The Zhitong Finance App learned that the latest research report released by Wedbush Securities, a well-known investment agency on Wall Street, shows that the photolithography giant ASML.US (ASML.US) exceeded expectations and raised its 2027 performance guidelines, which is very beneficial to the bullish outlook for AI computing power infrastructure related to the most advanced process logic chips such as CPU/GPUs and dynamic random access memory (that is, DRAM memory chips). The latest views of Wall Street institutions such as Wedbush, as well as the strong performance and optimistic outlook for the future of Asmack and TSMC, have undoubtedly provided a multi-needle counterattack on the AI computing power theme, which has fallen into an extremely sharp sell-off due to overcrowded and highly leveraged long positions.
From Wedbush's optimistic rhetoric on the topic of AI computing power, to the recently announced explosion performance and strong performance outlook for the two most important upstream AI computing power industry chains, Asmack and TSMC, to the latest analysis by research firm SemiAnalysis showing that ANTHROPIC is moving from long-term losses to a stage of rapid overall profit growth, it is actually sending an important signal to the global stock market: the AI computing power industry chain has gradually entered the “AI capital expenditure supercycle for training AI models” into a “large-scale application of smart devices” Driven by the “exponential expansion of demand for AI inference”, these latest signals can be described as hitting the pessimistic rhetoric of “excess computing power” that has led to a recent sharp decline in AI computing power topics, especially the AI semiconductor sector.
Wedbush's core judgment is that Asmack raised its 2027 outlook and is considering continuing to increase EUV production capacity in 2028, which is fully consistent with the nearly endless demand for advanced process logic chips (that is, logic chips such as CPU/GPU/TPU below the 5nm process) and high-end data center DRAM to support the AI data center construction process; the increase in Asmack's lithography equipment shipments in 2027 is expected to be transformed into initial wafer manufacturing output from the end of 2027 to the beginning of 2028, especially high-end DRAM/HBM memory chip production capacity, but the volume is Continued growth in AI-related capital expenditure by chip manufacturing giants such as electronics and global tech giants such as Microsoft and Meta still means that it is uncertain when supply will actually catch up with demand.
Matt Bryson, a senior analyst at Wedbush Securities, wrote in his latest report to clients: “We believe that Asmack's management unexpectedly raised its outlook for 2027 and the possibility of a significant increase in EUV lithography equipment production in 2028, which is almost in line with the pace we are seeing growing demand for advanced process logic chips and high-end DRAM for data centers. These requirements are designed to support the ongoing global AI computing power infrastructure construction process.”
“The timeline for the increase in shipments is also very much in line with our benchmark judgment, that the increase in capital expenditure related to AI computing power in 2027 will eventually drive an increase in advanced process logic chips and DRAM production from the end of 2027 to the beginning of 2028, especially the production of HBM/DRAM memory chips; however, given that capital expenditure continues to grow, we are still uncertain when supply will eventually catch up with the expanding demand curve.” A team of Wedbush stock market analysts led by Bryson said.
In addition to the 2027 performance guidelines, Asmack's management currently anticipates that total net sales (that is, total revenue expectations) for the full year of 2026 will be between 43 billion euros and 45 billion euros, far higher than the previous forecast range of 36 billion euros to 40 billion euros, and higher than the average Wall Street analysts' expectations of 39.3 billion euros.
From EUV to 5GW Super Clusters: Quadruple Industrial Chain Signals Break Down the “AI Overpower Theory”
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.
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.
More importantly, Asmack's management plans to increase the annual production capacity of about 65 low-numerical aperture EUV units by 30% in 2027, and study an additional 30% increase in 2028. The DUV immersion lithography machine is also adopting a similar production expansion path. This is not a short-term inventory replenishment; rather, global fabs lock in the upstream capabilities of advanced logic, HBM/DRAM, and advanced packaging with multi-year equipment commitments.
According to AI Industry News this week, Facebook parent company Meta expanded the investment and construction scale of its Louisiana Hyperion data center park from the initial announcement of 10 billion US dollars to a supercluster with over 50 billion US dollars and more than 5 GW of computing power, and 27,500 Rubins ignited the “Japan Robotics National Team”. In addition, Nvidia CEO Huang Renxun went to Japan this time to cooperate with established industrial giants such as Fanuc, Yaskawa Electric, Kawasaki Heavy Industries, and Fujitsu to promote physical AI collaboration, and will cooperate with established industrial giants such as SoftBank, NEC, Hitachi, Sony, Preferred Networks and others are incorporated into the Cosmos ecosystem, superimposed by Nvidia's exclusive AI chip manufacturer -- the recently announced strong performance exceeding expectations and the increasingly optimistic future outlook for AI computing power demand, which is sufficient to prove that the global AI computing power investment cycle is far from over. The expansion of computing power is now in a single arms race led by large-scale US cloud vendors, and upgraded to “cloud AI, sovereign AI, and physical AI” with multi-center resonance.
TSMC's performance and outlook have proven from the wafer manufacturing side that this demand is not limited to equipment orders: second-quarter US dollar revenue of $40.2 billion, up 33.7% year on year; net profit reached NT$706.56 billion, up 77.4% year on year, significantly higher than market expectations of NT$632.6 billion; high performance computing business accounts for 66% of revenue, and advanced manufacturing processes of 7 nm and below account for 77% of wafer business revenue.
