How should a group of "shadow assets" positioned on the equity, computing power, and enterprise distribution chain choose their investments? The first trillion-dollar AI startup is about to be born. As of this writing, according to Nasdaq Private Market (NPM) data, Anthropic's valuation has soared from $650 billion in early May to approximately $992 billion, just shy of the trillion-dollar mark. It's important to note that the NPM valuation is not the same as Anthropic's latest official funding valuation, nor is it the same as a simple order price on a platform. It integrates various data, including secondary market transactions, buy and sell quotes, and previous funding rounds, to provide an estimate closer to the current fair value in the private market. In other words, while Anthropic hasn't officially confirmed its "trillion-dollar valuation" in its official funding announcement, the private market has already begun to place it within the trillion-dollar asset pricing range. This raises a more practical question: aside from pre-IPO channels like MSX, it's difficult for ordinary investors to directly buy Anthropic. Are there other ways in the public market to capture the spillover opportunities brought about by Claude's rising valuation? The answer may not lie solely in traditional "AI stocks." I. Why did Anthropic break the trillion-dollar mark first? Undoubtedly, Anthropic's soaring valuation has become one of the most explosive stories in the global capital markets of 2026. Looking at the numbers alone, it's a very steep curve: In March 2025, Anthropic completed a $3.5 billion funding round, valuing the company at $61.5 billion post-money; in September 2025, its valuation rose to $183 billion; in February 2026, it completed another $30 billion funding round, reaching a valuation of $380 billion; and according to recent media reports, it is currently undergoing a new round of funding exceeding $900 billion, surpassing OpenAI's valuation of approximately $852 billion; coupled with the nearly trillion-dollar pricing signals given by private market data such as NPM, on the surface, this is another instance of the capital market paying for AI. The deeper change lies in the fact that the capital market is beginning to redefine the ceiling for cutting-edge model companies. This isn't entirely the same as the early consumer-facing explosion like ChatGPT, because OpenAI's strengths lie in consumer-level entry points, developer ecosystems, and brand awareness, while Anthropic's advantages are increasingly concentrated in enterprise scenarios, especially in industry applications with high security requirements, such as code and agent automation. For the capital market, this means Claude is not just a chatbot, but rather a kind of underlying infrastructure that can be embedded into daily enterprise production processes. This shift is directly reflected in revenue and profit expectations.According to recent media reports, Anthropic expects its revenue to reach $10.9 billion in the second quarter of 2026, a significant increase from $4.8 billion in the first quarter, and is expected to achieve a quarterly operating profit of $559 million. If this performance materializes, Anthropic will become one of the few cutting-edge AI companies that, despite its intensive investment in computing power, is still close to achieving profitability. This is also a core difference between its narrative and that of OpenAI. Of course, Anthropic is not running wildly in a market without competition. In recent days, OpenAI has been continuously improving its product performance and reputation in scenarios such as Coding and Agentic Workflow through Codex+5.5. Although Gemini and others have experienced a decline in reputation, they have not withdrawn from this model arms race. In other words, Anthropic's high valuation is not a result of a "determined outcome," but rather the market's willingness to pay a higher premium for its growth curve, the performance of products such as Claude Code, and a clearer commercialization path. It is precisely because of this scarcity that Anthropic has become one of the most watched targets in the pre-IPO market. MSX's recently concluded second Pre-IPO round included Anthropic as one of its two selected targets, with a subscription price of 855 USDT, corresponding to a valuation of approximately $950 billion. For many ordinary investors, this type of Pre-IPO product did provide a more accessible entry point closer to private market pricing, given that Anthropic was not yet listed and the primary market had extremely high barriers to entry. However, Pre-IPO is not the only perspective. When Anthropic's valuation reaches a trillion dollars, the market may need to re-evaluate not only Anthropic itself, but also the entire shadow chain surrounding Claude: whoever invested in it, whoever provided it with computing power, and whoever