Samsung Electronics' foundry division has begun "allocation" in some processes to apportion production volumes by customer, according to industry sources. As demand surges on the back of an expanding artificial intelligence (AI) semiconductor market and rising orders from global big tech, the company appears to be pursuing a "selection and concentration" strategy by taking new customer orders on a limited basis.
According to the industry on the 2nd, Samsung Electronics' foundry division has recently been adjusting supply priorities by allocating volumes to existing customers first and selectively taking orders from new customers. Similar shifts in supply and demand are being detected among major domestic design houses (DSPs) that make up the Samsung foundry ecosystem.
A design house industry official said, "Starting this year, allocation has been applied to the Samsung foundry process," and noted, "Rather than taking every customer order unconditionally, the mood is to choose and focus on projects with clear prospects."
◇ Tight supply in some processes as AI Semiconductor demand expands
The industry says explosive growth in the AI market is fundamentally reshaping foundry demand. Demand for advanced processes, once centered on smartphone application processors (APs), has recently shifted to AI accelerators, application-specific integrated circuits (ASICs), and chips for high-performance computing (HPC), leading to a flood of orders from global big tech corporations.
In fact, Samsung Foundry is producing Tesla's Autonomous Driving chips and AI startup Groq's AI inference chips, while also expanding collaboration with global AI corporations such as Nvidia and Google. The industry assessed that this big tech demand is tightening supply and demand in some processes.
In particular, Samsung Foundry's 4-nanometer (nm) process is reportedly almost entirely sold out through next year, and parts of the 8-nm process are said to be running at virtually full capacity.
Accordingly, there is analysis that order-taking strategies are being reorganized around large customer projects with higher line-operation efficiency to maximize production efficiency at fabs whose utilization has reached its limits.
From a foundry operations perspective, evaluations indicate that, compared with a "high-mix, low-volume" approach that mass-produces numerous product types, a "low-mix, high-volume" approach focusing on a few large projects is more advantageous for fab efficiency and profitability.
Another industry official explained, "From the factory's standpoint, it is far more efficient operationally to focus on producing a few large projects than to mass-produce many types of products."
◇ Windfall from TSMC's supply shortage… absorbing demand for diversified supply chains
The industry believes that a shortage in advanced process supply at Taiwan's TSMC, the world's No. 1 foundry, and the resulting demand from big tech for supply chain diversification (multi-foundry) have contributed to higher utilization at Samsung Foundry.
It is said that customers who left Samsung for TSMC in the past over issues such as Production yield are now reviewing Samsung Foundry again or adopting it as a second source (alternative supplier) to spread supply chain risk and strengthen bargaining power on chip unit prices.
Buoyed by this expanding demand, some assess that Samsung's market position has strengthened to the point where it can raise supply prices for certain processes by around 15% to 20%.
The market views this as a sign of recovery for Samsung Electronics' foundry division, which has suffered prolonged slump and losses. Brokerages, citing the expansion of AI Semiconductor demand, rising utilization, and the impact of large customer orders, forecast that the Samsung foundry division could return to profit in the second half of this year or next year.
However, because higher utilization is concentrated in certain nodes, some say further observation is needed to determine whether it will lead to profitability improvement for the division as a whole, given the burden of large depreciation costs.