AMD Q4 2025 Earnings

AMD’s fourth-quarter and full-year 2025 results, painted a picture of a company firing on almost all cylinders. While the headline numbers showed record-breaking revenue and earnings per share, the market’s reaction, a 9% drop in after-hours trading, suggests that for AI-adjacent semiconductor giants, good is no longer good enough as valuations hit sky-high expectations head on.

Performance vs. Expectations and Q1 Outlook

The Q4 Numbers

AMD reported record quarterly revenue of $10.3 billion, up 34% year-over-year, edging out analyst estimates of $9.7B.

• Non-GAAP EPS: $1.53 (Record), beating the consensus estimate of $1.32.
• Non-GAAP Gross Margin: 57%, a significant jump from 54% in the previous year.
• Data Center Segment: $5.4 billion (Record), up 39% Y/Y.
• Client & Gaming Segment: $3.9 billion (Record), up 37% Y/Y.

The Q1 2026 Outlook

For the first quarter of 2026, AMD provided the following guidance:

• Revenue: Approximately $9.8 billion (+/- $300 million) which is above expectations of $9.4B but I think investors were hoping for a larger beat showing that AMD is really making strides in taking share from NVDA.
• Non-GAAP Gross Margin: Approximately 55%.
• The China Factor: The guidance includes ~$100 million of AMD Instinct MI308 sales to China.

Why the 9% After-Hours Slide?

Despite beating on the top and bottom lines for Q4, the stock plummeted. Several factors likely contributed to this:

1. Whisper Numbers: In the current AI bull market, investors often look for beats and raises that far exceed consensus. AMD’s Q1 revenue guide of $9.8 billion wasn’t too far away from expectations on the bottom end, offering no major shock and awe upside. Given that the P/FC and P/E ratios are already high, investors are re-assessing whether the valuation stretched a bit.
2. China Revenue Quality: A portion of the Q4 gross margin beat was driven by a $360 million release of inventory reserves related to the MI308 chips for China. While this boosted the reported margin to 57%, management clarified that without this one-time benefit, margins would have been 55%.
3. Gaming and Embedded Softness: While Data Center is the star, the Gaming segment isn’t growing fast, and the Embedded segment is only just beginning a modest recovery.
4. Profit Taking: AMD stock has seen a massive run-up leading into earnings with the stock up over 100% in the last year; any guidance that isn’t perfect often triggers a sell-off in a high-valuation environment.

What’s Going Well vs. What’s Going Poorly

The Bulls (What’s Going Well)

• Server Market Share: AMD continues to eat Intel’s lunch. The 5th Gen EPYC Turin CPUs are seeing rapid adoption, now making up over half of server revenue.
• AI Momentum: The Data Center AI business is accelerating. AMD now has 8 of the top 10 AI companies using Instinct GPUs for production workloads.
• Client Strength: The Ryzen PC business is at record levels. Desktop CPU sales set records for four consecutive quarters, and commercial notebook adoption grew over 40%.
• Free Cash Flow: AMD generated a record $2.1 billion in FCF in Q4, nearly doubling year-over-year. Still this is a company that is expected to have less than $10B in free cash flow next year trading at a market cap just shy of $400B. A 2% forward free cash flow yield and a high 30s forward P/E might be too rich for some and bakes in a lot of expected top line and bottom line growth.

The Bears (What’s Going Poorly)

• Gaming Consolidation: Semi-custom revenue (consoles) is expected to decline by a significant double-digit percentage in 2026 as the current console cycle reaches its 7th year although they did let it slip that Microsoft is potentially releasing a new Xbox in 2027.
• Concentration Risk: The partnership with OpenAI is massive, but it makes AMD’s 2026/2027 success heavily dependent on a single, albeit major, customer.
• Export Restrictions: Continued uncertainty regarding U.S. export controls on high-end GPUs to China remains a persistent headwind and complicates forecasting.

The Fight Against NVIDIA: Technology and Strategy

AMD is not just the other guy these days. Dr. Lisa Su has shifted the company’s roadmap to a yearly cadence to match the pace of the AI industry.

Tech Advances to Counter NVDA:

1. The MI400 Series and Helios: AMD’s answer to NVIDIA’s Blackwell is the MI400 series. The Helios rack-scale platform is designed for yotta-scale AI infrastructure.
2. Memory Leadership: AMD is leaning heavily into HBM (High Bandwidth Memory). The upcoming MI500 series (planned for 2027) will utilize HBM4E and 2nm process technology.
3. The Open Software Play (ROCm): AMD’s biggest hurdle has been NVIDIA’s CUDA software moat. AMD is countering this with ROCm, an open-source ecosystem. In Q4, they achieved Day Zero support for leading AI models and integrated AMD GPUs into vLLM, a popular inference engine.
4. CPU/GPU Synergy: Unlike NVIDIA, which is primarily GPU-first, AMD has the #1 server CPU (EPYC). This allows them to optimize head nodes, the CPUs that manage AI clusters, creating a better together hardware stack.

Management Q&A: Key Takeaways

During the earnings call, Dr. Lisa Su and CFO Jean Hu addressed several critical investor concerns:

• OpenAI Partnership: Management confirmed the multi-year partnership to deploy 6 gigawatts of Instinct GPUs is on track, with the MI450 ramp starting in H2 2026.
• Supply Chain: Unlike previous years of tightness, Lisa Su expressed high confidence that AMD will not be supply-limited for its aggressive 2026 AI ramp.
• OpEx Leverage: Investors were worried about rising operating expenses. Jean Hu clarified that while they are investing heavily in R&D, revenue is expected to grow faster than OpEx in 2026, leading to margin expansion.
• Inference vs. Training: AMD noted that as AI moves into the inference phase (running models vs. training them), the efficiency and tokens per dollar of their chiplet architecture become a major competitive advantage.

Valuation: Is AMD a Buy?

At the current price (post-9% drop), AMD presents a complex valuation case.

• The Targets: Management reiterated their Financial Analyst Day goal: $20+ in annual EPS in the 3-to-5-year timeframe with a revenue CAGR of 35%.
• The Multiple: Historically, AMD trades at a high P/E because of its growth trajectory. If they can achieve $20 EPS by 2030, the current stock price looks incredibly cheap. If they’re still growing at that point then a 20-30x multiple isn’t crazy putting the price point at $400-$600 meaning a near double or triple from here post earnings drop.
• The Risk/Reward: The 9% drop has de-risked the entry point for long-term investors. However, the stock is still sensitive to broader AI sentiment and a CAGR of 35% for the next 3-5 years won’t be easy and really depends on continued AI demand.

Verdict:

In my mind, AMD is a buy on the dip for long-term believers of AI. The transition from a PC/Server company to an AI Infrastructure company is not fully complete. While NVIDIA holds the crown today, AMD’s yearly product cadence and leadership in CPUs make them the only viable alternative for hyperscalers (AWS, Google, Microsoft) who are desperate for a second source of high-end AI chips.

If you can stomach the volatility of the gaming cycle and the high expectations of the AI market, the fundamental engine at AMD is stronger than it has ever been. One thing to remember is that technology can develop at a pretty rapid clip and it’s not impossible for AMD to catch up to NVDA which is a $4T+ company. I think the risk/reward here is worth it although it will likely come with significant volatility.

Disclaimer: I am long AMD and may be long other stocks mentioned in this article in the near future. This article is for informational purposes only and does not constitute financial advice. Investors should perform their own due diligence or consult with a financial advisor before making any investment decisions.