Samsung 4nm Capacity Fully Booked Through 2027, Driven by NVIDIA Orders and HBM4; 5nm Emerges as Alternative
Miles Bennett
Samsung's 4nm foundry line is fully booked through 2027, locked up by HBM4 base-die production and Nvidia's LP30 inference chip. The overflow is feeding its own 5nm node, opening a second front built on maturity rather than cutting-edge risk.
Why is 4nm full through 2027?
Two clients have consumed all available capacity. First, Samsung's own memory division — HBM4 base dies (the logic chip at the bottom of a high-bandwidth memory stack) are ramping to mass production, making the memory unit the foundry's single largest customer.
Second, Nvidia — its Groq-architecture LP30 inference processor is already in mass production. CEO Jensen Huang publicly thanked Samsung at GTC, and shipment volumes have exceeded initial expectations.
This means → Samsung's 4nm isn't booked by one big order. Internal and external demand lines are pulling at full capacity simultaneously, leaving no room for new clients.
Why can 5nm absorb the overflow?
With 4nm full and 2nm / 3nm still carrying high yield and cost risk, Samsung is pitching 5nm as a "low-risk alternative" to fabless chip designers.
The selling point is straightforward: stable yields and mature performance. Demand has visibly picked up over the past three to four months.
In plain terms = clients are calculating that they don't need the most advanced node — they need one that can ship reliably right now. 5nm sits exactly in that sweet spot.
Which automotive chip orders has 5nm landed?
Samsung's SF5A automotive platform has locked in multiple customers. BOS Semiconductor's Eagle-N automotive AI accelerator will use 5nm, targeting mass production by late 2026.
Telechips' Dolphin7 in-vehicle infotainment processor is confirmed on Samsung 5nm. A separate advanced driver-assistance chip is tied to Hyundai's 2030 production timeline.
This means → 5nm isn't surviving on downgraded 4nm clients alone. Automotive chips are an independent demand line — longer cycle, higher stickiness.
How are TSMC capacity constraints and geopolitics helping Samsung?
5nm applications now extend to servers, high-performance AI chips, and NPUs — neural processing units, modules dedicated to AI inference workloads. Overseas fabless companies have signed agreements.
Some Chinese and Indian chip designers, navigating U.S. export controls and geopolitical risk, are listing Samsung as a backup supplier.
This reflects a broader trend: the tighter TSMC's advanced nodes get and the higher the geopolitical barriers rise, the more valuable Samsung 5nm becomes as a credible second source.
What about 2nm — can it pick up where 4nm leaves off?
Customer inquiries remain active. Samsung and Broadcom have signed an MoU covering 2nm and more advanced nodes. At Samsung's SAFE 2026 client event, 2nm-related inquiries were notably prominent.
But converting inquiries into production orders requires sustained yield improvement and stability — still an unresolved variable.
In plain terms = Samsung's near-term playbook is to defend the 4nm fortress (HBM4 + Nvidia) with one hand, catch overflow demand with mature 5nm with the other, and buy time to get 2nm yields production-ready.
Content is for reference only, not financial advice.