Anthropic has entered into a new collaboration with Alphabet and Broadcom, securing approximately 3.5 gigawatts of computing power through next-generation Tensor Processing Units (TPUs) to address rapidly growing market demand.
**Trend Analysis** The cooperation agreement between Anthropic, Alphabet, and Broadcom signifies a major step in diversifying the AI compute supply chain. Under the agreement, Anthropic will gain access to roughly 3.5 GW of AI computing power starting in 2027, supported by Alphabet's TPU processors and provided by Broadcom. This partnership not only reflects Anthropic's rapid business growth—revenue from its AI model "Claude" has surged from approximately $90 billion at the end of 2025 to over $300 billion currently—but also underscores Broadcom's strategic position in the custom AI chip sector. Broadcom's CEO, Hock Tan, previously forecasted that the company's AI-related revenue would exceed $100 billion by 2027, and this agreement reinforces that outlook. The collaboration also represents a structural shift in the AI compute market: alongside the dominant "first supply chain" led by NVIDIA's GPUs, a "second supply chain" combining Alphabet's TPUs and Broadcom's manufacturing capabilities is emerging. This trend towards diversification helps alleviate compute supply constraints and provides more sustainable infrastructure support for the AI industry.
**In-Depth Perspective** Anthropic, Alphabet, and Broadcom are deepening their partnership to deploy next-generation TPU computing power on a large scale. Anthropic has signed an agreement with Alphabet and Broadcom to jointly develop multiple gigawatts of new-generation TPU compute capacity, expected to be gradually deployed starting in 2027. This project represents Anthropic's largest compute investment to date, primarily located within the United States, further advancing its $50 billion U.S. compute infrastructure investment plan. This collaboration continues the strategic alignment established in October 2025, when Anthropic announced an expanded partnership with Alphabet to deploy up to one million TPUs (exceeding 1 GW of power) to support demand for the Claude model. TPUs, as Alphabet's self-developed AI chips, are being actively adopted by other large model developers, such as Meta Platforms, Inc., highlighting their competitiveness in high-performance inference and training scenarios.
**Surging Compute Demand Drives Competition in Foundational Hardware Ecosystems** Anthropic's commercialization progress shows explosive growth, with annualized revenue surpassing $300 billion and enterprise customers exceeding 1,000, leading to continuously rising compute demands. To reduce reliance on a single type of AI compute chip, major model developers are actively seeking diversified compute solutions. The competition between TPUs and GPUs is indirectly driving development in the wafer foundry, advanced packaging, and memory supply chains. By supplying TPUs, Alphabet is strengthening its cloud ecosystem, creating differentiated competition against NVIDIA's GPUs. The future landscape of the AI chip market may see further fragmentation.
**Investment Rationale** Anthropic's agreement with Alphabet and Broadcom for next-generation TPU compute, securing approximately 3.5 GW of support, significantly enhances the compute infrastructure for its AI model Claude, addressing rapidly growing market demand. Anthropic's annualized revenue has surpassed $300 billion, its enterprise customer count has doubled, and its commercialization process is experiencing explosive growth. The collaboration between Alphabet's TPUs and Broadcom reinforces the diversification and competitiveness of the AI chip supply chain, alleviating pressure on TSMC's CoWoS advanced packaging capacity. As a core supplier of custom AI chips, Broadcom anticipates its AI revenue will exceed $100 billion by 2027, and this agreement further solidifies its market position. Rising AI inference demand is propelling developments in memory technology, with HBM and on-chip SRAM becoming key focus areas, creating new growth opportunities for the supply chain. Capital expenditure by North American cloud providers continues to grow, driven by strong AI compute demand. ASIC chips are experiencing explosive growth, boosting prospects for key segments like wafer foundry, advanced packaging, and memory. The deepened collaboration between Anthropic and Alphabet not only reduces dependence on a single AI chip supplier but also optimizes cost structures, providing the AI industry with more sustainable and efficient compute solutions, benefiting the overall growth outlook for the AI产业链.
To capitalize on the AI compute opportunity, attention is drawn to the ChiNext Artificial Intelligence ETF (159363)—the largest and most liquid AI-themed ETF focused on the STAR and ChiNext markets—and its corresponding feeder funds (Class A: 023407, Class C: 023408), which stand to benefit directly from the growth surge driven by AI technology commercialization. From a sector perspective, the ChiNext Artificial Intelligence portfolio allocates approximately 70% to compute power (including leaders in optical modules/CPO) and about 30% to AI applications, representing not just a core "compute" play but also a genuine "AI application" investment.
**Risk Disclosure:** The Huabao ChiNext Artificial Intelligence ETF passively tracks the ChiNext Artificial Intelligence Index. The index's base date is December 28, 2018, and its release date was July 11, 2024. The annual performance of the ChiNext Artificial Intelligence Index from 2021 to 2025 was as follows: +17.57%, -34.52%, +47.83%, +38.44%, and +106.35%, respectively. The composition of the index's constituents is adjusted according to its compilation rules, and its past performance does not indicate future results. The index constituents mentioned are for illustrative purposes only; descriptions of individual stocks are not investment recommendations of any form and do not represent the holdings or trading动向 of any fund managed by the management company. The fund manager has rated this fund's risk level as R4 (Medium-High Risk), suitable for Aggressive (C4) and higher investor profiles. Suitability assessments should be based on the selling institution's evaluation. Any information appearing herein is for reference only, and investors are responsible for any independent investment decisions. Furthermore, any views, analysis, or forecasts contained herein do not constitute investment advice to readers and shall not be liable for any direct or indirect losses resulting from the use of this content. Fund investment carries risks; past performance of a fund does not guarantee future results, and the performance of other funds managed by the fund manager does not constitute a guarantee of this fund's performance. Fund investment should be approached with caution.
Comments