انویدیا با میلیاردها دلار گلوگاه تراشه‌های هوش مصنوعی را می‌شکند

Nvidia has announced the signing of massive, multi-billion-dollar agreements with supply chain partners across substrates, advanced packaging, semiconductor testing, and optical engines to address severe production bottlenecks for artificial intelligence chips. Revealed during the GTC event in Berlin, these long-term agreements aim to resolve critical constraints that have limited the development of current and future AI infrastructures, with production plans extending up to 2028 to ensure stability and sufficient capacity for AI accelerators.

Nvidia CEO Jensen Huang stated that the company is in the middle of the largest infrastructure expansion in human history, noting that demand is extraordinary while manufacturing capacity remains the primary bottleneck. Huang expressed gratitude to the company's global supply chain partners for expanding their capabilities to scale production and help deliver AI supercomputing systems for this new industrial revolution.

The commercial agreements cover two fundamental areas: hardware connections and outsourced semiconductor assembly and test services, known as OSAT. To secure the physical substrate layer, Nvidia has partnered with three major global manufacturers—Ibiden, Unimicron, and Samsung Electro-Mechanics—to achieve a 2.5-fold increase in dedicated capacity over the next three years. In the optical sector, an agreement with SPIL will result in a 15-fold increase in the production capacity of integrated optical engines by mid-2027.

Chip packaging is being pursued through a joint collaboration with Amkor by expanding facilities in South Korea, which will increase interposer wafer capacity by more than 300 percent by the end of 2028. Additionally, Nvidia has allocated 1.1 billion dollars to reserve capacity with King Yuan Electronics, or KYEC, for the final testing of its accelerators, aiming to double chip testing capacity by the year 2027.