The surge of intelligent agents ignites CPU + GPU as AMD partners with TSMC to accelerate production! Striving to become the "computing foundation" of the AI era

According to Zhitong Finance APP, Lisa Su, the CEO of AMD (AMD.US), a leader in high-performance chips for PCs and data centers in the United States, stated on Friday that as demand far exceeds expectations, squeezing the global data center CPU and GPU markets, AMD is significantly increasing production capacity in collaboration with partners in Taiwan. AMD is actively upgrading itself from an "Nvidia GPU challenger" to a "computing power base supplier for AI data center CPU + GPU + high-performance network infrastructure + rack-level systems," driven by the ongoing demand for CPU + GPU brought about by the AI inference and Agentic AI frenzy, and will further deepen its integration with TSMC's cutting-edge processes and advanced packaging ecosystem in the future.

After visiting China, Lisa Su stated that she has met with AMD's largest customers in China and globally to ensure that its long-term partner, the chip manufacturing giant TSMC (TSM.US), known as the "king of global chip foundries," has the supply capacity to support a significant increase in the output of data center central processing units (CPUs) and AI GPU computing clusters.

AMD also announced a significant investment of "over $10 billion" in Taiwan's AI ecosystem to deepen strategic partnerships and enhance advanced AI chip manufacturing and server cluster assembly capacity. Lisa Su also mentioned that AMD is collaborating with partners like TSMC to accelerate capacity expansion to meet the unexpected demand for CPU + GPU driven by AI inference and Agentic AI.

From the perspective of AI data center engineering logic, Agentic AI does not rely solely on GPUs to train large models but requires CPUs to handle task scheduling, data movement, network stacks, storage access, security isolation, multi-agent orchestration, and inference service management. At this time, AMD's push for capacity expansion in Taiwan is not simply about "producing more chips," but ensuring that the supply chains for EPYC CPUs, Instinct GPUs, advanced packaging, substrates, assembly, and rack-scale systems are synchronized and scaled up.

Against the backdrop of a massive explosion in demand for data center CPUs, Wall Street financial giants are increasingly bullish on the two x86 architecture CPU super giants—Intel (INTC.US) and AMD (AMD.US)—as well as Arm Holdings (ARM.US), the owner of the ARM instruction set architecture. This year, the stock prices of these three CPU giants have all surged, with Intel's stock price increasing by as much as 222% since 2026. According to target prices compiled by TipRanks from Wall Street analysts, AMD's highest target price reaches $625—corresponding to about 40% potential upside. AMD's stock price has already risen by 110% this year, making it one of the "AI super bull stocks" in the global stock market.

From 2nm Venice to $10 billion investment, AMD strives to seize the high ground in AI computing power infrastructure capacity

Taiwan plays a key role in the global AI computing power infrastructure supply chain of North American tech giants, including Nvidia (NVDA.US) and Apple (AAPL.US), supported by TSMC, the world's largest contract chip manufacturer Taiwan Semiconductor Manufacturing Company (TSMC) is the core chip supplier for AMD, NVIDIA, and chip giants like Apple, Broadcom, and Qualcomm.

Dr. Lisa Su stated, "The overall demand in the CPU market is significantly higher than any of us predicted a year ago." She added, "I would say the CPU market is tight on supply." She mentioned that AMD is rapidly increasing production capacity and expects significant supply increases each quarter this year, with plans to provide significantly more supply in 2027 and beyond.

Dr. Su indicated that this strong demand growth is driven by AI inference and agent-based AI systems. As enterprises shift en masse to agent-based AI—AI systems that perform autonomous functions—CPUs have become the core focus, surpassing GPUs in popularity, and the demand for computing power has expanded significantly beyond GPUs used for training large AI models.

Dr. Su noted that China accounts for about 20% of AMD's revenue and added that China remains a very important market for the American chip design company. She stated, "Frankly, looking at the size of this market and the scale of our product portfolio, we will continue to work very closely with our Chinese customers."

She mentioned that AMD will continue to work closely with Chinese customers while complying with U.S. export controls; these export controls restrict the shipment of some of its high-end AI chips to the Chinese market.

On Thursday, AMD announced it would invest over $10 billion in Taiwan's AI sector to deepen strategic partnerships and accelerate the expansion of its advanced AI computing power cluster's capacity.

Dr. Su stated that this investment will focus on chip manufacturing, advanced packaging, substrates, and rack-level system manufacturing. She added, "Because some of these investments have a relatively long lead time, they must ensure land, buildings, and manufacturing capacity to complete these matters." She mentioned that AMD is co-investing with partners to ensure sufficient expansion capacity through 2026 and beyond, including up to 2029.