More importantly, TSMC's revenue guidance for the third quarter was further raised to 44.6 billion to 45.8 billion US dollars. At the same time, the company drastically raised its 2026 capital expenditure from 52 billion to 56 billion US dollars to 60 billion to 64 billion US dollars, raised the annual US dollar revenue growth guide to slightly more than 40%, and added 100 billion US dollars in US investment, bringing the total commitment in the US to about 265 billion US dollars. As the final production capacity of core AI chip customers such as Nvidia and the manufacturing chain undertaker to meet AI computing power requirements, TSMC has shown through profit, utilization, and capital expenditure that the visibility of AI computing power infrastructure demand around advanced process logic chips, cutting-edge 2 nm manufacturing processes, and advanced packaging is still rising, rather than global AI computing power demand falling into a period of weakness.
Asmack and TSMC did not mechanically accept customer predictions, but only approved the expansion of production after reviewing data center construction and terminal requirements. Therefore, this latest set of guidelines strongly disproves the assertion that “there is currently a systemic excess of computing power.” The demand side also continues to expand: Noetra, supported by the Japanese government, plans to purchase 27,500 Nvidia Rubin chips to build physical AI infrastructure. The project is scheduled to start construction in April 2027 and put into operation in June 2028; Meta is expanding the Hyperion Park in Louisiana from a $10 billion project initially announced in 2024 to a supercluster of over 50 billion US dollars and 5 gigawatts of computing power. This quadruple industrial chain signal will undoubtedly break through the “AI excess computing power theory.”
Anthropic is expected to achieve a major shift in profit trajectory, and the AI computing power bull market will enter the “era of token compound interest”
OpenAI and Anthropic PBC, the strongest competitor in AI applications for a long time, have been racing to develop more advanced artificial intelligence agents (i.e. AI agent products) to streamline workflows in a wider range of fields. Previously, the two companies had achieved remarkable success with AI development tools that can automate code writing and complete debugging and actual deployment processes. Earlier this year, Anthropic launched a similar product, called Claude Cowork, with the goal of attracting a wider user base to join the unprecedented superwave of AI agents.
Both OpenAI and Anthropic have secretly submitted listing applications. An agency previously reported that Anthropic is expected to enter the US stock market as early as the fall of this year. OpenAI is considering launching next year.
Recent analysis by research firm SemiAnalysis reveals that Anthropic is reshaping the AI commercialization pattern by far exceeding the profitability and growth rate of its competitors. With a high-margin business model centered on APIs, Anthropic has become a leader in the B2B AI market. According to an in-depth report released by SemiAnalysis, the agency expects ANTHROPIC to achieve a profit of 1 billion US dollars before GAAP interest and tax in the third quarter of 2026, corresponding to a profit margin of about 6%. Meanwhile, its annual recurring revenue (ARR) has soared from $9 billion at the end of 2025 to over $60 billion now. The agency predicts that if Anthropic maintains a net ARR (NNARR) rhythm of about $15 billion per month, ARR is expected to reach 300 billion US dollars by the end of 2027, corresponding to a corporate value of 6 trillion US dollars, making it the company with the highest market capitalization in the world.
Anthropic's inflection point stemmed from the explosive popularity of Claude Code. According to statistics compiled exclusively by SemiAnalysis, Claude Code currently accounts for more than 7% of all code submissions on GitHub, directly driving the company's ARR increase in a single month in the first quarter to a crazy jump from $3 billion in January to $11 billion in March. Furthermore, according to SemiAnalysis estimates, Anthropic's current comprehensive gross margin has risen to the mid-60% range, while in 2024 this figure is negative 94%; of these, the gross margin of the API business exceeds 80%.
Since this year, the global capital's grand investment narrative of “seeking silicon-based inflation and weakening the carbon base” is essentially a shift of capital from “carbon-based assets” that rely on population, resources, and linear economic growth, such as traditional manufacturing, automobiles, consumption, real estate, and energy, to a high-end manufacturing chain around silicon wafers related to AI computing power infrastructure. Therefore, GPT-5.6, along with the arrival of ChatGPT Work and Anthropic commercialization data, have strengthened a core investment judgment: the unprecedented AI computing power infrastructure demand cycle is not over, but rather from the AI big model training drive stage to the AI inference application driven stage. The real AI computing power infrastructure supercycle may come from global companies deploying AI agents on a large scale as a new generation of digital employees. This also means that the current correction in the AI semiconductor sector is a healthy adjustment, not a sharp decline in the bear market driven by “excess computing power.”
Recent research by the well-known research institute Exponential View shows that the AI industry is upgrading from a capital expenditure cycle that mainly relies on cutting-edge model training to a two-wheel cycle where “training continues to expand, and reasoning becomes the main engine of incremental growth.”
Exponential View's latest estimates from the bottom up, excluding repeated calculation of supply chain paths, show that the revenue scale of terminals related to generative AI has reached 110 billion US dollars in the past 12 months, with an annual revenue operating rate of more than 175 billion US dollars in recent months, which is about three times the growth rate of the Internet and mobile waves; more importantly, AI computing power infrastructure-related revenue attributable to hyperscale cloud vendors can roughly cover depreciation expenses for new assets.
Exponential View's exclusive model also shows that for every 10% drop in token price, usage will increase by 12% to 18%, which means that a decrease in unit inference costs will not necessarily depress total revenue; on the contrary, it may expand overall computing expenses through demand elasticity. In other words, AI computing power infrastructure clusters, where the construction process is in full swing led by large AI tech giants such as Microsoft, Meta, and Google, is shifting from expected transactions of “building first, then finding revenue” to a closed economic loop of actual token consumption and verification of corporate payments.