Austrian technology giant AT&S has officially announced a strategic partnership with AMD and another undisclosed IT company. The goal is a large-scale expansion of production capacity for IC substrates, critical components for artificial intelligence chips and high-performance computing.

The total capital investment in projects in Malaysia and China is estimated to range from €1.5 billion to €2 billion. This is one of the largest investment moves in the semiconductor infrastructure segment in recent months.

Geography and Production Plans

The main focus of the expansion is on the site in Kulim, Malaysia. Here, AT&S intends to utilize existing factory capacities, as well as equip a previously vacant building on the premises of a second plant. Simultaneously, the company is ramping up production in Chongqing, China, to fully cover the demand of a key customer.

Project financing is based on long-term contractual commitments from clients. The finalization of these agreements is expected shortly, indicating a high level of trust from customers and a long-term planning horizon.

Financial Metrics and Forecasts

Against the backdrop of the deal, AT&S has significantly revised its forecast for the 2026/27 fiscal year. The company now expects revenue growth of 45–55%, compared to the previously announced 30–35%. This jump suggests that demand for specialized substrates for AI chips is growing much faster than general market expectations.

For the 2025/26 fiscal year, AT&S revenue amounted to €1.8 billion, a 21% increase over the previous period. The company operates production sites in Austria, China, Malaysia, and India, supplying components for the automotive industry, medical equipment, and the mobile device market.

Analytical Commentary

This agreement is a clear indicator of a tectonic shift in the semiconductor industry. Traditional manufacturers are refocusing on the AI segment, where margins and order volumes are significantly higher. AT&S's investment of up to €2 billion is not just an expansion of capacity, but a bet that the AI chip race is just beginning, and infrastructure support will become a bottleneck within the next 2–3 years.