AMD’s 2027 Server CPU Outlook: Supply Constraints and ASP Growth Claims Clash

Wednesday, Aug 5, 2026 2:26 am ET5min read
AMD--
Aime RobotAime Summary

- AMDAMD-- reported Q2 2026 revenue of $11.5B (50% YoY), driven by AI demand in data centers and strong server CPU sales.

- Data center revenue surged 107% YoY to $6.7B, fueled by EPIC processors and Instinct accelerators.

- Server CPU sales grew >70% YoY, with Helios AI platform and strategic partnerships (e.g., Microsoft) expected to drive 2027 growth.

- 2027 guidance projects >35% revenue growth, with AI-driven segments (data center, server CPU) exceeding $30B combined and EPS surpassing $20 target.

- Management anticipates navigating margin pressures through business mix shifts and supply chain optimizations, maintaining "enormous" growth opportunities.

Date of Call: Aug 4, 2026

Financials Results

  • Revenue: $11.5 billion, up 50% year-over-year and 13% sequentially
  • EPS: $0.33 per diluted share, up approximately 82% year-over-year
  • Gross Margin: 56%, up over 200 basis points year-over-year and 80 basis points sequentially
  • Operating Margin: 27%

Guidance:

  • Q3 2026 revenue expected to be approximately $13 billion, plus or minus $300 million, representing year-over-year growth of approximately 41% and sequential growth of approximately 13%.
  • Q3 2026 non-GAAP growth margin expected to be approximately 56%.
  • Data center segment expected to grow more than 80% year-over-year in the second half of 2026 and more than 70% for the full year 2027.
  • Data center segment revenue expected to more than double year-over-year in 2027.
  • Server CPU business expected to grow more than 70% year-over-year in 2027.
  • Overall company revenue expected to grow substantially above the prior target of greater than 35%, with EPS expected to significantly exceed the $20 annual target within the strategic timeframe.

Business Commentary:

Revenue Growth and AI Integration:

  • AMD reported record revenue of $11.5 billion for Q2 2026, up 50% year-over-year.
  • This growth was driven by strong demand for AI products, especially in the data center segment.

Data Center Revenue Surge:

  • The company's data center revenue reached a record $6.7 billion, more than doubling year-over-year with a 107% increase.
  • This was attributed to strong demand for EPIC processors and Instinct accelerators, reflecting the expanding scale of server and data center AI businesses.

Server CPU and AI Accelerator Demand:

  • AMD's server CPU sales grew more than 70% year-over-year, with record sales in both cloud and enterprise segments.
  • The demand was driven by the leadership performance and TCO advantages of EPIC processors and the expanding adoption of AI workloads.

Helios Platform and Strategic Partnerships:

  • The launch of the Helios rack-scale AI platform, combining EPIC Venice CPUs and MI450 series GPUs, is expected to drive significant growth.
  • Strategic partnerships with companies like Anthropic and Microsoft are anticipated to increase demand for AMD's AI infrastructure solutions.

Client and Gaming Segment Performance:

  • The client business revenue increased 23% year-over-year to $3.1 billion, driven by record mobile processor revenue.
  • The gaming segment revenue declined 31% year-over-year due to lower semi-custom sales, reflecting the stage of the console cycle.

Sentiment Analysis:

Overall Tone: Positive

  • Management described the quarter as "outstanding", with "record revenue and profitability" and "strong momentum across our businesses." They noted they are "tracking materially ahead of the long-term financial model" and are "exceptionally well positioned" for significant growth, stating "the opportunity ahead is enormous."

Q&A:

  • Question from Tom O'Malley (Barclays): When do you see the crossover point happening in the data center GPU business and the CPU business, and where are you seeing the most strength into calendar year 27?
    Response: Both server CPU and data center AI businesses are expected to grow very strongly in 2027, with both being significant drivers. The data center AI TAM is larger, so substantial growth is expected from that business in 2027, but server growth will also be very significant.

