Samsung Electronics (005930) has entered final-stage negotiations with Anthropic, the developer of generative AI model Claude, to produce next-generation AI semiconductors. If the deal is finalized, adding another marquee customer alongside Tesla, Nvidia, and Apple, it would signal a full turnaround for Samsung’s foundry business, which has struggled in recent years.

According to foreign media reports and semiconductor industry sources on July 3, Anthropic has initiated early-stage work on developing its own AI chips and is strongly considering Samsung Electronics as its manufacturing partner. U.S. tech publication The Information, citing multiple sources, reported that Anthropic is discussing plans to leverage Samsung Foundry’s cutting-edge 2-nanometer (nm) process and advanced packaging technologies.

The collaboration extends beyond simple contract manufacturing. Samsung Electronics is expected to work closely with Anthropic from the micro-circuit design stage within the chip. High-difficulty packaging technology that combines multiple chips to function as one will also be applied. Samsung, however, maintained its standard position, stating it “cannot confirm details regarding customers.”

2nm Process Emerges as the Answer, Samsung Rises as TSMC Alternative

Anthropic’s selection of Samsung as a partner is seen as a result of Samsung recently proving its technological competitiveness by raising 2nm process yields above 60%. Additionally, as orders for AI chips flood TSMC, the world’s top foundry, supply bottlenecks have intensified, prompting Big Tech companies to actively seek alternatives.

“Big Tech firms feeling the burden of TSMC’s process service pricing are beginning to view Samsung Electronics as a realistic option,” an industry insider said, adding that “the likelihood of additional order wins is high.” Indeed, Samsung recently secured production deals for Nvidia’s inference chip, the Grok-3 LPU, and Tesla’s autonomous driving chip, the AI6. Meta is also pursuing plans to manufacture the third generation of its next-generation AI accelerator, MTIA, using Samsung’s 2nm process.

‘Post-GPU’ Strategy Reduces Nvidia Dependency

This collaboration is part of Anthropic’s strategy to internalize its AI infrastructure. Anthropic is pursuing the long-term construction of its own data centers with a capacity of approximately 1GW (gigawatt). Total investment is estimated at around $50 billion (approximately ₩76.5 trillion), with roughly half expected to be allocated to AI semiconductors. Faced with soaring prices and supply shortages of Nvidia GPUs, the company is accelerating efforts to secure its own chips.

This trend is spreading across global Big Tech. Google continues to advance its proprietary Tensor Processing Units (TPUs), while OpenAI partnered with Broadcom to unveil its first custom inference chip, ‘Halapero.’ Anthropic also recently established a dedicated team by hiring Clive Chan, an early member of OpenAI’s custom chip development team.

Samsung Electronics is considered the biggest beneficiary of this market shift, as it is the only company with integrated semiconductor capabilities spanning memory, foundry, and packaging. When Anthropic conducted its $65 billion Series H funding round last May, Samsung participated as a “strategic infrastructure partner” alongside SK Hynix (000660) and Micron. Anthropic emphasized at the time that “the technology of these companies plays a key role in the global supply of memory, storage, and logic chips.” Among the three, Samsung is the only one with logic chip foundry capabilities.

Green Light for Foundry Profitability

The string of major order wins is expected to directly improve the performance of Samsung’s foundry division. Industry observers project that Samsung Foundry’s mid-to-long-term order backlog is approaching ₩50 trillion (approximately $32.7 billion). Han Jin-man, head of the foundry business, recently stated at an internal briefing that “the likelihood of achieving profitability by 2028 is high,” but market expectations are growing that an operating profit turnaround could occur as early as the fourth quarter of this year.

Samsung’s Taylor, Texas facility, set to begin full-scale operations early next year, is also expected to contribute. The plant, which will produce Tesla’s AI chips based on the 2nm process, is poised to become a key hub for Samsung to secure additional Big Tech customers in the United States.

Anthropic Strengthens ‘Technological Sovereignty’ Amid China Tracking Code Controversy

Meanwhile, Anthropic recently experienced a controversy after embedding tracking code in its coding tool ‘Claude Code’ to identify users in China, which it later withdrew. The move was an attempt to block “distillation” techniques used by China’s Alibaba Group and others to unauthorizedly collect Claude’s responses and train their own AI models.

Anthropic has consistently criticized Chinese AI firms such as DeepSeek, Moonshot AI, MiniMax, and Alibaba for illegally using Claude to train competing models. In response, Alibaba has directed all employees to stop using all of Anthropic’s products, including Claude, starting July 10, and to use its proprietary coding platform ‘Qoder’ instead. Reuters reported that the measure stems from escalating tensions between the two companies over technology leakage.

As global conflicts over AI technological supremacy intensify, Anthropic’s move to secure independent semiconductors is interpreted as a strategic effort to strengthen technological independence and security, beyond mere cost reduction. For Samsung Electronics, this represents a golden opportunity to position itself as a critical supply chain partner in this race for technological sovereignty.