Advanced Micro Devices (AMD) painted an ambitious vision of reaching "billions in annual revenue" from its AI business during its latest earnings call, but failed to fully address investor concerns about near-term growth momentum.
CEO Lisa Su for the first time quantified AMD's long-term AI revenue target, projecting that data center AI business could reach "billions in annual revenue" by 2027. She also revealed that the first batch of next-generation MI450 series accelerators under its landmark collaboration with OpenAI would begin deployment in the second half of 2026.
However, these optimistic long-term projections, coupled with better-than-expected Q4 revenue guidance, couldn't prevent AMD shares from falling over 3% in after-hours trading. According to reports, some investors had expected even stronger guidance, suggesting the market believes AMD's AI payoff may come slower than previously anticipated.
During the call, Su faced repeated analyst questions about when AI chip growth would "truly accelerate." A key takeaway was that AMD's traditional server business growth last quarter slightly outpaced its much-hyped AI chip division - contrasting sharply with market expectations positioning AMD as "the next Nvidia" in AI.
Key Call Highlights:
1. AI Revenue Target: CEO Lisa Su explicitly stated AMD's AI business is entering a new growth phase targeting "billions in annual revenue" by 2027.
2. OpenAI Collaboration: Details emerged about the OpenAI partnership, with the first 1GW of MI450 series accelerators planned for H2 2026 deployment, seen as strong validation of AMD's full-stack solutions.
3. Next-Gen Roadmap: The MI400 series AI chips and Helios rack-scale solutions are on track for 2026 launch, with deep technical collaborations already secured with major clients including OpenAI and Oracle.
4. ROCm Software Progress: The release of ROCm 7 showed significant performance improvements, with open-source strategy gaining support from developers like Hugging Face.
Growth Concerns Persist: Despite management's "billions" AI revenue vision, analysts repeatedly pressed about near-term catalysts. Management acknowledged that Q3 traditional server CPU growth slightly outperformed the data center AI (GPU) business, becoming a focal point of investor concern.
While Q4 revenue guidance (~$9.6B) beat average analyst estimates, it failed to meet some investors' lofty expectations for explosive AI-driven growth. Combined with anticipated gaming segment declines, this triggered market reassessment of growth trajectory.
China Business Uncertainty: AMD confirmed it has obtained some export licenses for China-specific MI308 chips, but Q4 guidance includes no revenue from this product, adding near-term uncertainty.
AI Growth Pace vs. Server Business Surprise: For a company positioned as Nvidia's AI challenger, AI growth pace is the ultimate metric. Yet the biggest surprise was AMD's traditional server CPU business outperforming AI GPUs. When asked directly about Q3 data center growth, management stated server growth was "directionally similar but slightly better" than GPU growth - validating some investor worries about the AI growth narrative supporting AMD's valuation.
Client business (PC processors) grew 73% YoY, far outpacing data center's 22% growth, further highlighting AI business growth hasn't yet met investor expectations.
Long-Term Vision vs. Near-Term Growth: Facing growth pace questions, Su redirected focus to the longer term, reiterating optimism about the AI market potentially exceeding previous $500B TAM estimates. However, investors remain concerned about the growth path before next-gen MI400 series and Helios solutions begin major deployments in H2 2026. Analysts specifically raised concerns about potential "gap periods" as customers might pause MI350 purchases ahead of MI400 availability. Su responded that MI350 should continue growing through H1 2026, but market skepticism remains.
OpenAI Details and Whale Clients: The OpenAI collaboration emerged as a call highlight, with confirmation of 6GW Instinct GPU deployments starting H2 2026. Su emphasized this as a "very important" multi-year, multi-generation partnership, with AMD working closely with OpenAI and cloud partners to ensure power and supply chain readiness. She positioned OpenAI's choice of AMD for most complex workloads as validation of Instinct GPU and ROCm stack's performance and TCO advantages, potentially creating "well over $100B" revenue in coming years.
Other major wins include Oracle as MI450 launch partner planning "tens of thousands" GPU deployments from 2026, and the U.S. Department of Energy selecting upcoming MI430X GPUs and Epic Venice CPUs for its next-gen "Discovery" supercomputer. Su stated AMD aims for broad customer diversification while scaling supply chain to support multiple OpenAI-scale clients.
Financial Performance: Q3 revenue grew 36% YoY to record $9.2B, with data center up 22% ($4.3B), client up 73% ($4B), and gaming up 181% ($1.3B). Embedded declined 8% ($857M). Q4 guidance projects ~$9.6B revenue (+25% YoY), with data center expected to show double-digit sequential growth. Gross margin guidance stands at ~54.5%.
During Q&A, analysts focused on: - CPU/GPU growth dynamics and MI350-MI400 transition management - OpenAI partnership specifics and scalability - Rack-scale solution adoption timeline - Supply chain constraints (power, components) - ROCm software ecosystem progress - China business potential - Customer concentration risks - Gross margin trajectory - Long-term AI TAM potential
Su emphasized strong customer momentum building toward 2026 MI450/Helios launch, with multiple customers expected to reach significant scale. The company continues investing heavily in ROCm development and sees the OpenAI partnership accelerating software stack improvements. While not providing 2026 specifics ahead of Financial Analyst Day, management expressed confidence in demand environment and execution capabilities to capture meaningful AI market share.
Comments