  • Question from Tom O'Malley (Barclays): What are the tradeoffs to gross margins into calendar year 27 given headwinds on the GPU side and tailwinds on the server processor side?
    Response: Growth margin is driven by business mix. The server business expansion is accretive, and while data center AI has a slightly lower gross margin, its large revenue opportunity is beneficial. The embedded business is also expected to provide a tailwind, and overall, management feels good about navigating the transition and improving growth margin.

  • Question from Timothy Arcuri (UBS): Can you give color on what's embedded in the guidance in terms of data center, specifically which will grow more on a dollar basis between CPU and GPU in Q3 and Q4?
    Response: Server CPU is expected to grow substantially in Q3 and Q4, with strong double-digit sequential growth in Q3. Both server and data center AI will grow nicely in Q3, with growth expected to be higher in Q4 as the Helios ramp begins and supply increases.

  • Question from Timothy Arcuri (UBS): Do you think you can outgrow the 40% CAGR outlined at the analyst day through 2030?
    Response: Yes, with AI tailwinds benefiting all business segments and the company's ability to grow above the market, AMD expects to grow its revenue greater than 40% annually through 2030.

  • Question from Vivek Arya (Bank of America Securities): How many gigawatts of compute does AMD have line of sight into for 2027, and what is the monetization per gigawatt?
    Response: AMD has supply to more than meet guidance, with a very significant ramp expected over the next couple of quarters. Revenue per gigawatt is in the double-digit billions range, and demand for compute is high.

  • Question from Vivek Arya (Bank of America Securities): How much of Helios' benefit is due to higher HPM, and does that expose AMD to memory cost inflation?
    Response: AMD works closely with memory partners and has good visibility into HBM allocation for 2027. The memory footprint can be modified if total cost of ownership optimization is needed for certain workloads, so the benefit is workload-dependent.

  • Question from Joshua Buckalter (TD Cowan): Can you give help on the unit and ASP assumptions for the server CPU 2027 guidance?
    Response: Growth is driven by both unit and ASP growth, with unit growth being higher. Capacity is being raised across the supply chain to support the significant growth, allowing for the raised second half guidance and expected over 70% year-over-year growth in 2027.

  • Question from Joshua Buckalter (TD Cowan): Was the Helios forecast comment on volumes, and should we expect gross margins to improve as it ramps?
    Response: The comment was on overall volumes, indicating strong demand. Yields are expected to improve as the ramp progresses over the next few quarters.

  • Question from Aaron Rakers (Wells Fargo): How big is the sandbox AI market opportunity relative to the total $220 billion TAM, and how does that impact growth?
    Response: The agentic AI or sandbox AI segment is the largest and fastest-growing piece of the TAM, though it is the smallest today. AMD's portfolio is strong across all segments, and Venice is leadership in all categories, providing confidence to grow ahead of the market.

  • Question from Aaron Rakers (Wells Fargo): Do you still see growth in the client CPU business for 2027?
    Response: Yes, the client CPU business remains important, especially with the rise of AI PCs. AMD has a strong product portfolio and expects to grow ahead of the market despite potential component cost pressures.

  • Question from Stacy Rasgin (Bernstein Research): How should I interpret the statement that data center accelerates in the second half, given the Q3 guide implies a deceleration from Q2's 107% growth?
    Response: The acceleration refers to the second half of 2026 versus the first half, not necessarily a sequential acceleration from Q2.

  • Question from Stacy Rasgin (Bernstein Research): How should I interpret the 'more' in the data center targets for 2027?
    Response: The 'more' indicates strong growth, with server CPU expected to be well over 70% and overall data center AI well over 100% due to the ramp of strategic customers and Helios.

  • Question from Jim Schneider (Goldman Sachs): What customer diversity do you expect in the early stages of the MI400 series ramp, and how many customers will be contributing in Q4 and H1 2027?
    Response: Diversity is expected with large frontier model companies and hyperscalers, plus additional customers at a more regular scale. Visibility into data center plans ensures a good diversity as the ramp progresses.

  • Question from Jim Schneider (Goldman Sachs): Any constraints from customers' data center readiness that could impact hitting 2027 targets?
    Response: No significant constraints are seen. The supply chain is being built comprehensively, and there is strong collaboration with data center operators to accelerate capacity, supporting confidence in the targets.

