI350 (CDNA 4) launch — H2 2026 target
Expected Q3–Q4 2026
MI350 is AMD's CDNA 4 architecture GPU, built on TSMC's 3nm node. AMD has claimed 35x AI performance improvement over MI300X at peak theoretical compute (FP4/INT8 inference). More importantly, MI350 will feature HBM3e with higher bandwidth-per-chip than the current MI325X, addressing one of the primary customer objections to AMD vs. NVIDIA Blackwell for training workloads. If MI350 ships on schedule (H2 2026) and achieves hyperscaler qualification by Q4, AMD enters 2027 with a product that can credibly compete with NVIDIA's Vera Rubin for new cluster deployments. This is the single most important catalyst for AMD's 2027 AI GPU revenue trajectory.
FY2026 Data Center GPU revenue guidance — the annual print
May 5, 2026 (Q1 earnings call)
The Q1 earnings call on May 5 will almost certainly include AMD's updated view on full-year 2026 data center GPU revenue. Street consensus is currently modeling $13–15B in Data Center segment revenue for FY2026, with MI-series GPUs representing ~$9–10B of that. If AMD guides Data Center GPU revenue materially above these estimates — or if Lisa Su frames the MI325X demand as supply-constrained — the stock re-rates immediately. Alternatively, guidance below $12B would represent a meaningful deceleration from 2025 momentum and likely pressure the stock significantly.
ZT Systems AI rack solution ramp — full-stack positioning
2026
AMD closed ZT Systems in Q1 2025, retained system design capabilities, and sold manufacturing to Sanmina. The strategic bet: AMD AI clusters — MI325X/MI350 GPUs + EPYC Turin CPUs + ZT-designed rack architecture — compete with NVIDIA's DGX H200 and GB300 NVL72 as a complete AI infrastructure solution. The first full ZT-AMD clusters began qualifying at hyperscalers in late 2025. A large-scale ZT-AMD cluster win at a Tier 1 hyperscaler or CSP would be a signal that AMD has moved beyond chip-selling into AI infrastructure platform status — a significant stock catalyst and ASP/margin expansion lever.
EPYC continued server CPU market share capture
Ongoing
Intel's server CPU position continues to weaken — Granite Rapids faces EPYC Turin on performance-per-watt, and Intel's foundry transition has hampered roadmap execution. AMD's EPYC is now the default choice for new hyperscale CPU deployments at Oracle, Google, and increasingly AWS. Server CPU ASPs are ~$2–4K vs. GPU ASPs of $15–40K+, so CPU share gains contribute meaningfully but won't move the needle the way GPU wins do. Still, a sustained EPYC share gain to 28–30% by 2027 adds $2–3B annually to AMD's revenue with high margins.
📋
07 / 07
Pre-Earnings Preview — May 5, 2026
⏳  PRE-PRINT — Consensus estimates only. Actuals and commentary to be added after the May 5 earnings call close.

AMD reports Q1 2026 (quarter ended March 28, 2026) on May 5 after market close. Here's what Wall Street is modeling and what to watch on the call:

Metric Consensus Est. Estimate Range Actual (Post-Print)
Total Revenue ~$7.25B $7.1B – $7.45B TBD — post-print
Non-GAAP EPS ~$1.09 $1.03 – $1.18 TBD — post-print
Data Center Segment ~$3.3B $3.0B – $3.6B TBD — post-print
Client Segment ~$2.1B $1.95B – $2.25B TBD — post-print
Gaming Segment ~$0.5B $0.45B – $0.6B TBD — post-print
Non-GAAP Gross Margin ~54.0% 53.0% – 55.0% TBD — post-print
Q2 2026 Revenue Guide ~$7.4B $7.0B – $7.8B TBD — post-print
// Key watch items on the May 5 call
  • MI300X / MI325X quarterly GPU revenue commentary: AMD doesn't break out GPU revenue separately. Watch for Lisa Su to give a directional number or growth rate — the street is looking for implied MI-series GPU revenue in the $3B+ range for Q1 and a trajectory toward $9–10B+ for the full year. Any framing around supply constraints (demand exceeding supply) would be the most bullish signal.
  • MI350 timing confirmation: Any pull-forward from H2 to H1 2026 would be a major catalyst. Conversely, a slip signal to 2027 is a serious negative. Lisa Su's prior guidance was "available in the second half of 2026" — listen for whether that tightens to a specific quarter.
  • FY2026 Data Center GPU TAM update: AMD previously described the 2025 AI accelerator TAM as growing substantially. On this call, expect updated full-year guidance or at minimum a qualitative framing. If AMD implies FY2026 Data Center segment revenue above $14B, the stock re-rates materially. Below $12B would be a disappointment.
  • Gross margin trajectory: AMD has been guiding non-GAAP gross margins of ~54%. Any guidance for sequential compression (toward 52–53%) would signal pricing pressure from NVIDIA competition or higher MI350 NRE costs. Expansion toward 55%+ would confirm the Data Center mix shift is driving margin improvement as expected.
  • ZT Systems integration update: First full-rack AI cluster deployments in production. Lisa Su has not quantified ZT's revenue contribution separately — watch for whether she begins to characterize the revenue opportunity or any named hyperscaler wins.
  • Embedded recovery pace: AMD guided Embedded would return to seasonally normal revenue levels by Q2/Q3 2026. Q1 sequential performance vs. Q4 2025 will signal whether that recovery is on track.
The single most important number on this call

AMD doesn't separately report MI300X/MI325X GPU revenue. The most important signal is the combination of total Data Center segment guidance for Q2 and any qualitative commentary on MI-series demand versus supply. If Lisa Su says demand exceeds supply — and if she signals $4B+ Data Center for Q2 — AMD stock will likely respond positively. If she guides Q2 Data Center below Q1 (sequential decline), expect significant selling regardless of total revenue beat.

📬 Post-Print Update — AMD prints May 5 AMC

Get the post-print update. We'll email you within 24h of the print with actuals, thesis verdict, and an updated bull/bear range.

Data current as of April 28, 2026. Financial figures sourced from AMD Q4 2025 / FY2025 earnings release (January 2026), AMD Q3 2025 earnings release (October 2025), and SEC filings. Segment data sourced from AMD investor relations materials. Consensus estimates aggregated from Wall Street research (Wells Fargo, BofA Securities, Morgan Stanley, Wolfe Research). MI-series revenue estimates are analyst estimates — AMD does not publicly disclose GPU-specific revenue. Section 7 actuals to be populated following the May 5, 2026 earnings call. Forward-looking statements are Vektor analysis, not investment advice.
// You might also be interested in
⚡ Vektor Intelligence

Get a brief like this
for any ticker — $29/mo

Drop in any company name. 40 seconds. Full 7-section intelligence brief: snapshot, strategy, competitive position, financials, risks, catalysts, and earnings preview.

Start for $29/month → — or get your first brief free —
✓ You're in — we'll send your free brief shortly.

No credit card. Cancel anytime.