Nvidia’s Groq 3 LPX chip pushes Samsung foundry closer to profitability

Seoul [South Korea], August 26 (ANI): Nvidia's Groq chip ramp brings Samsung foundry closer to profit, as the American semiconductor firm initiated mass production of its Groq 3 LPX inference accelerator with Samsung Electronics manufacturing the critical processors, according to a news report by the Korea Herald.
The development raised market expectations that the South Korean technology giant's loss-making contract manufacturing unit could stage a financial turnaround as early as this year.
As per the news release, Nvidia announced at the Hot Chips 2026 conference in California that the Groq 3 LPX entered full-scale production, marking the company's push beyond graphics processing units into an integrated computing stack covering training, inference, systems, and software.
The accelerator is a rack-scale system containing 256 language processing units tailored for artificial intelligence inference. The hardware operates alongside Nvidia's Vera Rubin platform, allowing graphics processing units to handle training workloads while Groq silicon manages inference tasks for AI agents.
"Inference is the growth engine of AI," the news report quoted Nvidia CEO Jensen Huang. "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."
Samsung operates as the exclusive manufacturer for the Groq 3 LPX processors using its 4-nanometer fabrication line. The contract provides the company with substantial advanced-node production volume as it attempts to reverse operational losses. Huang previously confirmed at the GTC conference in March that Samsung was fabricating the hardware and that Nvidia was expanding output "as fast as we possibly can."
The production surge fueled forecasts of an operational rebound for Samsung during the second half of the year.
"The Samsung foundry is expected to enter a meaningful inflexion point in the third quarter," the news report quoted Kim Dong-won, head of research at KB Securities.
"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 per cent price increase and the ramp-up of 4-nanometer LPU production," Kim said.
Samsung widened business links with key global firms including Tesla, Apple, Google, Broadcom, and Anthropic. With market leader TSMC facing tight manufacturing capacity, Samsung positioned itself to secure contracts from enterprises seeking alternative advanced-node supply.
As per the news release, the expansion comes as major cloud providers develop proprietary chips. Amazon deployed its Trainium silicon, while Google prepared to offer its tensor processing units externally this year. Nvidia countered these shifts last December by acquiring Groq's intellectual property and hiring its core engineering team in a 20 billion-dollar agreement.
Nvidia also committed approximately 7 billion dollars to AI startup Poolside to integrate over 100 researchers for its upcoming Nemotron 4 model ahead of its fiscal second-quarter financial results. (ANI)