Shares of Advanced Micro Devices (NASDAQ: AMD) rallied more than 7% intraday before closing up 5.9% on Tuesday, as the chipmaker delivered a powerful counterargument to concerns that artificial intelligence infrastructure spending has peaked. During the Citi Global TMT Conference, AMD management unveiled a significantly more ambitious growth trajectory, doubling its 2030 AI total addressable market projection to an eye-watering $2 trillion.
The revised outlook hinges on a fundamental shift in how AI workloads are evolving, with inference replacing training as the primary growth engine. AMD executives noted that inference has become the dominant driver of AI compute demand over the past twelve months, with workloads moving from basic chatbots toward autonomous agent-based systems. This structural transformation, combined with the unexpected oversubscription of its Helios rack-scale systems by core customers including Meta Platforms Inc, has reinforced management's conviction that the investment supercycle remains firmly intact.
The company's data center segment is now expected to more than double next year, supported by the upcoming launch of the MI450 GPU this quarter, with volume shipments ramping through Q4 and continuing into 2027. The sequential acceleration positions AMD to capitalize on what management describes as a once-in-a-generation compute paradigm shift.
Beyond the headline TAM revision, the real validation came from order book dynamics. Meta Platforms Inc and two other unnamed AI research laboratories have placed orders exceeding their original purchase commitments for the Helios system, providing concrete evidence that demand continues to outpace supply. Furthermore, AMD is cultivating new business from emerging "neo-cloud" service providers, with 2027 Helios shipment expectations already surpassing initial projections.
The company's traditional server CPU business is also firing on all cylinders, with management guiding to year-over-year growth exceeding 80% in the second half of this year and more than 70% next year. While data center GPU margins remain below the corporate average, robust double-digit growth in high-margin server and embedded segments during Q2 and Q3 provides a solid offset, creating a diversified portfolio that supports overall profitability.
On the competitive front, AMD is systematically closing the gap with NVIDIA Corp. Management disclosed deep technical collaborations with three key customers on next-generation MI500 and MI600 chips, alongside a partnership with Cerebras Systems focused on low-latency inference solutions. The company also hinted at internal programs not yet publicly announced, signaling further strategic initiatives in the pipeline.
The first true test of these revised expectations will come with the quarterly shipment cadence of the Helios system following the MI450 launch this quarter. With orders already exceeding initial agreements, the stage is set for what management believes will be a decisive proof-point in the ongoing debate over AI compute demand sustainability.