Nvidia has begun mass production of its Groq 3 LPX inference accelerator at Samsung Electronics, potentially ending the Korean foundry's four-year losing streak as early as the third quarter of 2026. The move marks a significant turning point for Samsung's contract chipmaking business, which has struggled to compete with industry leader Taiwan Semiconductor Manufacturing Company (TSMC) in advanced chip production.

What Is Groq and Why Does It Matter for AI?

Groq represents Nvidia's strategic push beyond graphics processing units (GPUs) into a broader computing stack designed specifically for artificial intelligence. The Groq 3 LPX is a rack-scale accelerator built with 256 language processing units, or LPUs, which are specialized chips designed to handle AI inference, the process of running trained AI models to generate responses. Unlike GPUs, which handle general-purpose computing tasks like model training, LPUs are optimized for the rapid generation of text tokens, the small units of language that AI models produce one at a time.

Nvidia announced the full-scale production milestone at the Hot Chips 2026 conference in California on Tuesday. The company positioned Groq 3 LPX as a complementary technology to its latest Vera Rubin platform, where GPUs handle training workloads and Groq chips take on inference-intensive tasks required by AI agents, the autonomous systems that can reason and act independently.

"Inference is the growth engine of AI. Vera Rubin extends that vision for the era of agentic AI, with LPX enabling ultrafast token generation and another leap in AI throughput, efficiency and responsiveness," said Jensen Huang, Nvidia's chief executive officer.

Jensen Huang, Chief Executive Officer at Nvidia

How Samsung's Foundry Could Return to Profitability?

• Advanced Node Production: Samsung is the sole manufacturer of the Groq 3 LPX's LPUs using its cutting-edge 4-nanometer process technology, providing meaningful volume for advanced-node production that has been a source of losses.
• Recent Price Increases: Samsung foundry recently raised prices by 15 percent, improving margins on new orders and helping offset years of competitive pressure from TSMC.
• Expanded Customer Base: Beyond Nvidia, Samsung has been widening foundry ties with major technology companies including Tesla, Apple, Google, and Broadcom, while potential cooperation with Anthropic on custom AI chips has also drawn market attention.

Analysts tracking Samsung's financial recovery see the Groq ramp-up as a critical inflection point. According to research from KB Securities, Samsung's foundry operation is expected to enter a meaningful turning point in the third quarter, with a return to profitability expected for the first time in four years since 2022.

"The Samsung foundry is expected to enter a meaningful inflection point in the third quarter. Excluding costs such as provisions for performance bonuses, a return to profitability is expected for the first time in four years since 2022, supported by the recent 15 percent price increase and the ramp-up of 4-nanometer LPU production," stated Kim Dong-won, head of research at KB Securities.

Kim Dong-won, Head of Research at KB Securities

Why Is Nvidia Betting So Heavily on Inference?

Nvidia's aggressive expansion into inference reflects a broader shift in the AI chip market. The company's biggest customers, including Amazon, Google, and major cloud providers, are increasingly developing their own custom silicon to reduce dependency on Nvidia's dominant GPU offerings. By building out a complete AI computing stack, Nvidia aims to maintain its market leadership even as customers diversify their chip suppliers.

Nvidia acquired Groq's intellectual property and key personnel in December 2025 in a deal valued at approximately $20 billion, bringing engineers who previously worked on Google's Tensor Processing Unit (TPU) technology into the company. This acquisition allowed Nvidia to rapidly scale inference capabilities without building the technology entirely from scratch.

The company is simultaneously expanding further up the AI stack. Nvidia announced it would invest about $7 billion in AI startup Poolside, securing technology licenses and bringing more than 100 of its employees into Nvidia to work on Nemotron 4, its upcoming trillion-parameter AI model slated for release in fall 2026.

What Does This Mean for the Broader AI Chip Market?

Samsung's potential return to profitability signals that there is genuine demand for alternative advanced chip manufacturing capacity beyond TSMC. With TSMC facing tight advanced-node capacity constraints, analysts see room for Samsung to capture additional orders from AI companies seeking alternative supply sources. This diversification could reshape the semiconductor supply chain for AI infrastructure over the next several years.

The timing is significant. Nvidia is scheduled to report its fiscal second-quarter earnings early Thursday Korea time, with investors watching the results for fresh signals on AI infrastructure spending and the outlook for the broader semiconductor sector. The Groq production ramp will likely feature prominently in those discussions as evidence that Nvidia's inference strategy is gaining traction with enterprise customers.