Nvidia is the sweetest customer right now. It is the glue of artificial intelligence, investing millions in AI startups, and while Big Tech attempts to position itself as the new Nvidia, the reality is that Nvidia remains the only company with the necessary hardware to meet the demands of hyperscalers. The company’s real achievement lies in making the entire industry revolve around it, compelling everyone to work for its success.
Foxconn Joins Nvidia’s Ecosystem
And Foxconn is the latest company to align itself with Nvidia’s future.
Major Developments
Nvidia is gearing up for the launch of Vera Rubin, a state-of-the-art training and inference platform that integrates cutting-edge hardware from firms like Samsung and TSMC. According to CEO Jensen Huang, Nvidia will require extensive support this year to achieve its ambitious objectives, which is where Foxconn comes into play.
The iconic Taiwanese company has been selected as the exclusive supply chain provider for the necessary equipment related to the Groq 3 LPX. This recent agreement not only represents a tenfold increase in the volume of deliveries but also accelerates timelines, with fulfillment expected by the third quarter of 2026.
Nvidia’s Strategic Direction
This partnership showcases Nvidia’s intentions to innovate further. Previously focused on GPUs like the H200, which were highly sought after by Chinese tech firms, Nvidia is now emphasizing the Vera Rubin platform, capable of both training and inference—essential for modern, agentic AI applications. According to Nvidia, the Groq 3 LPX rack offers 35 times improved performance in inference for billion-parameter AI models over its predecessor, Blackwell.
Market Impact
With Intel pivoting towards Xeons for data centers and ARM stocks hitting record highs thanks to its AI CPU, Foxconn’s role in the LPX platform is vital, especially in accelerating the critical decoding phase of AI models. Additionally, Foxconn is a primary supplier of the Vera Rubin NVL72 cabinets, indicating its pivotal position in this AI revolution.
Industry sources project shipments of LP30 and LP35 chips that utilize these LPX cabinets to reach 1.5 million by 2026 and 2.5 million by 2027. This amounts to a total of 6,000 racks expected this year and 10,000 in the coming year alone.
The Dual Faced Partnership
Working with Nvidia comes with its pros and cons. For Nvidia, this represents a golden opportunity to dominate the AI sector, having effectively turned its back on clients it once depended on. Collaborating with Nvidia assures significant profits as long as its leadership in AI endures. For example, Samsung aggressively pursued the HBM4 memory race, realizing that securing contracts with Nvidia would bolster its competitive stance against larger clients.
Conversely, Foxconn is reaping considerable rewards, obtaining major contracts that extend beyond just Nvidia. It also supplies chips to tech giants like Google, Microsoft, and Amazon. For instance, it holds a 15% market share from Google and is currently capable of producing over 1,000 complete cabinets weekly, aiming to double that capacity by year’s end.
RAM Market Implications
Regarding potential impacts on RAM pricing, each of the thousands of racks produced by Foxconn will contain 256 chips with 128 GB of SRAM and 12 TB of DDR5 memory. This significant demand explains why major memory manufacturers are shifting their focus from DDR4 to DDR5 production, thereby stabilizing prices for the foreseeable future.
In summary, as Nvidia continues to ascend in the AI landscape, its partnerships, notably with Foxconn, exemplify its strategy of transforming the industry in its favor. As a result, the tech supply chain is poised for significant changes, upending traditional alliances and creating new opportunities within the sector.
Image | Hillel Steinberg