With the joint efforts of TSMC, Samsung and other manufacturers, NVIDIA's bottleneck problem is expected to be quickly alleviated, and GPU production capacity is about to usher in an inflection point.
According to Taiwanese media economic reports, Samsung will join the Nvidia HBM3** chain to alleviate the HBM3** problem caused by the previous Hynix family. HBM3 is currently the most advanced memory chip technology, mainly used in NVIDIA's flagship GPU H100.
In addition, NVIDIA's advanced packaging company, TSMC, has also further expanded its advanced packaging production capacity by changing equipment and strengthening external cooperation. Superimpose factors such as external demand-side changes in design and order adjustmentsTSMC is expected to break through the bottleneck of advanced packaging capacity in the first quarter of next year, one quarter to half a year earlier than the industry expects
In September, Wall Street News mentioned that TSMC believed that the shortage of CODOS packaging ** would take about 18 months to ease, dragging down the supply of NVIDIA GPUs. Now the shortage of advanced packaging capacity is expected to end early, which means that the problem of Nvidia's AI chips will also be alleviated. In the previous financial report, Nvidia confirmed that it had certified the production capacity of other CODOS advanced packaging manufacturers as backup. According to the Economic Times, the certification of part of the front-end and back-end production capacity of other COMOS advanced packaging manufacturers will help TSMC and its partners reach the target of about 40,000 pieces per month in the first quarter of next year.
Samsung said in a recent press release that Samsung has been working closely with TSMC to support the compatibility of the CODOS process and the interconnection of HBM in previous generations and current HBM technology. After joining the TSMC OIP 3dFabric Alliance in 2022, Samsung's memory chip division will further expand its scope of work and provide solutions for future HBM generations.
In addition to the expansion of production by the two major leading companies, NVIDIA's main customers, that is, cloud service providers, have successively changed their designs to reduce the amount of advanced packaging, and some repeat customers have deleted orders, which is also a key factor that makes GPU production capacity about to usher in an inflection point.
Industry insiders expect that after the advanced packaging capacity problem is alleviated, it is expected to take about three months from the upstream of semiconductors to the downstream systems and components, which is expected to be conducive to the increase in the shipments of major manufacturers of related supporting components quarter by quarter.
Wall Street news, welcome **app to see more.