Asmi expands production, TSMC increases investment, and Anthropic's profits leap: Is the theory of "excessive computing power" being challenged, and is the AI bull market moving towards the era of compound reasoning?

Wedbush Securities, a well-known investment institution on Wall Street, recently released a research report showing that the lithography giant ASML Holding NV ADR (ASML.US) has raised its 2027 performance guidance above expectations, which is very favorable for the future prospects of the bull market related to AI computing infrastructure themes associated with advanced process logic chips such as CPU/GPU, as well as dynamic random-access memory (DRAM) chips. The latest views of Wall Street institutions such as Wedbush, as well as the strong performance and optimistic outlook of ASML Holding NV ADR and Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, undoubtedly serve as a strong stimulant for the AI computing theme, which has been hit hard by extreme and violent selling due to overcrowding and highly leveraged long positions. From Wedbush's bullish outlook on the AI computing theme, to the explosive performance and strong performance outlook of ASML Holding NV ADR and Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, the two most important upstream capacity and supply forces in the AI computing industry chain, and to SemiAnalysis' latest analysis showing that Anthropic has transitioned from long-term losses to explosive profit growth, are all sending an important signal to the global stock market: the AI computing industry chain has gradually entered a new stage of "exponential expansion of AI inference computing power driven by the scale deployment of intelligent agents", These latest signals are a blow to the recent pessimistic view that "excess computing power" has led to a sharp decline in the AI computing theme, especially the AI semiconductor sector. Wedbush's core judgment is that the upward revision of ASML Holding NV ADR's 2027 outlook and consideration of further increasing 2028 EUV capacity is completely in line with the strong demand growth pace seen in advanced process logic chips (i.e., logic chips below the 5nm process, such as CPU/GPU/TPU) and high-end DRAM to support AI data center construction processes; ASML Holding NV ADR lithography equipment shipments are expected to increase in 2027, which is expected to be converted into initial wafer manufacturing output at the end of 2027 to early 2028, especially in high-end DRAM/HBM memory chip capacity, but larger chip manufacturers such as Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, as well as global tech giants like Microsoft Corporation, Meta, continue to grow their AI-related capital expenditures, indicating uncertainty about when supply will truly catch up with demand. Wedbush Securities' senior analyst Matt Bryson wrote in the latest report to clients: "We believe that ASML Holding NV ADR's unexpected upward revision of the 2027 outlook, along with the potential significant increase in 2028 EUV lithography machine production, is almost in line with the strong demand growth for advanced process logic chips and high-end DRAM aimed at supporting the flourishing global AI computing infrastructure construction process." "The timeline of the increase in shipments is also very consistent with our baseline judgment, that is, the increase in capital expenditures related to AI computing in 2027 will ultimately drive an increase in the production of advanced process logic chips and DRAM at the end of 2027 to early 2028, especially in HBM/DRAM memory chip production; however, given that capital expenditures are continuing to grow, we remain uncertain about when supply will eventually catch up with the expanding demand curve." stated the Wedbush stock market analyst team led by Bryson. In addition to the 2027 performance guidance, ASML Holding NV ADR management currently expects full-year 2026 total net sales (i.e., total revenue outlook) to range from 43 billion euros to 45 billion euros, well above the previous forecast range of 36 billion euros to 40 billion euros, and higher than the Wall Street analyst average expectation of 39.3 billion euros. From EUV to 5GW superclusters: Four-fold industry signals shatter the "AI computing excess" argument Financial data shows that ASML Holding NV ADR's second-quarter revenue reached 9.326 billion euros, ahead of market expectations of 8.8 billion euros; net profit was 2.918 billion euros, also exceeding expectations of 2.62 billion euros, with a gross margin of 54%. More importantly, the company has significantly raised its 2026 revenue guidance from 36-40 billion euros to 43-45 billion euros, its gross margin guidance from 51-53% to 54-56%, and expects third-quarter revenue to further increase to 11-12 billion euros. ASML Holding NV ADR management referred to the first half-year orders as "extremely strong", with customers accelerating the construction of advanced process logic chips and storage chip production capacity, hence planning to increase the low-numerical aperture extreme