GPU prices skyrocketed by 48% in a single month! Intel has completely bid farewell to being the AI supporting role and is embracing the inference era with a combination of CPU and GPU demand surge + chip foundry + advanced packaging.

According to market research data, the price of the Arc Pro B70 workstation GPU developed by Intel Corporation (INTC.US) has increased by as much as 48% within a month. The well-known Wall Street investment firm Wedbush Securities commented that this largely reflects the ongoing rise in storage chip prices, and highlights Intel Corporation's transition from a supporting role in AI to becoming a core player in the AI inference era with its CPU + GPU product array. Equipped with 32GB of GDDR6 ECC memory, the Arc Pro B70 is becoming a significant selling point for Intel Corporation in the AI inference era. Its large memory capacity and error correction capabilities make it particularly suitable for AI inference workloads, especially for enterprise-level AI inference tasks that demand long-term stability and high data accuracy. As the demand for large parameter inference models continues to increase in terms of memory capacity and bandwidth, Intel Corporation's workstation GPUs, along with Crescent Island, are becoming essential computing options outside of AI data center-level GPU/TPU accelerator card clusters. Here, error correction capability (ECC) refers to the memory's ability to automatically detect and correct single-bit errors that may occur during data transmission or storage, thereby minimizing computational anomalies, result distortions, or system crashes during model inference. Intel Corporation is rapidly transitioning from being an "AI supporting role" in the previous AI large model training era to a multi-tier beneficiary of AI computing infrastructure led by Agentic AI in the AI inference era. The "CPU and GPU demand surge + chip foundry + advanced packaging" paradigm is reshaping Intel Corporation's growth trajectory. Not only has the increasing demand from data centers driven a substantial rise in sales of its x86 architecture central processing units (i.e., data center CPUs), but the 32GB of large memory also enables it to handle local large model inference, small model fine-tuning, and professional workstation tasks. Compared to expensive NVIDIA Corporation professional cards, it offers a capacity-price advantage, resulting in a significant upsurge in workstation GPU demand; however, this is mainly due to the scarcity premium of high-memory workstation inference cards, which does not indicate that the Arc ecosystem has suddenly surpassed CUDA, nor does it mean that Intel Corporation is starting to capture data center GPU market share. The 32GB ECC memory is riding the wave of AI inference! The price of Intel Corporation's Arc Pro B70 surged by 48% in a month. Matt Bryson, a senior analyst with Wedbush Securities, wrote in a report to clients: "According to research data reported by Toms Hardware, the price of the Intel Corporation Arc Pro B70 workstation GPU has increased significantly by 30% in the U.S. market. The ASRock Creator variant is currently priced at $1,299, up from $999 last month, while its suggested retail price is $949. The Arc Pro B70 workstation GPU has also increased by 26% in the German market; in South Korea, it has surged by 48%, from 1,889,980 KRW (approximately $1,334) on July 21 to 2,798,980 KRW (approximately $1,975) on August 16." "Intel Corporation's official price in South Korea is currently 2,817,000 KRW. This graphics card, equipped with 32GB GDDR6 ECC memory, is particularly attractive for AI-related inference workloads. We believe this news once again reflects the intensified transmission of storage chip costs to end prices, as well as the beginning shift of Intel Corporation's GPU + CPU product line into an increasingly important role in the AI inference era," Bryson added in his report to clients. The sudden price increase of the Arc Pro B70 should not be simply interpreted as an explosive demand for Intel Corporation GPUs. The core reason, as noted by Wedbush, is the transmission of storage chip costs. The B70's 32GB GDDR6 ECC memory represents a significantly higher material cost (BOM) weight compared to ordinary consumer graphics cards; as production capacity shifts to HBM, server DDR5, and high-end storage, the supply cost of GDDR6 has risen, and tight channel inventory has further amplified regional premiums. At the same time, the large 32GB memory allows it to accommodate local large model inference, small model fine-tuning workloads, and professional workstation tasks, providing a "capacity-price advantage" compared to expensive NVIDIA Corporation professional cards and indeed resulting in incremental demand; however, this is primarily due to the scarcity premium of high memory workstation inference cards, not an indicator that the Arc ecosystem has suddenly surpassed CUDA, nor does it mean that Intel Corporation is desperately trying to capture data center GPU market share. The rise in memory prices has ignited the trend towards a 1:1 ratio of Arc Pro, CPU, and GPU, with Intel Corporation transitioning from an AI supporting role to becoming a core beneficiary of AI computing demand. As described above, Intel Corporation is upgrading from being an "AI supporting role" during the training era to becoming a multi-tier beneficiary of the AI computing infrastructure in the inference age. However, it is still premature to say that it will become the dominant player in the inference era. Agentic AI's technology-driven workflow requires CPUs to handle control plane tasks such as task planning, retrieval-augmented generation (RAG), vector database queries, API calls, state management, and multi-agent orchestration, while GPUs continue to manage matrix calculations and high-throughput reasoning on the data plane. Therefore, Bank of America has raised its potential market size for server CPUs by 2030 to over $210 billion, with an annual compound growth rate increased to 36%. It also predicts that the CPU to GPU ratio will narrow from about 1:4 during the training phase to about 1:2 in the inference phase, and ultimately approach 1:1; this will significantly amplify the value of Intel Corporation's Xeon data center CPU installation base. The potential to significantly reshape Intel Corporation's growth trajectory lies in the four-layer combination of "CPU + inference GPU + wafer foundry + advanced packaging," which can be considered a composite repair asset in the AI agent-led inference power expansionthis means that the rising demand density for Xeon Intel Corporation data center CPUs provides a performance foundation, exclusive GPUs with high memory types open differentiated inference channels, and the 18A and advanced packaging provide valuation options. For Intel Corporation's new fundamental growth trajectory, Xeon will support AI orchestration and independent CPU rack demands; Gaudi 3 and Crescent Island with 160GB LPDDR5X target cost-sensitive, air-cooled enterprise inference markets, and Arc Pro B70 aims at local large model inference, small model fine-tuning, and professional workstation tasks; the 18A advanced process technology will support future server CPU iterations; EMIB bridging and Foveros Direct 3D hybrid bonding can cater to third-party AI accelerators and large-scale chiplet systems. In the second quarter of 2026, Intel Corporation's data center and AI-related revenue reached $6.3 billion, a year-on-year growth of 59%, and the wafer foundry business revenue reached $5.8 billion, a year-on-year growth of 31%. This sufficiently indicates that the strong demand for AI server computing clusters has already begun to enter its reports; however, Intel Corporation's 18A/14A chip foundry and advanced packaging revenues currently include a large amount of internal settlements. Only the mass production, yield improvement, and large-scale conversion of EMIB/Foveros 3D advanced packaging orders for external customers can prove that it has formed an independent profit engine. As of 2026, Intel Corporation's stock price has increased significantly by 150%. Looking at the stock price prospects for Intel Corporation, the most optimistic target price on Wall Street is $200, issued by HSBC analyst Frank Lee, suggesting a potential upside of approximately 115.5% from Intel Corporation's closing price of $92.80 on August 19; this analyst has doubled the target price from $100 and maintained a "buy" rating, with the core logic being that this institution has included Intel Corporation's foundry business (Intel Foundry) in its valuation for the first time and believes that the 18A/14A advanced processes, EMIB/Foveros advanced packaging, and potential external customers will position Intel Corporation as an important alternative choice apart from Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, while the continuing expansion of data center CPU and AI accelerator demand strengthens profit recovery.