This is not investment advice. The author has no position in any of the stocks mentioned. Wccftech.com has a disclosure and ethics policy.
Intel Corporation has scored a win in the artificial intelligence space, with South Korea’s first online portal, Naver, switching to its central processing units (CPU) for some artificial intelligence workloads from graphics processing units (GPUs) offered by NVIDIA Corporation. Naver was previously an NVIDIA customer, but troubles with the supply chain and high prices have made the firm switch its fraud detection AI platform to Intel’s products. Naver’s switch marks the first stage of a shortage of NVIDIA’s products that has plagued analysts’ minds since the start of the AI boom earlier this year.
NVIDIA’s AI Accelerators Double In Price & Lead Times Stretch Out To A Year – Forcing Naver To Switch to Intel
When it comes to the three primary U.S.-based semiconductor designers, namely Intel, NVIDIA and AMD, Intel is not only the biggest company but also the one that is most plagued by troubles. Successive manufacturing timeline slippage and a slow personal computing market have stressed the firm’s income statement and affected its share price.
At the same time, advances in semiconductor manufacturing processes spearheaded by the Taiwan Semiconductor Manufacturing Company (TSMC) have enabled GPUs and alternatives to the x86 processor model to infiltrate the semiconductor market. NVIDA’s GPUs have risen to become the gold standard for AI applications. At the same time, firms such as Apple have shifted to Arm-based processors for personal computing applications traditionally performed by x86 chips.
Now, according to a report from The Korea Economy Daily, Naver has replaced NVIDIA’s GPUs with Intel’s CPUs for its Naver Place map platform. Naver had previously used NVIDIA’s GPUs to run an AI platform to separate fake listings from real ones in response to user queries.
High prices and long lead times have forced Naver to make the switch, according to industry sources quoted by the KED. These sources share that the price of an AI accelerator that uses NVIDIA’s chips has seen prices soar from an earlier 40 million Korean Won to 80 million Won (roughly equivalent to $59,193). To make matters worse, the lead time for these accelerators has jumped to 52 weeks.
In procurement and supply chain operations, lead time refers to the time it takes for an order to get fulfilled after it is placed, and longer lead times result in longer upgradation time for critical data centers and other information technology infrastructure equipment.
According to the sources, before switching to Intel, Naver tested the products for a month. It will also continue to use Intel’s Sapphire Rapids CPU to upgrade its location and search products computing infrastructure. A key bottleneck that has hampered NVIDIA’s AI accelerator shipments is TSMC’s packaging process. This is one of the final stages of product development, and it involves preparing the chips for final use.
Reports from the Taiwanese press suggest that TSMC has ramped up its orders for packaging equipment by as much as 30% in response to the high demand for NVIDIA’s products. However, this equipment has lead times of its own, which creates shortages in the industry as capacity takes time to come online.
Media reports have also claimed that TSMC’s packaging suppliers have entered a hot production run in response to its orders. These September reports forecasted AI chip prices and shared that some components are restricted by production capacity that should double soon. TSMC has announced a $2.89 billion investment into a chip packaging facility, and supply-demand mismatches can often evolve into a bigger problem in the chip sector as companies place more orders in response to shortages. This drives up prices, but as supply stabilizes, expensive inventory becomes more challenging to clear.