Asmack writes the boom in AI computing power into the 2028 production expansion table! Wall Street is betting that AMD (AMD.US), which holds the CPU+GPU, dominates the AI inference era

The Zhitong Finance App learned that Wall Street's recent bullish sentiment towards Nvidia's biggest competitor of AI GPUs, AMD (AMD.US), the leader in high-performance chips in US PCs and data centers, is getting stronger. Financial giants such as Bank of America, Goldman Sachs, and Barclays have recently upgraded their ratings or bullish margins for AMD. Wall Street financial giant Goldman Sachs recently raised the target price of Chaowei Semiconductor (AMD) from $450 to $640, while maintaining the most optimistic “buy” rating for the stock. Another agency, Cantor Fitzgerald, gave $700 (a huge increase from the previous $500). The most optimistic target price came from KeyBanc—the agency raised AMD's target price from $530 to $725 and maintained an “gain” rating.

Wall Street analysts' core bullish logic is not simply “AI GPU market share catching up with Nvidia,” but the proxy AI wave (that is, the big wave of AI agents) is driving the full explosion of AI CPU demand, and the ratio between CPU and GPU narrows. With data center server-level CPUs, AI accelerators, Helios rack server clusters, and cloud customer base, AMD is expected to become one of the most important computing power beneficiaries in AI computing power infrastructure other than Nvidia.

Wall Street's bullish logic can be described as having expanded from the early “AMD is the second AI GPU training/inference accelerator supplier outside of Nvidia” to resonating with the three supergrowth curves of server CPUs, AI GPUs, and large-scale rack-scale server clusters. Analysts are betting that AMD, which has CPU+GPU production capacity, will co-dominate the pricing power of the computing power infrastructure in the AI inference era.

According to information, AMD and RackSpace Technology recently signed an important final agreement to deploy the first 30 megawatts of artificial intelligence computing infrastructure based on AMD technology in RackSpace's data centers around the world. This legally binding contract formally implements the preliminary memorandum of understanding announced by the two companies in May. The relevant artificial intelligence computing power is scheduled to be put into operation in stages from the end of 2026 to 2028. The two companies said that this deployment will enable the workloads of regulated enterprises to use the computing power infrastructure built by AMD Instinct AI GPUs and AMD EPYC central processing units (CPUs) in the RackSpace enterprise artificial intelligence cloud architecture and AI agent deployment workflows.

Gayan Kandia, CEO of Rackspace Technology, said the partnership integrates AI computing infrastructure with a regulated operating model and provides services by a partner with unified responsibility. Dan McNamara, senior vice president and general manager of AMD Computing and Enterprise Artificial Intelligence Business, said the partnership will help regulated companies deploy scalable artificial intelligence computing power infrastructure. The two companies also jointly committed to investing in joint sales and marketing resources to reach corporate customers in regulated industries.

At a time when AI agents are popular around the world, the main line of AI computing power investment is shifting from a “single-point computing power competition around GPUs” to a “full-stack computing power system driven by AI agents”. The next round of excess alpha earnings will no longer only belong to the list of the strongest leaders in the AI GPU/AI ASIC field, but will also spread systematically to data center high-performance CPUs, DRAM/NAND/HBM storage, AI PCBs, liquid cooling systems, data center optical interconnect systems, ABF carriers/glass substrates, MLCCs, electronic cloth, and extensive foundry of wafers AI computing power infrastructure layer, and in this AI mainline narrative shift, data center CPUs, optical interconnects, and memory chips are probably the biggest winners.

From lithography machines and HBM to server CPUs, the computing power supercycle ushered in a full chain of testing

The outlook for global AI computing power demand can be described as becoming more optimistic, rather than moving towards “excess computing power” as some pessimists might expect. In particular, the latest results announced by lithography giant ASML.US (ASML.US) on Wednesday are more forward-looking industrial evidence than orders from a single AI chip company.

