The demand for AI inference chips is heating up, reshaping the order pattern of Samsung's wafer foundry.
On March 10, according to the Chosun Ilbo citing informed sources, AI chip startup Groq has requested Samsung Electronics to increase its AI chip production from about 9,000 wafers to 15,000 wafers, an increase of approximately 70%.
Reportedly, Groq was &#34;indirectly acquired&#34; by NVIDIA for about $20 billion last December. NVIDIA did not directly take over management control but stated that it would cooperate through a non-exclusive technology licensing agreement. The industry generally believes that this move aims to extend NVIDIA's ecosystem into the inference field to complement its dominant position in the AI training chip market.
It is noteworthy that NVIDIA is expected to release a chip based on Groq's design, optimized for inference, at GTC 2026. Reports cite sources that this chip may use SRAM instead of the HBM commonly used in traditional AI chips, a design choice that is expected to significantly reduce inference latency and power consumption.
Meanwhile, according to South Korean tech media The Elec, Tesla has delayed multiple project wafer production plans, causing the next-generation NPU mass production schedule of South Korean AI chip company DeepX to be postponed by about six months.
These developments reflect the complex situation currently faced by Samsung's wafer foundry: on one hand, the rapidly expanding demand for AI inference chips brings new customers and incremental orders; on the other hand, the production rhythm changes of major customer Tesla have had a substantial impact on other customers on the same production line.
<h2>Groq Expands Production, Samsung's Wafer Foundry Enters the Inference Chip Market</h2>
According to reports citing industry insiders, AI chip startup Groq's production at Samsung Electronics is moving from the sample stage to mass production. Previously, Groq's production at Samsung was primarily focused on sample chips for evaluating the applicability of AI inference scenarios, and this year it is considered to officially enter the early stage of large-scale commercialization.
Although the absolute scale of Groq's orders entrusted to Samsung remains relatively limited, this order is expected to help Samsung's wafer foundry establish a foothold in the inference AI chip market.
In addition, the processor developed by South Korean inference AI chip startup HyperAccel is also entirely produced by Samsung's wafer foundry using a 4-nanometer process, further highlighting Samsung's layout in this niche market for inference chips.
<h2>Tesla Delays Affect DeepX, 2-Nanometer Production Line Faces Scheduling Conflicts</h2>
Samsung's wafer foundry's 2-nanometer production line is facing order disruptions from Tesla. According to The Elec citing informed sources, Tesla has delayed its multiple project wafer (MPW) production plans, directly impacting the South Korean AI chip company DeepX, which also relies on this production line.
MPW is a model that allows customers to share wafers to share prototype manufacturing costs. DeepX was originally scheduled to start MPW production of its second-generation NPU chip DX-M2 in April this year, but it has now been postponed by about six months. Reports indicate that the DX-M2 is the first external customer chip designed specifically for Samsung's 2-nanometer process, while Tesla is also using this process to produce its AI6 chip
The specific reasons for Tesla's delay in MPW have not yet been disclosed. According to reports citing industry insiders, adjustments to the production schedules of autonomous vehicles and humanoid robots, as well as changes in supercomputer investment plans, may have collectively contributed to this delay.
DeepX's clients reportedly include Samsung Electronics, Hyundai Motor, and Intel, and the company recently delivered chips to Baidu.
<h2>Tesla May Significantly Expand Production of AI6 Chips, Samsung's 2-Nanometer Capacity Under Scrutiny</h2>
Despite recent order disruptions, Tesla's partnership with Samsung continues to deepen. According to The Elec, Tesla has recently dispatched a procurement executive to Samsung Electronics to discuss expanding the production of its 2-nanometer AI6 chips.
The original contract stipulated a production volume of 16,000 wafers per month, and if a new agreement is reached, the monthly output may increase to approximately 40,000 wafers, more than doubling the output.
This potential expansion plan indicates that Tesla's long-term reliance on Samsung's 2-nanometer process may significantly deepen, and it also means that Samsung's 2-nanometer foundry capacity will face tighter scheduling pressures. Coordinating capacity allocation among multiple clients such as Tesla and DeepX will become a key operational challenge for Samsung in the near future.
Risk Warning and Disclaimer
The market carries risks, and investment should be approached with caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at one's own risk

SSNGY

NVDA

NVDS

With the rising demand for AI inference chips, Groq has requested Samsung Electronics to increase the production of AI chips from 9,000 units to 15,000 units, an increase of approximately 70%. Groq was indirectly acquired by NVIDIA last year and is expected to release a new chip based on Groq's design at GTC 2026. Meanwhile, Tesla has delayed its production schedule, affecting the mass production timeline of the South Korean AI chip company DeepX. This dynamic reflects the complex situation of Samsung's foundry in the AI inference chip market, with new customer incremental orders as well as challenges from the production pace of major clients

- Samsung's foundry is reshaping due to increased demand for AI inference chips, particularly from AI startup Groq, which aims to raise its wafer production from 9,000 to 15,000 units.  
- However, Tesla's delays in multiple project wafer production are impacting South Korean company DeepX's timeline for its next-generation NPU.  
- Samsung faces challenges in balancing production capacity among clients, as the demand for AI chips grows while major customers adjust their production schedules.

Wallstreetcn

Samsung Electronics 

NVIDIA

AXS 1.5X NVDA Bear Daily ETF

SPDR S&P Semicon

VanEck Semiconductor ETF

YieldMax NVDA Option Income Strategy ETF

Direxion Semicon Bull 3X

Invesco Semiconductors ETF

Direxion Daily NVDA Bull 2X Shares

T-Rex 2X Long NVIDIA Daily Target ETF

GraniteShares 2x Long NVDA Daily ETF

iShares Semiconductor ETF

NVIDIA expands its AI territory, Groq's Samsung AI chip orders may increase by 70% to 15,000 units