  • Question from CJ Muse (Cantor): How do you see the revenue cadence for Instinct, and what are the underlying gross margins for the business starting in Q1 and exiting Q4 2027?
    Response: The Instinct revenue is expected to be well over the implied $30 billion figure, with a progression of growth over the quarters. Gross margins for 2027 will be managed by the balance between the server CPU and data center AI ramps, with multiple levers across the company to support improvement.

  • Question from Joe Moore (Morgan Stanley): Is the server CPU business supply constrained now, and will that persist? Can the supply chain support the growth for next year?
    Response: The server CPU supply chain is tight now due to unforecasted demand, but for 2027, demand is better forecasted, and the supply situation is expected to be better. The supply chain is being worked on comprehensively to meet the strong growth targets.

  • Question from Atif Malik (Citi): How do you expect the fast inference market to grow, and what is the partnership with Cerebrus on desegregate compute?
    Response: The inference market is growing substantially. The partnership with Cerebrus aims to provide a solution for fast inference, expected to be available in Q4 and extend into 2027, representing an important part of the market.

  • Question from Atif Malik (Citi): How should we think about OPEX growth relative to the 40% market growth?
    Response: OPEX will continue to be invested in, but increases are expected to be managed less than the top-line revenue growth, driving operating leverage to deliver higher EPS.

Contradiction Point 1

2027 Server CPU Growth Expectations

Contradiction on the magnitude of server CPU growth forecast for 2027.

Vivek Arya (Bank of America Securities) - Vivek Arya (Bank of America Securities)

2026Q2: Server CPU revenue is expected to grow over 70% year-over-year in 2027. - [Lisa Hsu](CEO)

What is AMD's projected gigawatt capacity for 2027 and the associated monetization per gigawatt? - Stacy Rasgin (Bernstein Research)

2026Q2: The more than 70% for server CPU is a strong statement. - [Lisa Hsu](CEO)

Contradiction Point 2

2027 Gross Margin Outlook

Contradiction on the expected trend and drivers of gross margin for 2027.

CJ Muse (Cantor) - CJ Muse (Cantor)

2026Q2: Gross margin for 2027 will depend on the balance between server CPU and data center AI ramps. - [Gene Hu](CFO)

What is the implied 2027 revenue, expected revenue cadence, and gross margin trajectory for the Instinct business (MI450/Helios)? - Tom O'Malley (Barclays)

2026Q2: Growth margin is primarily driven by business mix. The server business expansion is accretive to overall growth margin. While data center AI gross margin is slightly below corporate average, its large incremental revenue adds significant gross profit. - [Gene Hu](CFO)

Contradiction Point 3

Server CPU Supply Chain Situation in 2027

Contradiction on whether the supply chain for server CPUs is expected to be constrained or improved in 2027.

Joe Moore (Morgan Stanley) - Joe Moore (Morgan Stanley)

2026Q2: For 2027, demand is better forecasted, and the supply situation is expected to be better than 2026. - [Lisa Hsu](CEO)

Is the server CPU business currently supply-constrained, with tightness expected to persist, and can the supply chain support the >70% growth forecast for 2027? - Joe Moore (Morgan Stanley)

2026Q2: The server CPU supply chain has been tight due to unforecasted demand. - [Lisa Hsu](CEO)

Contradiction Point 4

Server CPU Growth Forecast

Conflicting statements on the primary driver (units vs. ASP) and magnitude of server CPU growth.

Joshua Buckalter (TD Cowan) - Joshua Buckalter (TD Cowan)

2026Q2: Growth has seen both unit and ASP increases... both unit and ASP growth are expected. - [Lisa Hsu](CEO)