brought it into enterprise software could be re-marked by capital. Therefore, besides Pre-IPO, which companies are positioned in Claude's equity, computing power, and enterprise distribution chain? II. What will the market buy if Anthropic isn't available? If we categorize Claude concept stocks based on their "closeness to Anthropic," they can be roughly divided into three categories: The first category consists of equity shadow stocks that have directly participated in Anthropic's investment; the second category comprises cloud and chip companies that meet Claude's computing power needs; and the third category consists of software platforms that integrate Claude into enterprise workflows. While all three categories are called "Claude concept stocks," their ways of benefiting are completely different: equity shadow stocks benefit from the book value revaluation resulting from Anthropic's increased valuation; computing power chains benefit from the orders generated by the expansion of Claude's training and inference needs; and enterprise software platforms benefit from whether Claude can become a native capability in their own products. 1.Tier 1: "Shadow Stocks" with Direct Equity Investments and Accessible in the Public Market. Public information shows that Anthropic has undergone seven rounds of financing from Series A to Series G since its inception. Major shareholders include Google, Amazon, Nvidia, Microsoft, Sequoia Capital, Blackstone, and GIC. After completing its Series G financing (post-investment valuation of approximately $380 billion), the shareholding ratios of the major investors were: Amazon (9%), GIC (8%), Microsoft (7%), Coatue Management (6%), Google (6%), Nvidia (5%), the founders and team held 21%, and the employee option pool held 19%. However, the one that most resembles a "hidden shadow stock" is not actually the tech giants like Amazon, Google, and Microsoft, which have invested tens or even hundreds of billions, but Zoom. Back in 2023, Zoom announced a strategic partnership with Anthropic, investing approximately $51 million through Zoom Ventures (at a time when Anthropic was valued at only $4.1 billion). The collaboration aimed to integrate Claude into the Zoom platform, gradually expanding its product lines to include Team Chat, Meetings, and more. What's interesting about this partnership is that while Zoom's initial investment was relatively small, even considering subsequent funding rounds, the stake is worth between $2 billion and $4 billion, or even more. Compared to Zoom's current market capitalization of approximately $29 billion, this single investment represents 7% to 15% or even more of its market value. For giants like Amazon and Google, similar investments might be considered non-core items on their financial statements, but for Zoom, the significance is entirely different. Zoom's current market capitalization is around $300 billion; if Anthropic's stake is valued at several billion dollars, it would be a crucial variable influencing the market's reassessment of its asset value. This is what makes Zoom unique—a small temple holding a large Buddha. Over the past few years, Zoom's core video conferencing business has seen a gradual slowdown in growth, and market expectations for it have significantly diminished compared to the pandemic period. However, if we consider the Anthropic stake, Claude integration, and the AI-driven enterprise customer service center and collaboration scenarios together, Zoom is no longer just a video conferencing company experiencing slowed growth; it's a publicly traded super-shadow stock that unexpectedly holds an early ticket to Claude. A similar logic of shadow equity can be extended to SK Telecom. In 2023, SK Telecom announced an additional $100 million investment in Anthropic and partnered with Anthropic to develop a multilingual large-scale model for the telecommunications industry. Compared to Zoom, SK Telecom is unique in that it is a traditional telecommunications operator with a relatively small market capitalization. Therefore, the book value of its Anthropic stake may have a more significant impact on its overall valuation. For this reason, overseas markets have also viewed SK Telecom as a stranger but more direct shadow asset of Anthropic. 