According to financial data, Asmack's second-quarter revenue reached 9.326 billion euros, higher than market expectations of 8.8 billion euros; net profit of 2,918 billion euros also exceeded expectations of 2.62 billion euros, and gross margin reached 54%. More importantly, the company drastically raised the 2026 revenue guide from 36 billion to 40 billion euros to 43 billion to 45 billion euros, raised the gross margin guide from 51% to 54% to 56%, and expected a further increase in revenue to 11 billion to 12 billion euros in the third quarter. Asmack's management said that orders in the first half of the year were “extremely strong,” and customers are speeding up production capacity for advanced process logic chips and memory chips. Therefore, they plan to increase the production capacity of both low-value aperture extreme UV lithographers and immersion deep UV lithographers by about 30% in 2027, and study an increase of about 30% in 2028. This means that major fabric-type customers such as TSMC and Intel have been voting for strong AI computing power demand from 2027 to 2028 using lithography equipment promises rather than verbal predictions.

The performance warning from the established US tech giant IBM (IBM.US) proved from the demand side that the AI computing power infrastructure procurement frenzy is spreading from hyperscale cloud vendors to traditional enterprise business models. IBM revealed that in order to avoid continued shortages and price increases in key AI computing power infrastructure such as AI chips, server CPUs, storage and memory chips, customers switched their quarterly capital budget from software projects to core hardware procurement related to AI infrastructure. As a result, its revenue was only 17.2 billion US dollars, lower than market expectations of 17.86 billion US dollars, and the stock price plummeted by about 25% in a single day.

IBM's own large contract extensions and sales execution errors are also important reasons for the performance gap, so the whole problem cannot be attributed to industry trends; however, enterprise customers have begun to lock down infrastructure ahead of time, indicating that AI computing power demand is no longer just a huge AI capital expenditure story of Microsoft, Meta, and Google, but is changing the top priority level of the IT budget of ordinary enterprises.

Goldman Sachs drastically raised AMD's target price from $450 to $640. The core is not only betting on its Instinct AI GPU market share to speed up with Nvidia in the AI chip field, but believes that the big wave of AI agents will significantly increase the server central processor load required for data retrieval, task scheduling, model orchestration, network and storage management, thus making the EPYC CPU cluster the second-core growth engine underestimated by the market.

Asmack's rare drastic expansion of production means that advanced logic chip manufacturers such as TSMC and Intel are preparing larger wafer production capacity for customized AI ASIC accelerators such as larger data center server-level CPUs, GPUs, and TPUs; while Samsung's latest performance and the continued severe shortage of memory chips jointly released by SK Hynix CEO on the first day of the US stock ADR listing also clearly demonstrates from another core link in the AI computing power industry chain that demand for complete AI servers is still extremely strong.

Wall Street is also bullish on memory chips. In terms of SK Hynix's US stock ADR, which hit the US stock market last week, Wall Street financial giant Barclays gave it a target price of up to 330 US dollars for American Depositary Receipts (that is, US stock ADR), which implied about 117% upward space compared to the closing price before it was covered. Barclays said that industry representatives generally expect DRAM manufacturers to only meet the current demand of about 75% to 80%. The 2027 satisfaction rate may even drop to about 60%, and long-term supply agreements are expected to increase the visible profit margin for the next two to three years.

Asmack, IBM, and SK Hynix each jointly proved in the three directions of upstream equipment orders, downstream enterprise procurement, and key memory chip supply: the AI computing power super demand cycle did not peak due to large-scale pullbacks in technology stocks or semiconductor stocks, and AI computing power bottlenecks began to gradually migrate from GPU/TPU/AI ASICs to key parts of the AI computing power industry chain such as CPUs, HBM, server memory and NAND storage components, extensive wafer manufacturing, and even semiconductor devices such as lithography machines. As a result, AMD is expected to be re-evaluated from an “Nvidia replacement” to the second-largest full-stack computing power platform spanning EPYC central processing units, Instinct accelerators, and rack-level systems. However, whether the benefits can be realized still depends on mass production of advanced products, software ecology, TSMC wafer distribution, HBM supply, and gross profit margin; in addition, the severe shortage of HBM/DRAM/NAND memory chips not only supports terminal demand, but may also limit AMD's short-term shipments.

Are the two major x86 chip giants joining hands to go to the Wild Bison Market?

According to Wall Street analysts, AMD is definitely not simply replicating Nvidia's GPU hegemony, but is moving from the entire AI infrastructure ecosystem around EPYC CPUs, Instinct GPUs, Helios/Rack-scale platforms, 2nm Venice, and advanced packaging to try to restructure its valuation logic in this AI data center industry chain in the era of AI inference and Agentic AI.