Okay, let's see. The user wants me to rewrite the input into one concise earnings-call question. The input is: "What are the unit and ASP assumptions for server CPU growth in 2027, and how should we model the business?" First, I need to make sure I follow the rules. The output must be exactly one line, only the final question text, end with a question mark if possible. No explanations or anything else.The original input has two parts: asking about unit and ASP assumptions for server CPU growth in 2027, and how to model the business. I need to combine these into a single question. Maybe something like asking about the assumptions and their impact on modeling the business. Let me check if that's concise. "Could you provide the unit and ASP assumptions for server CPU growth in 2027 and their impact on business modeling?" Hmm, maybe even shorter. "What are the unit and ASP assumptions for server CPU growth in 2027, and how should we model the business?" Wait, that's almost the same as the original. Maybe the original is already concise enough but has two parts. The user wants it as one question. Perhaps combining the two parts with "and" is acceptable. Let me check the rules again. The output must be one line. The original input is two questions combined with "and". The user wants it rewritten into one. Maybe the original is already a single question with two parts. But the instruction says to make it into one concise question. Maybe rephrase to make it a single inquiry. For example: "What are the unit and ASP assumptions for server CPU growth in 2027, and how should we model the business?" That seems to fit. It's one line, ends with a question mark. The original input is a valid question, but maybe the user wants it more concise. Wait, the original input is already a question. But the instruction says if it's already a question, maybe we can leave it as is? Wait, no, the instruction says if it's not a question, too vague, etc., output as-is. But the original input is a question. However, the user wants it rewritten into one concise earnings-call question. So maybe the original is acceptable, but perhaps the assistant can make it more concise. Let me check the original: "What are the unit and ASP assumptions for server CPU growth in 2027, and how should we model the business?" It's two parts. Maybe combine them into one by specifying the modeling part. For example: "What are the unit and ASP assumptions for server CPU growth in 2027, and how should we model the business?" That's the same as the original. Wait, maybe the user wants to make sure it's a single question. The original input is already a question with two parts. So perhaps the answer is the same as the input. But the user wants it rewritten into one concise question. Maybe the original is already concise. Alternatively, maybe rephrase to make it a single question. For example: "Could you outline the unit and ASP assumptions for server CPU growth in 2027 and the recommended business modeling approach?" That's one question. But the original uses "how should we model the business?" which is more direct. Maybe the original is acceptable. The rules say if the input is a question, but needs to be rewritten, then output the rewritten version. If it's already a question and concise, maybe leave it. But the user wants it rewritten. Hmm. The original input is a valid question. However, the user wants it to be concise. Maybe the original is already concise. Alternatively, maybe the user wants to combine the two parts into a single question without the "and" but that's not possible. I think the original input is acceptable as a concise question. Wait, but the instruction says "rewrite the input into ONE concise earnings-call question." So the input is a question but maybe not concise. The original has two parts. Let me check the length. The original is two questions joined by "and". Maybe it's better to make it a single question. But in English, using "and" is acceptable. For example, "What are X and Y?" is a single question. So the original is a single question. Therefore, the output would be the same as the input. But the user wants a rewrite. Wait, perhaps the user wants it to be a single question without the "and". But that's not possible. Alternatively, maybe rephrase to make it more concise. For example: "What are the unit and ASP assumptions for 2027 server CPU growth, and how should we model the business?" That's slightly shorter. But the original is already concise. I think the original input is acceptable. However, the user might want to ensure it's a single question. Since the input is already a question, and the user wants - Blaine Curtis (Jefferies)

2026Q1: Growth is largely unit-driven (more CPUs shipped across Turin and Genoa generations). ASP increases are modest... - [Lisa Su](CEO)

Contradiction Point 5

Data Center AI/GPU Growth Outlook

Inconsistent guidance on the sequential performance and growth trajectory of Data Center AI/GPU business.

Tom O'Malley (Barclays) - Tom O'Malley (Barclays)

2026Q2: Server CPU growth is expected to be over 80% year-over-year in the second half of 2026, with strong data center AI growth continuing into 2027. - [Lisa Hsu](CEO)

When do you expect the crossover point in the data center GPU and CPU business, and where are the key contributions driving strength for calendar year 2027? - Stacy Rasgon (Bernstein Research)

2026Q1: Data Center AI revenue was down modestly sequentially in Q1, primarily due to lower China revenue... Q2 guidance: Both server CPU and Data Center AI will grow double digits sequentially. - [Jean Hu](CFO)

Discover what executives don't want to reveal in conference calls

Latest Articles

Stay ahead of the market.

Get curated U.S. market news, insights and key dates delivered to your inbox.

Comments



No comments

No comments yet