NVIDIA Soars Near New Highs: Tech Giants Rally Around AI Leader

On Tuesday, global gaming and computing powerhouse Nvidia saw its stock surge over 3%, reaching $132.24—just shy of its historic closing high of $135.58 from June. While there was no groundbreaking news from Nvidia itself over the past couple of days, many tech giants highlighted the importance of Nvidia's advanced chips during their business updates.

On the same day, Foxconn held its annual Technology Day, where Chairman Liu Yangwei announced plans to construct a major AI server manufacturing facility in Mexico. This development aims to meet the soaring global demand for Nvidia's latest GB200 system. Liu noted the extraordinary demand for Blackwell chips and emphasized that the Mexican facility will have significant production capacity, utilizing Nvidia's Omniverse software to create a highly automated environment designed for AI server production.

Meanwhile, Microsoft’s Azure cloud services proudly revealed they had acquired servers equipped with GB200 chips, becoming the first cloud service provider globally to implement the Blackwell system. These new servers feature Infiniband interconnect technology and innovative closed-loop liquid cooling systems.

Interestingly, as Microsoft celebrated its new servers, news outlet The Information reported that OpenAI—valued at over $150 billion—may be planning to reduce its reliance on Microsoft. Reports suggest that OpenAI is negotiating with Oracle to lease the entire Abilene data center, which is expected to support a scale of 1 gigawatt by 2026—sufficient to power large clusters of Nvidia chips.

Nvidia's influence extends beyond end-user chips to the semiconductor manufacturing industry. The company announced that leading chipmaker TSMC is utilizing Nvidia's computing lithography platform, cuLitho, to accelerate production and push the physical limits of next-generation advanced semiconductor chips. Computing lithography is a critical step in chip production, involving the transfer of circuits onto silicon wafers. Nvidia highlighted that a typical chip mask set may require over 30 million CPU computing hours, but with Nvidia's GPU hardware, just 350 H100s can replace a system of 40,000 CPUs, offering faster production at a lower cost, space, and energy consumption. Furthermore, leveraging AIGC can double the speed of workflows accelerated by cuLitho.

In a noteworthy gesture, South Korean memory giant Samsung issued a rare public apology after announcing a profit warning due to delays in delivering HBM3E chips to a major client. In the competitive field of high-bandwidth memory suppliers for AI servers, SK Hynix and Micron are key suppliers for Nvidia.