AMD's move is not merely an increase in TSMC's foundry orders but is linked to expanding strategic capacity based on TSMC's 2nm advanced process, integrating packaging, substrates, ODM/servers, and rack-level system ecosystems. This reflects that the competition for AI computing power has upgraded from individual GPUs/CPUs to a full-chain capacity competition of "advanced process + chiplet advanced packaging + packaging substrates + rack-scale AI systems."

AMD also announced on Thursday that it has begun using TSMC's 2nm process technology to ramp up production of its Venice CPU. According to AMD, the EPYC "Venice" CPU is the sixth generation EPYC data center CPU and the world's first 2nm process data center CPU, as well as the "world's first 2nm process HPC/data center CPU entering the ramp-up phase," marking the imminent large-scale commercial production of 2nm data center CPUs.

Dr. Su stated, "In fact, we made two bets. The first bet is on TSMC, which I think has proven to be a great bet for us." She added that AMD's second major bet is that increasingly complex silicon manufacturing technology will require chips to be split into smaller parts and integrated through advanced packaging technology, a method that is now widely adopted across the semiconductor industry—known as "chiplet advanced packaging technology." The Two Major x86 Chip Giants Join Forces to Rush into the Wild Bull Market?

According to Wall Street analysts, AMD is definitely not simply replicating NVIDIA's GPU dominance, but is instead attempting to reconstruct its valuation logic in the AI data center supply chain by focusing on the entire AI infrastructure ecosystem surrounding EPYC CPUs, Instinct GPUs, Helios/rack-scale platforms, 2nm Venice, and advanced packaging.

Wall Street analysts are expanding the narrative of AI computing power infrastructure from "GPU monopoly/single-core driven" to a full-stack computing revaluation of "AI GPU/ASIC + CPU + HBM/DRAM/NAND storage chips + optical interconnect-dominated high-speed connection systems in data centers."

Against the backdrop of a massive explosion in demand for data center CPUs, Wall Street financial giant Citigroup significantly raised its 12-month target prices for the two major x86 architecture CPU super giants—Intel (INTC.US) and AMD (AMD.US) this week. This financial giant also substantially upgraded its expectations for the data center CPU and overall CPU market size in its latest research report.

At the 54th Global Technology, Media, and Telecommunications Conference hosted by Wall Street financial giant JP Morgan, Intel CEO Pat Gelsinger stated that Intel 18A (the advanced chip process below 2nm, specifically at the 1.8nm level) has supported the mass production of Panther Lake, with yield rates improving by about 7% each month, exceeding Intel's internal expectations. Gelsinger also mentioned that as the focus of AI computing power infrastructure shifts from training to inference, CPUs are becoming increasingly important and indispensable in the AI era, with the configuration ratio of CPUs to GPUs accelerating from 1:8 towards 1:1, and potentially reaching 4:1. Additionally, Intel's business planning indicates that it is actively pursuing ASIC business, offering customized AI CPU or AI GPU chip solutions.

The shift of AI from training to inference and Agentic AI will significantly increase the strategic weight of CPUs in AI data centers. GPUs are responsible for large-scale matrix calculations, but CPUs handle scheduling, I/O, memory management, task orchestration, security isolation, database access, network stacks, and multi-agent workflow execution; as AI applications transition from "single inference" to "continuously running software labor," the demand for CPUs will rise from traditional server upgrade cycles to AI infrastructure expansion cycles. Citigroup's latest model echoes this point: it predicts that the total potential market size for data center server CPUs (CPU TAM) will expand from $29.3 billion in 2025 to $131.5 billion in 2030, with a compound annual growth rate of approximately 35%.

TipRanks' compilation of Wall Street analysts' target prices shows an average target price of about $460, with the highest target price for AMD reaching as high as $625—corresponding to about 40% potential upside. So far this year, AMD's stock price has surged by 110%, making it one of the "AI super bull stocks" in the global stock market. The highest target price of $625 from Wall Street comes from Baird analyst Tristan Gerra, who recently raised AMD's target price from $300 to $625 and maintained an "Outperform" rating The rating's bullish logic is primarily based on the unprecedented increase in CPU/GPU demand driven by Agentic AI, as well as the revaluation of AMD's data center platform.

Analyst Tristan Gerra stated that AMD's opportunities in AI server CPUs, GPUs, and accelerated computing platforms have been repriced, especially under the joint drive of the AI inference wave, Agentic AI, EPYC data center CPUs, Instinct GPUs, and the expansion of next-generation AI infrastructure. The market is beginning to view AMD as a key beneficiary of the "CPU + GPU + super-platform solutions" among the leaders in the AI computing industry chain