The NVIDIA GTC 2026 conference is not only a product launch show but also a reshuffling of the supply chain landscape. As the architectural details of NVIDIA's next-generation AI platform Vera Rubin gradually come to light, the roles played by the three major chip giants—Samsung, Micron, and Intel—are also emerging.
According to TrendForce, in the most closely watched supply chain dynamics, NVIDIA CEO Jensen Huang publicly confirmed for the first time that the Groq 3 LPU under its umbrella is produced by Samsung; Micron announced that HBM4 has entered mass production in the first quarter of 2026, breaking the previous rumors of being excluded from the Vera Rubin supply chain. These two pieces of news directly impact the competitive landscape of the HBM market and the bargaining power of suppliers.
Meanwhile, Intel also solidified its partnership with NVIDIA at this conference, confirming that its Xeon 6 processors will provide computational support for the DGX Rubin NVL8 system. Looking further ahead, according to Wccftech, Intel is expected to participate in the packaging production of NVIDIA's next-generation Feynman GPU, set to launch in 2028, as a wafer foundry.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/64177a9f-1da3-4d9b-a380-8e2f67b784cc.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/64177a9f-1da3-4d9b-a380-8e2f67b784cc.png"/>
<h2>Samsung Secures LPU Foundry Order, Jensen Huang Confirms</h2>
Groq 3 is one of the most anticipated releases at this GTC. This LPU, designed for high-speed inference, will be integrated into the Vera Rubin platform and is scheduled to start shipping in the second half of 2026. According to South Korea's Chosun Ilbo, Jensen Huang publicly confirmed at the conference that Groq 3 is produced by Samsung's foundry, continuing the existing foundry agreement between Groq and Samsung prior to NVIDIA's $20 billion acquisition of Groq last year.
On a technical level, the design logic of Groq 3 significantly differs from mainstream AI accelerators. According to Tom's Hardware, each Groq 3 LPU contains 500MB of SRAM—an ultra-fast memory typically used for CPU and GPU caches.
Although this capacity is much smaller than the 288GB HBM4 equipped with the Rubin GPU, its bandwidth reaches approximately 150TB/s, far exceeding the 22TB/s provided by HBM4. For bandwidth-intensive AI inference decoding tasks, this design is expected to significantly enhance inference performance.
Samsung's acquisition of the foundry order means its role in NVIDIA's supply chain extends from HBM4 memory supplier to the logic chip foundry domain, strengthening its strategic position on the Vera Rubin platform.
<h2>Micron HBM4 Mass Production Launched, SK Hynix's Monopoly Premium Under Pressure</h2>
Micron officially announced at this conference that the 36GB 12-layer stacked HBM4 has begun mass production for NVIDIA's Vera Rubin platform in the first quarter of 2026. This product features a pin rate exceeding 11 Gb/s and a bandwidth exceeding 2.8 TB/s, representing a 2.3-fold increase over HBM3E, while power efficiency has improved by over 20%
In addition, Micron has begun sending 48GB 16-layer stacked HBM4 samples to customers, which represents a 33% increase in capacity compared to the 12-layer version.
The market significance of this development lies not only in Micron's own technological breakthrough. According to an analysis by Joseilbo.com, Micron's acceleration in mass production will reduce the concentration of HBM suppliers, creating greater pressure on existing suppliers in terms of shipment distribution and price negotiations. The report points out that the core impact of this move is not to directly erode SK Hynix's market share, but to weaken the monopoly premium formed during peak HBM demand.
Samsung also faces more direct competitive pressure. Joseilbo.com notes that although Samsung has officially advanced HBM4 production to showcase its technological strength, Micron's large-scale supply to NVIDIA's Vera Rubin platform may shift the industry's competitive benchmark from &#34;ability to mass produce&#34; to &#34;actual adoption scale,&#34; posing a new challenge to Samsung.
<h2>Intel's Dual-Line Layout, Feynman Packaging Collaboration Emerges</h2>
Intel's presence at this year's GTC is also noteworthy. Intel has officially confirmed that its Xeon 6 processors will support NVIDIA's DGX Rubin NVL8 system. According to Tom's Hardware, this product offers a 2.3 times increase in memory bandwidth compared to the previous generation, providing scalable high-performance AI computing power for next-generation GPU-accelerated workloads.
In a longer-term strategy, according to Wccftech, NVIDIA intends to collaborate with Intel in the wafer foundry sector, leveraging Intel's advanced packaging technologies, including EMIB, to provide packaging support for the Feynman GPU, which is set to debut in 2028. Notably, the chip itself for the Feynman GPU is expected to be produced using TSMC's 1.6nm process, with Intel's involvement primarily focused on the packaging stage.
The Feynman platform will also introduce 3D chip stacking technology, which may be NVIDIA's first use of 3D stacking design in GPU products. In terms of storage, NVIDIA plans to equip Feynman with customized HBM, rather than standard specification next-generation HBM products, further strengthening its differentiated competitive advantage in AI data center platforms

512760

588780

09747

Jensen Huang confirmed that Samsung is manufacturing Groq 3 LPU, Micron's HBM4 has officially entered mass production, breaking the rumors of "exiting," directly impacting SK Hynix's monopoly premium; Intel, on the other hand, is entering the market with Xeon 6 and positioning itself for the 2028 Feynman GPU with advanced packaging technology, completing a low-profile dual-line bet

- The GTC 2026 conference showcased Nvidia's next-gen AI platform Vera Rubin, revealing major supply chain shifts involving key chipmakers like Samsung, Micron, and Intel.  
- Nvidia confirmed that its Groq 3 LPU will be manufactured by Samsung, while Micron announced large-scale production of HBM4, enhancing competition in the HBM market.  
- Intel established its collaboration with Nvidia, providing its Xeon 6 processors for AI systems and signaling potential joint initiatives for future GPU packaging solutions.

Wallstreetcn

Guotai CES Semiconductor Chip Industry ETF

GTJA Allianz SSE STAR Chip Design Thematic ETF

XL2CSOPSMSN-U

T-Rex 2X Long NVIDIA Daily Target ETF

Samsung Electronics 

Direxion Semicon Bull 3X

Invesco PHLX Semiconductor ETF

Direxion Daily MU Bull 2X Shares

Guotai CSI All Share Integrated Circuit ETF

iShares Semiconductor ETF

XL2CSOPHYNIX

Samsung Electronics Co., Ltd. Sponsored GDR

Invesco S&P 500 Equal Weight Tech ETF

INTEL-T

Harvest CSI All Share Integrated Circuit ETF

SAMSUNG SEMICON

Spdr Select Tech

SPDR S&P Semicon

GraniteShares 2x Long NVDA Daily ETF

XI2CSOPNVDA

GraniteShares 2x Short NVDA Daily ETF

VanEck Semiconductor ETF

YieldMax NVDA Option Income Strategy ETF

China Merchants CSI Semiconductor Industry ETF

Guotai CSI Semiconductor Material Equipment Theme ETF

Invesco Semiconductors ETF

GraniteShares 2x Long MU Daily ETF

GF Fund CNI Semiconductor Chips ETF

Micron Tech

XL2CSOPSMSN

Samsung, Micron, and Intel share the big orders from NVIDIA: Samsung is manufacturing LPU, Micron is mass-producing HBM4