ultraviolet lithography machines and immersion deep ultraviolet lithography machines by about 30% for 2027, and to study a further 30% increase for 2028. This means that major customers like Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, Intel Corporation, have voted for strong demand for AI computing in 2027-2028 by committing to years of lithography machine equipment, rather than oral predictions. The rare significant increase in production capacity by ASML Holding NV ADR means that major advanced logic chip manufacturers such as Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, Intel Corporation are preparing for larger-scale wafer capacity for custom AI ASIC accelerators such as CPUs, GPUs, and TPUs at the server level, while Samsung's latest performance and the storage chip shortage signaled by SK Hynix's CEO on the first day of its ADR listing on the US stock market, significantly validates the strong demand for AI server hardware. Furthermore, ASML Holding NV ADR management plans to increase the annual production capacity of about 65 low-numerical aperture EUV lithography machines by 30% in 2027, and to study a further 30% increase in 2028, with similar expansion path for DUV immersion lithography machines. This is not just about short-term inventory replenishment, but about global wafer factories locking in advanced logic, HBM/DRAM, and advanced packaging capabilities through multi-year equipment commitments. This week's news in the AI industry shows that Meta, Facebook's parent company, has expanded the investment and construction scale of its Hyperion data center campus in Louisiana from the initial $10 billion project announced in 2024, to over $50 billion and a supercluster with a computing power of over 5GW, and has engaged 27,500 NVIDIA CorporationRubin chips to ignite the "Japanese Siasun Robot & Automation national team", along with NVIDIA Corporation CEO Jensen Huang's collaboration with established Japanese industrial giants such as FANUC, Yaskawa Electric, Kawasaki Heavy Industries, and Fujitsu on the first day of the ADR listing on the US stock market, and including SoftBank, NEC, Hitachi, Sony, Preferred Networks into the Cosmos ecosystem, and coupled with Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, the exclusive manufacturer of NVIDIA Corporation AI chips, recently announced strong performance above expectations and an increasingly optimistic outlook for the future of AI computing demand, all of which prove that the global AI computing investment cycle is far from over, and the expansion of computing power is transitioning from a unilateral arms race led by US hyperscale cloud providers to a multi-center resonance of "cloud AI, sovereign AI, and physical AI". The performance and outlook of Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR from the semiconductor manufacturing end verifies that this demand is not only present in equipment orders: the company's second-quarter US revenue reached $40.2 billion, a year-on-year increase of 33.7%; with a net profit of NT$706.56 billion, a year-on-year increase of 77.4%, significantly higher than the market's expected NT$632.6 billion; the high-performance computing business accounted for 66% of revenue, and the 7nm and below advanced processes accounted for 77% of wafer business revenue. Even more significant is that Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR's third-quarter revenue guidance has been raised to $44.6 - $45.8 billion, the company has also significantly raised its 2026 capital expenditure from $52 - $56 billion to $60 - $64 billion, raised its annual revenue growth guidance to slightly above 40%, and added $100 billion in US investment, bringing its total promised investment in the US to about $265 billion. As the final capacity provider and manufacturing chain receiver to meet the AI computing demand for core AI chip customers such as NVIDIA Corporation, Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, through profitability, utilization, and capital expenditure, all indicate that the visibility of the demand for AI computing infrastructure around advanced process logic chips, cutting-edge 2nm processes, and advanced packaging is on the rise, rather than being in a sluggish period of global AI computing demand. ASML Holding NV ADR and Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR are not merely mechanical passive recipients of customer forecasts, but are approving expansion plans after reviewing data center construction and end demand, thereby providing strong evidence to counter the argument of "systematic excess of computing power". The demand side is also continuing to expand: the Noetra project, supported by the Japanese government, has purchased 27,500 NVIDIA Corporation Rubin chips to build a physical AI infrastructure, with the project scheduled to start construction in April 2027 and be operational in June 2028; Meta has expanded its Hyperion campus in Louisiana from a $10 billion project announced in 2024 to a supercluster with a value of over $50 billion and a computing power of 