Samsung and SK Hynix Valuations Set to Rise as Nvidia’s Monopoly Cracks in the AI Chip Market [Chip Talk]
OpenAI Unveils Its Own 'Jalapeno' Chip
Amazon and Google Join In... Nvidia Dependency Drops
Memory Sector Gains Bargaining Power as Client Base Grows
"Long-Term Production Strategies Must Be Refined"
Following in the footsteps of Google and Amazon, OpenAI has also joined the race to develop its own artificial intelligence (AI) semiconductors, accelerating the global big tech companies' move to reduce their reliance on Nvidia. Analysts suggest that this shift could present new opportunities for domestic memory and foundry (semiconductor contract manufacturing) companies.
Sam Altman, CEO of OpenAI, is speaking at a press conference held at The Plaza Hotel Seoul in Jung-gu, Seoul, on the morning of February 4 last year. Photo by Yonhap News
View original imageOpenAI Completes AI Chip in Just 9 Months
On June 24 (local time), OpenAI unveiled its self-developed chip, "Jalapeno," which is specialized for AI model inference. Developed jointly with global semiconductor design company Broadcom since last year, the chip is scheduled for full-scale deployment in actual data centers by the end of this year.
Hock Tan, CEO of Broadcom, emphasized the chip's performance, stating, "It is on par with Nvidia's Blackwell chip and Google's Tensor Processing Unit (TPU)." According to preliminary test results, Jalapeno demonstrated superior performance per watt compared to current state-of-the-art technologies. The company explained that it reduced data movement—a bottleneck in existing AI chips—while balancing compute, memory, and networking resources to achieve near-maximum utilization.
What stands out in particular is the development speed. It took only nine months for Jalapeno to go from initial design to the "tape-out" stage, where the blueprints are handed over to the foundry for manufacturing. Both companies highlighted this as "the fastest application-specific integrated circuit (ASIC) development cycle ever," attributing it to the use of OpenAI's own AI models during the design and optimization process.
OpenAI has also confirmed a concrete roadmap, planning to release the next-generation chip in 2028 and to introduce a new chip every year thereafter. While Jalapeno was developed with a focus on inference, the company indicated that future chips may expand into other areas such as training.
'De-Nvidia' Acceleration Triggers Seismic Shift
This trend of developing proprietary chips is spreading beyond OpenAI to big tech as a whole. Notably, Amazon has achieved success with its self-developed AI accelerator "Trainium." Since its launch in 2020, Trainium has been adopted by companies such as OpenAI and Anthropic, helping Amazon secure $225 billion (about KRW 342 trillion) in revenue commitments as of April this year. Recently, it has been reported that Amazon is negotiating to sell Trainium chips directly to external company data centers, bypassing Amazon Web Services (AWS). Previously, Trainium chips could only be used within AWS, but now the strategy is to boost market share by selling directly, similar to Nvidia's approach.
Alphabet, the parent company of Google, also announced in April that it would begin supplying its proprietary AI semiconductor, TPU, directly to select customers, formalizing its external sales plan. Anthropic is also reportedly considering developing its own chips. As Nvidia's key big tech customers pursue proprietary chip development, there is growing analysis that Nvidia's dominant position in the AI chip market is being challenged.
‘Good News’ for Korean Firms, ‘Production Strategy’ Still a Challenge
The semiconductor industry views this trend toward "chip internalization" by big tech companies as likely to have a positive impact on domestic memory companies such as SK hynix and Samsung Electronics. The more companies attempt to develop their own chips, the greater the demand for essential components like DRAM and high bandwidth memory (HBM). An industry insider explained, "As demand, which was previously limited to Nvidia GPUs, diversifies to include chips from OpenAI, Amazon, and Google, supply channels are expanding. With limited supply capacity, an increase in customers will allow companies to charge a premium for their products, thereby solidifying their profit structure."
From the foundry perspective, this trend could also present new opportunities. Since various AI chips from Amazon, Google, and OpenAI will all require contract manufacturing, there is a chance to win new orders. Yongseok Kim, Distinguished Professor at Gachon University's College of Semiconductor, stated, "While TSMC is likely to maintain its lead, Samsung Electronics could emerge as a differentiated alternative if it can sufficiently improve the yield of its advanced processes."
However, the ability to forecast demand and optimize production portfolios by process will remain a key challenge for semiconductor companies. Since memory chip production capacity is limited, how resources are allocated between different products could significantly affect profitability over the next two to three years.
Hot Picks Today
"Two Spend 500,000 Won in a Flash, Yet Wait Two Hours?" ... 'Dreamland' Shunned by Kids and Foreign Tourists
- The Era of Getting 200 Million Won After 30 Years Is Over... A Japanese Bank's "Perpetual Department Head" Who Just Clocked In Cries Out
- Spain Hit by Its Worst Tragedy... Lee Kangin, Moreno and Real Madrid Join Soccer World in Mourning
- They Said, “Make It Your Home for Life” - Then Suddenly Kicked 200 Passengers Off... What Happened to the Around-the-World Cruise?
- "130 Rice Bowls Sold Out in a Day"... 'Unfiltered Face' Adds 8 Million Won to Night Market Sales: China's Bungeunwoo
An industry insider cautioned, "If the market is misjudged and the share of a particular product is increased excessively, it could result in missing out on the supply opportunity for another higher-margin product, which would have adverse effects. Rather than focusing on short-term gains, companies must accurately read mid- to long-term market demand and optimize product production ratios—this production strategy will ultimately determine future performance."
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.