Samsung’s $73 Billion AI Chip Bet Resets Barrier to Entry in Semiconductor Wars
Record capital allocation signals Samsung's all-in push to break SK Hynix's memory dominance and challenge TSMC's foundry supremacy — only firms with $50B+ annual capex need apply.
Samsung announced a record 110 trillion won ($73.3 billion) investment for 2026 chip capacity expansion and R&D, a 22% increase from 2025 spending that positions the company as the highest single-year capital allocator in semiconductor history. The commitment, disclosed on Bloomberg, surpasses TSMC’s ~$50 billion capital expenditure and marks a strategic inflection point in the AI chip wars. Samsung is targeting three simultaneous battles: reclaiming HBM (high-bandwidth memory) market share from SK Hynix, validating its foundry process for AI accelerators, and establishing itself as the primary US-aligned semiconductor manufacturer outside Taiwan.
The HBM Market Realignment
SK Hynix currently commands Reuters 53-57% of the HBM market as of Q3 2025, with Samsung holding 35% and Micron at 11%. The investment is designed to close that gap as HBM capacity remains sold out through 2026 and the total addressable market accelerates from $35 billion in 2025 to a projected $100 billion by 2028, per Introl — a ~40% compound annual growth rate driven by AI infrastructure buildout. Samsung is competing for 20-30% of Nvidia’s HBM4 allocation, with TrendForce projecting the company will surpass 30% global market share in 2026.
“The emergence of agentic AI is driving an extraordinary spike in customer orders, encompassing both memory components and enterprise storage solutions.”
— Jun Young-hyun, Vice Chairman and Co-CEO, Samsung Device Solutions Division
Jun Young-hyun disclosed at a shareholder meeting that Samsung HBM4 customers told him “Samsung is back,” signaling renewed competitiveness after trailing SK Hynix in prior HBM generations. The company is accelerating construction of P4 and P5 fabs at its Pyeongtaek campus, with P5 expected operational by 2028, according to Business Korea.
Foundry Validation Through Nvidia Partnership
Samsung’s Foundry business secured critical validation through its manufacturing agreement with Nvidia for the Groq 3 LPU (language processing unit) inference chip. The company is producing the chip on its 4nm process, with shipments expected in Q3 2026, per Tom’s Hardware. Production ramped from ~9,000 wafers to ~15,000 wafers as Samsung transitioned from samples to commercial manufacturing in March 2026.
The foundry wins extend beyond Nvidia. Samsung secured a $23 billion contract to produce Tesla’s A16 AI chip at its Texas facility, validating the company’s 2nm process and providing geopolitical risk hedging through US-based production, according to AInvest. The company also partnered with Nvidia to deploy 50,000 GPUs in an “AI Megafactory” configuration, achieving 20x performance gains in computational lithography — effectively using AI to manufacture AI Chips.
Capital Intensity as Competitive Moat
| Company | 2026 Capex | YoY Change |
|---|---|---|
| Samsung | $73.3B | +22% |
| TSMC | ~$50B | — |
| Intel | ~$25B (est.) | — |
The $73.3 billion investment resets expectations for barrier to entry in advanced semiconductor manufacturing. Only Samsung, TSMC, and potentially Intel can sustain $50+ billion annual capital expenditure across foundry, memory, and packaging. South Korea’s government reinforced this positioning with a $23 billion support package and K-Chips Act offering up to 25% tax incentives for Samsung and SK Hynix, per the Wilson Center.
Jun Young-hyun emphasised Samsung’s unique vertical integration: “The Device Solutions division is the world’s only semiconductor company capable of providing one-stop solutions ranging from logic to memory, foundry contract chip manufacturing, and packaging.” This end-to-end capability — spanning HBM memory, 2nm/3nm foundry processes, and advanced packaging — differentiates Samsung from pure-play foundries like TSMC or memory specialists like SK Hynix.
The investment occurs against intensifying US-China semiconductor decoupling and Taiwan Strait risk concerns. Samsung’s US-aligned positioning and planned Texas foundry operations provide supply chain diversification for American customers wary of Taiwan concentration. The K-Chips Act mirrors US CHIPS Act incentive structures, coordinating industrial policy between allies while excluding China from advanced node access.
What to Watch
Samsung’s HBM4 qualification timeline with Nvidia and AMD will determine whether the company recaptures market share from SK Hynix in 2026. Watch for announcements of additional foundry design wins at the 2nm node — particularly from cloud hyperscalers (Amazon, Google, Microsoft) developing custom AI accelerators. The P5 fab construction schedule and 2028 operational target will signal whether Samsung can maintain $70+ billion annual capex through the decade. Finally, monitor SK Hynix’s response: if the incumbent matches Samsung’s investment scale, the HBM market consolidates into a two-player duopoly with Micron marginalised. If SK Hynix cannot match the capital intensity, Samsung’s vertical integration advantage becomes decisive.