At the Citi Global TMT Conference on Tuesday, AMD significantly raised its forecast for the total addressable market (TAM) of AI in 2030 from the previous $1 trillion level to $2 trillion. The company also disclosed that actual orders for its Helios rack systems from core customers, such as Meta, had exceeded initial agreements. Backed by both macroeconomic expectations and micro-level order momentum, AMD’s shares surged as much as 7% during intraday trading before closing up 5.9%.
$Advanced Micro Devices (AMD.US)$ Through a dual reassessment of market size and orders, it is signaling to the market that the investment cycle for AI computing power is far from over.
On Tuesday, at the Citi Global TMT Conference, AMD significantly raised its forecast for the total addressable market (TAM) for AI in 2030 from the previous $1 trillion level to $2 trillion. More critical than this doubled macro expectation is the fact that its Helios rack-level systems have secured excess orders from core clients such as Meta, surpassing initial procurement agreements.
These two positive developments directly drove AMD’s stock price up by as much as 7% during Tuesday’s trading session, closing with a 5.9% gain. This serves as a tangible response to ongoing market debates regarding the sustainability of AI capital expenditure and whether demand for computing power has peaked.
From the surge in inference demand to the strong recovery in server CPUs, AMD’s strategic focus is shifting from mere product launches to the substantive fulfillment of orders and expansion of market share.
Inference Reshapes TAM: The Logic Behind the Shift from $1 Trillion to $2 Trillion
AMD’s doubling of its 2030 AI TAM forecast is not merely a numbers game, but is based on a structural shift in AI computing demand.
According to Citi’s summary of the conference, AMD management stated that "over the past twelve months, inference has become the primary driver of growth in AI computing demand," with related workloads accelerating their migration from basic chatbots to agent-based AI systems capable of autonomous decision-making.
To capitalize on this market expansion, AMD’s data center business is projected to achieve more than double-digit growth next year. The company’s new generation of AI accelerators, the MI450 GPU, will be officially launched this quarter, with volume shipments continuing in the fourth quarter and into 2027.
Helios Excess Orders: Micro-Level Validation of Computing Power Demand
Beyond macro expectations, the order performance of the Helios rack-level systems provides micro-level evidence that demand is still expanding.
Currently, Meta and two other unnamed AI laboratories have become core customers of the system, with their actual demand forecasts exceeding their initial purchase agreements.
As Helios shipments for 2027 are expected to surpass initial forecasts, AMD is continuously tapping into incremental demand from emerging "neo-cloud" service providers. This surge in orders provides a practical response to market debates over whether AI computing demand has peaked.
Server CPU Recovery: Accelerating Market Share Gains
Beyond AI GPUs, AMD's traditional computing business is also demonstrating strong momentum. Company guidance indicates that server CPU revenue is projected to grow by more than 80% year-over-year in the second half of this year, with growth exceeding 70% next year.
Although the current profit margin of the data center GPU business is below AMD's overall average, a diversified product portfolio effectively offsets this drag.
High-margin server and embedded businesses achieved robust double-digit growth in the second and third quarters, supporting overall profitability.
Narrowing the Gap with NVIDIA Through Technological Iteration; Deepening Presence in the Inference Market via Partnership with Cerebras
In response to NVIDIA's ecosystem barriers, AMD is narrowing the competitive gap through technological iteration and external collaborations. Management revealed that it is engaging in in-depth technical cooperation with three core customers on the next-generation MI500 and MI600 chips.
Furthermore, AMD disclosed its partnership with Cerebras and revealed internal, undisclosed plans aimed at further penetrating the low-latency inference market.
With the MI450 GPU launching this quarter and ramping up volume in the fourth quarter, whether Helios can meet its quarterly shipment trajectory will serve as the first key indicator to validate the credibility of the doubled TAM expectation.
Editor/lambor