5GW, these four-fold industry signals undoubtedly shatter the argument of "AI computing excess". Anthropic is poised to achieve a significant leap in profitability trajectory, ushering the AI computing bull market into the era of "Token compounding" OpenAI and its strongest competitor in the AI application field, Anthropic PBC, have been competing to develop more advanced artificial intelligence agents to simplify workflows in a wider range of areas. Both companies have already achieved significant success with AI development tools that can automate code writing, complete debugging, and deployment processes. Earlier this year, Anthropic launched a similar product called Claude Cowork, which aims to attract a wider user base to join the unprecedented wave of AI intelligent agents. Both OpenAI and Anthropic have submitted secret applications to go public. Some institutions have previously reported that Anthropic may debut on the US stock market as early as this fall. OpenAI is considering going public next year. The latest analysis from the research firm SemiAnalysis reveals that Anthropic is reshaping the commercial landscape of AI with profitability and growth rates far exceeding its competitors. With a high-margin business model centered around APIs, Anthropic has become a leader in the B2B AI market. A deep report released by SemiAnalysis shows that the agency expects Anthropic to achieve $1 billion in GAAP EBITDA in the third quarter of 2026, corresponding to a profit margin of about 6%. At the same time, its annual recurring revenue (ARR) has surged from over $9 billion at the end of 2025 to over $60 billion. The agency predicts that if Anthropic maintains a net new ARR (NNARR) pace of around $15 billion per month, its ARR is expected to reach $300 billion by the end of 2027, corresponding to an enterprise value of $6 trillion, making it the world's most valuable company. Anthropic's inflection point in performance stems from the explosive popularity of Claude Code. Statistics compiled exclusively by SemiAnalysis show that Claude Code currently accounts for over 7% of all code submissions on GitHub, directly driving the company's ARR from a wild surge of $3 billion in January to $11 billion in March. In addition, SemiAnalysis' data calculations show that Anthropic's comprehensive gross profit margin has risen to the mid-60% range, compared to negative 94% in 2024; with the API business gross margin exceeding 80%. The grand investment narrative of global funds seeking "silicon-based inflation, weakening carbon-based assets" this year essentially involves capital shifting from traditional manufacturing, automotive, consumer, real estate, energy, and other "carbon-based assets" that rely on population, resources, and linear economic growth, to the high-end manufacturing chain based around silicon chips related to AI computing infrastructure. Therefore, the arrival of GPT-5.6 and ChatGPT Work, as well as the commercialization data around Anthropic, together reinforce a key investment judgment: the unprecedented demand cycle for AI computing infrastructure is not over, but rather transitioning from a phase driven by training large AI models to a phase driven by the deployment of AI inference applications, the true super cycle of AI computing infrastructure may come from global enterprises deploying AI intelligent agents as the next generation of digital workers on a large scale, which also means that the recent pullback in the AI semiconductor sector is a healthy adjustment, and not a bear market crash driven by "excess computing power". The latest research from renowned research firm Exponential View shows that the AI industry is transitioning from a capital expenditure cycle primarily driven by front-end model training to a dual-cycle of "continued expansion in training and inference becoming the main driver". Exponential View's exclusive model, calculated from the bottom up and eliminating duplicate calculations in the supply chain path, shows that the terminal revenue scale related to generative AI in the past 12 months has reached $110 billion, with a recent annualized revenue run rate of over $175 billion, a growth rate about three times that of the internet and mobile waves during the same period; more importantly, revenue generated from AI computing infrastructure related to super-hyperscale cloud providers has roughly been able to cover the depreciation expenses of new computing assets. Exponential View's exclusive model also shows that for every 10% decrease in token price, usage will increase by 12-18%, indicating that a decrease in unit inference cost will not necessarily depress total revenue; instead, it may expand total computing expenditure through demand elasticity. In other words, the construction of AI computing infrastructure clusters around super-large cloud providers such as Microsoft Corporation, Meta, and Alphabet Inc. Class C, is transitioning from an expected trade of "build first, then find revenue" to an economic cycle verified by actual token consumption and enterprise payment validation.