Samsung expects capex to rise ~60% to ~$63B this year, with HBM capacity sold out through 2027 and 10 long-term supply deals secured with key customers

High-bandwidth memory chip capacity is now sold out through 2027, locking in costs for AI infrastructure builders.

Samsung Electronics reported a record operating profit of 107.4 trillion won (USD 80.5 billion) and a record revenue of 195.3 trillion won (USD 146 billion) for the last quarter. This marks the company's fourth consecutive quarter of record profit and revenue, driven by high demand for its memory chips used in AI systems. Samsung anticipates that demand for its AI-related chips will continue to accelerate through 2027, exceeding supply as customers build out AI infrastructure. The company plans to prioritize supplying advanced memory chips and storage products for AI servers, including high-bandwidth memory (HBM), server DRAM, and solid-state drives. This comes as Samsung and competitor SK Hynix, which together dominate the global memory market, invest heavily in semiconductor manufacturing to meet AI-driven demand.

Samsung's announcement of a 60% increase in capital expenditures to approximately $63 billion for the current year, coupled with secured long-term supply deals with key customers, underscores the immense, sustained demand for AI-specific semiconductor components. The company's assertion that HBM capacity is fully booked until 2027 signals a potential bottleneck for AI development, as clients face limited options for securing the necessary memory hardware. This tight supply-and-demand dynamic for crucial AI chips suggests a sustained period of high pricing and significant investment focus within the semiconductor industry. Samsung's strategy emphasizes advanced memory and storage for AI servers, indicating a clear market direction toward specialized hardware solutions catering to the burgeoning AI sector. The substantial capital expenditure planned by Samsung, alongside similar moves by SK Hynix, points to an arms race in semiconductor manufacturing capacity aimed at capturing the AI market share. This investment surge, while promising for meeting demand, also raises questions about long-term profitability and the potential for oversupply if AI growth projections do not materialize as expected.

The market is signaling a significant shift towards sustained, high-level investment in AI-enabling hardware, with memory chips being a critical choke point. Samsung's proactive securing of long-term deals and aggressive capacity expansion demonstrates a strategy to solidify its position in this lucrative, albeit volatile, market segment. This will likely influence the competitive landscape, potentially consolidating market power among a few key players capable of meeting the specialized demands of AI. Businesses reliant on AI infrastructure will need to factor in extended lead times and potentially higher costs for critical components like HBM. Furthermore, the sustained focus on memory and storage for AI servers suggests a continued evolution of data center architectures, prioritizing speed and data throughput to support complex AI computations. Investors and competitors will be closely watching Samsung's execution and market response to these ambitious plans.