Anthropic CEO Dario Amodei announced new supply agreements with Samsung Electronics and SK hynix at an AI summit in South Korea on July 27, though he withheld specifics on components, deal value, or timeline, Business Today reported.
“We recently signed investments and supply agreements with some of the world’s biggest memory providers, including Samsung and the SK Group,” Amodei said, according to a Reuters transcript of his remarks.
The announcement follows AMD’s commitment of up to $5 billion in equity investment in Anthropic, announced July 22, to deploy Instinct MI450-series GPUs in Helios rack-scale systems starting H1 2027. Two major infrastructure deals in five days signals that securing compute supply has become an operational priority for frontier AI labs, not a routine procurement task.
Custom Chips Move from Plan to Procurement
SK Group Chairman Chey Tae-won separately disclosed that Anthropic has submitted chip material supply requests to SK Hynix, according to Winzheng. Chey publicly commented that an AI company’s ambition to develop its own chips is “a significant undertaking.”
Anthropic’s custom chip direction first surfaced in April 2026, covering both ASIC and GPU designs, according to Winzheng’s analysis. SK Hynix participated in Anthropic’s Series H funding round in May 2026. The chip material supply request moves the self-development timeline from design phase to active procurement.
The Supply Chain Map
Anthropic now has supply relationships with AMD (GPUs, $5B equity partnership), Samsung (memory and potentially foundry capacity), SK hynix (HBM memory, chip materials), and Micron (named as a strategic technology partner during Anthropic’s May fundraise at a $965 billion post-money valuation), according to Business Today.
The pattern mirrors what Google, Amazon, and Meta have done: start with data center construction, then extend to chip design. In November 2025, Anthropic announced a $50 billion data center investment plan with Fluidstack for hyperscale facilities in Texas and New York, according to Winzheng. The Samsung and SK deals represent the next layer down: securing the physical materials and manufacturing capacity for custom silicon.
Competitive Context
Every major frontier AI lab is now building or acquiring custom hardware. Google has TPU, Amazon has Trainium, and Meta is investing in custom chip development. Anthropic joining the custom chip race with active supply requests to SK Hynix, while simultaneously locking in AMD GPU capacity, suggests the company is hedging: buying current-generation GPUs for near-term training while developing proprietary silicon for longer-term cost control.
The specific terms of neither the Samsung nor SK hynix agreements have been disclosed. Whether both parties will reveal procurement volumes or technical specifications in the second half of 2026 will indicate how far the custom chip program has progressed beyond the material request stage.