Wall Street analysts are expanding the AI computing power infrastructure narrative from “GPU dominance/single-core driver” to “AI GPU/ASIC+CPU+HBM/DRAM/NAND memory chip+data center high-speed connection system collaboration dominated by optical interconnection” to a full-stack computing power reassessment.

According to Goldman Sachs, the AI computing power super bull market is far from over. Instead, it has moved from the “AI chip purchase frenzy” to the second stage of “large-scale AI factory construction” — that is, the next round of excess alpha revenue will no longer only belong to the list of the strongest leaders in the AI GPU/AI ASIC field, but will also spread systematically to data center high-performance CPUs, DRAM/NAND/HBM storage, AI PCBs, liquid cooling systems, data center optical interconnection systems, ABF carriers/glass substrates, MLCC, electronic distribution, and extensive foundry of “AI factories” full-stack AI computing Power infrastructure layer.

Brian Nowak, a senior analyst from Wall Street financial giant Morgan Stanley, led the analysis team and released the latest research report on July 12, once again significantly raising the 2027/2028 capital expenditure forecasts for the five largest hyperscale cloud computing and technology giants (Meta, Amazon, Microsoft, Google, SpaceX), which have reached approximately $1.2 trillion and $1.4 trillion, respectively. The agency's capital expenditure forecast for major US tech giants in 2026 was drastically raised from 433 billion US dollars a year ago to 805 billion US dollars.

Wall Street's bullish sentiment is not only growing stronger about AMD, but also about Intel, another x86 CPU giant. Against the backdrop of the explosion in data center CPU demand, several Wall Street financial giants have recently drastically raised their target stock prices for the two major x86 CPU supergiants, Intel (INTC.US) and AMD (AMD.US). These financial giants also drastically raised their expectations for data center CPUs and the overall CPU market size in the newly released research report.

At the 54th Global Technology, Media, and Communications Annual Conference hosted by Wall Street financial giant J.P. Morgan Chase, Intel CEO Chen Liwu said that Intel 18A (that is, 1.8nm advanced chip manufacturing process below 2nm) already supports mass production of Panther Lake, increasing yield by about 7% every month, exceeding Intel's internal expectations. Chen Liwu also said that as the focus of AI computing power infrastructure gradually shifts from training to reasoning, CPUs are becoming more and more important and indispensable in the AI era, and the CPU to GPU configuration ratio accelerates from 1:8 to 1:1, and can even reach 4:1.

Notably, Intel's investment logic is fundamentally changing: the market no longer sees it as just a traditional consumer electronics central processor manufacturer waiting for the PC cycle to recover, but is beginning to reprice the full-stack AI computing power infrastructure platform based on “data center server CPU+ advanced process chip manufacturing/foundry/advanced packaging”. This is why international financial giant HSBC (HSBC) raised the agency's price target for Intel by 100% to $200 — this is also the highest target point for Intel among Wall Street analysts. Intel can once again be described as one of the most popular semiconductor stocks among retail and institutional investors around the world.

Based on Intel's stock price of about $107.76 and a market value of about US$547.7 billion on July 15, the target price of 200 US dollars means a potential increase of about 85.6%; under the assumption that the share capital in circulation remains largely unchanged, the corresponding market value is about 1.02 trillion US dollars, that is, according to HSBC, Intel is expected to re-enter the “trillion-dollar chip giant.”

At a time when AI agents (or Agentic AI) are rapidly becoming popular around the world, cloud computing giants and AI leaders are rapidly expanding their artificial intelligence (AI) infrastructure spending, and global AI data center computing power infrastructure is in full swing, HSBC made this aggressive bullish judgment. Although Nvidia (NVDA.US) has always dominated the AI GPU computing power infrastructure market, CPU, advanced packaging, and wafer/semiconductor manufacturing capacity expenses are becoming an increasingly critical part of the AI computing power supply chain system. HSBC believes Intel can benefit from the continued surge in data center CPUs brought about by AI agents and the global AI semiconductor production capacity expansion frenzy led by Musk TeraFab's “superchip factory.”