SK Group Chair Chey Tae Won says Anthropic has asked SK Hynix for supplies to make its own chips, calling it remarkable that an AI developer has chip ambitions
AI developer Anthropic PBC has asked SK Hynix Inc., one of the world's biggest manufacturers of memory chips …
Context & Ripple Effects
Anthropic’s approach to SK Hynix follows reports that it had begun early work on a custom AI server chip and discussed manufacturing with Samsung, making memory procurement a concrete extension of that early custom-chip effort.
The request arrives as SK Hynix is being pushed to add capacity amid a reported memory shortage, after the company had already committed heavily to HBM manufacturing in its broader AI investment plan HBM-focused investment plan. That makes a new prospective buyer consequential even before a supply agreement is disclosed.
First-order effects
- Anthropic gains a potential memory-supply path for its chip program; SK Hynix gains another prospective AI-chip customer, though the report does not establish a deal or volume.
- The request puts memory availability alongside chip design and foundry access as an immediate dependency for Anthropic’s in-house hardware effort.
Second-order effects
- Demand from an AI developer pursuing custom silicon could add pressure to SK Hynix’s allocation and expansion decisions, particularly while the company is already responding to capacity constraints reported memory-capacity pressure.
- Other AI developers weighing custom chips have a clearer signal that securing memory supply may need to begin during design, rather than after a processor reaches production.
Third-order effects
- If AI model developers increasingly source components for proprietary chips, competition will shift from buying finished compute to coordinating a broader hardware supply chain spanning design, manufacturing and memory.
- This points toward a more vertically coordinated AI infrastructure market, but whether it changes supplier power depends on these exploratory requests becoming sustained production demand.
The trend: AI developers are moving from dependence on merchant accelerators toward selectively integrated hardware stacks, with memory supply becoming a strategic constraint.