2.The second tier: the computing power ecosystem, from cloud vendors and AI ASICs to Neo-cloud. However, beyond Zoom and SK Telecom, the larger Claude industry chain is actually hidden in the computing power layer. If equity shadow stocks are about the book revaluation brought about by the upward valuation of Anthropic, then computing power ecosystem stocks are about another issue: the larger Claude becomes, the more enterprises call for it, and the heavier the code and agent scenarios become, who will undertake the training, inference, and data center needs behind it? This chain cannot be viewed solely by NVIDIA, nor solely by traditional cloud vendors. More accurately, Claude's computing power ecosystem can be divided into at least three groups: the first group is cloud platforms such as AWS, Google Cloud, and Azure; the second group is AI ASICs such as TPU and Trainium and their supply chains; and the third group is Neo-cloud, which specializes in providing AI computing power leasing, such as CoreWeave, Nebius, Lambda, and Crusoe. For example, Anthropic's partnership with Amazon is the earliest and deepest, as Amazon has already invested billions of dollars in Anthropic. AWS is also one of Anthropic's most important cloud and training partners, with both sides engaging in in-depth cooperation in areas such as AWS Trainium, the Neuron software stack, and Project Rainier. Therefore, for Amazon, Anthropic is not just a financial investment project, but also a crucial tool for AWS to compete for cloud workloads in the era of generative AI. Google represents a different approach, as it invested in Anthropic early on and subsequently expanded its cloud and TPU cooperation with Anthropic. In 2026, Anthropic, Google, and Broadcom further expanded their cooperation, planning to obtain multi-GW-level next-generation TPU computing power starting in 2027 to support the expansion of Claude models and enterprise applications. Objectively speaking, Anthropic's diversified computing power strategy—using AWS Trainium, Google TPUs, and accessing NVIDIA architecture through Microsoft Azure—reduces its reliance on a single vendor and indirectly benefits more publicly traded companies from Claude's growth. This also brings to mind a previously overlooked area: the AI ASIC chain. In the past, discussions about AI computing power primarily focused on NVIDIA GPUs. However, as leading model companies increasingly prioritize inference costs, supply stability, and unit token costs, the importance of cloud vendors' self-developed chips and custom ASICs is rising. Therefore, tools like AWS Trainium and Google TPUs essentially aim to provide a more controllable cost structure for large-scale model training and inference, in addition to GPUs.Within this chain, Broadcom is one of the companies most worthy of separate discussion regarding Claude-related stocks. While not a publicly disclosed equity investor in Anthropic, it is a key chip and network supplier behind the Google TPU ecosystem. If Anthropic's future growth does indeed rely on larger-scale TPU deployments, then Broadcom will become an indispensable hardware and network node in this chain. Extending further, Marvell, TSMC, advanced packaging, optical interconnects, and high-speed networks can also be observed within the broader AI ASIC industry chain: Broadcom's connection to Claude is more direct, as it stands at the intersection of Google TPU and Anthropic's expanded cooperation; other ASIC and semiconductor supply chain companies are more like industry betas driven by the expanding demand for AI computing power, and not necessarily exclusive beneficiaries of Claude's success. Microsoft and NVIDIA entered into a clearer cooperation framework by the end of 2025, with NVIDIA and Microsoft committing to invest up to $10 billion and $5 billion respectively in Anthropic. This is interesting, as it signifies that Microsoft is buying "insurance beyond OpenAI" for its AI ecosystem. NVIDIA's logic is more direct: regardless of whether Anthropic uses Trainium, TPU, or NVIDIA GPUs, as long as the competition in cutting-edge models continues to escalate, NVIDIA will remain one of the most unavoidable core computing power providers. However, in Claude's line of thought, NVIDIA is not the only winner, because Anthropic emphasizes the diversification of computing power sources more than many model companies. Besides traditional cloud providers and AI ASICs, there is a newer group of computing power beneficiaries: Neo-cloud. Simply put, these are new cloud providers that specialize in providing high-density GPU/accelerated chip computing power leasing for AI training and inference. Unlike AWS, Azure, and Google Cloud, which offer a wide range of cloud services, they focus more on AI workloads, localized high-performance clusters, GPU-as-a-Service, and the elastic computing power needed by model companies. Within this chain, CoreWeave's relationship with Anthropic is the most direct. In April 2026, CoreWeave announced a multi-year agreement with Anthropic, under which Anthropic will use CoreWeave's cloud platform to run production-grade workloads. This also means that Claude's computing power ecosystem is not a binary choice between "cloud vendors vs. chip companies," but rather a multi-layered structure: the bottom layer includes different chip options such as NVIDIA GPUs, Google TPUs, and AWS Trainium; the middle layer includes traditional cloud platforms like AWS, Google Cloud, and Azure; and the remaining layer includes Neo-clouds such as CoreWeave, Nebius, Lambda, and Crusoe, providing a more flexible computing power leasing and delivery layer.Therefore, to understand the Claude computing power ecosystem more comprehensively, Broadcom represents the AI ASIC and custom chip chain, CoreWeave represents the Neo-cloud computing power leasing chain, and Amazon, Google, and Microsoft represent the computing power entry points for traditional cloud platforms. These three groups of companies collectively illustrate one point: the higher Anthropic's valuation, the more the market reprices not only Claude itself, but also the increasingly complex network of computing power procurement, chip design, and cloud infrastructure behind it. 3. The Third Tier: Software Platforms Bringing Claude into Enterprise Workflows Besides equity and computing power, there is a third category of companies that are more easily overlooked: enterprise software platforms. The most typical examples here are Salesforce, SAP, Snowflake, and ServiceNow. Salesforce Ventures has been involved in Anthropic since its early funding rounds and has continued to support it in subsequent rounds of funding. More importantly, Claude has already been integrated into Salesforce's product ecosystem, including Slack and Agentforce. Especially in industries with higher security and compliance requirements, such as finance, healthcare, and the public sector, Claude has the potential to become a key model in Salesforce's enterprise AI solutions. SAP's logic, on the other hand, leans more towards core enterprise systems. In 2023, SAP announced strategic investments in generative AI companies like Anthropic, Cohere, and Aleph Alpha; in 2026, SAP further announced an expanded partnership with Anthropic, planning to use Claude as one of the main inference and agent capabilities in SAP's Business AI Platform, Joule, and Joule agents. This path is significant because SAP connects to the core ERP, finance, HR, supply chain, and operational management systems of enterprises. If Claude can enter SAP, it's not just entering a software entry point, but rather entering the most fundamental business processes and data structures of global enterprises. Snowflake and ServiceNow represent another type of enterprise AI distribution path. Snowflake and Anthropic have expanded their collaboration, committing $200 million to jointly promote the Claude model in Snowflake Cortex AI, Snowflake Intelligence, and enterprise data analytics agents. ServiceNow announced that Claude has become the default model for its ServiceNow Build Agent and is being used for application development, industry workflows, and internal employee productivity improvements. ServiceNow also stated that it has deployed Claude to tens of thousands of employees. These companies are not direct beneficiaries of Anthropic's increased valuation, but they represent another more important direction: Claude is evolving from a standalone AI product into an inference engine and workflow engine within enterprise software. For the public market, this may actually be a more sustainable trend.Because the book value of equity shadow shares has an upper limit, and computing power orders are easily affected by capital expenditure cycles, if Claude truly becomes the default intelligent layer in enterprise software, then companies like Salesforce, SAP, Snowflake, and ServiceNow have the opportunity to alleviate market concerns about "AI disrupting SaaS." In short, Anthropic's rise doesn't necessarily mean only a threat to traditional software companies; it could also mean that a group of enterprise software companies have the opportunity to repackage their valuation logic. Furthermore, if we extend enterprise workflows further to government, intelligence, and defense scenarios, Palantir deserves special mention. It is becoming an important distribution platform for Claude to enter high-security scenarios in the US government: In 2024, Palantir, Anthropic, and AWS announced a collaboration to integrate Claude 3 and Claude 3.5 series models into Palantir AIP, providing services to US intelligence and defense agencies; subsequently, Anthropic joined the Palantir FedStart project, promoting Claude for Enterprise to enter government departments with FedRAMP High and DoD IL5 standards. III. How to See the Whole Picture of "Claude Concept Stocks" Therefore, to create a clearer framework for "Claude concept stocks," we cannot simply look at who appears in the same press release as Anthropic, nor can we package all AI collaborations as equity benefits. What truly needs to be distinguished are three completely different ways of benefiting: The first is equity benefit: Zoom, Salesforce, SAP, Amazon, Google, Microsoft, NVIDIA, and financial institutions such as Blackstone, Goldman Sachs Alternatives, and JPMorgan Chase that appear in Anthropic Series G can all be observed within this framework; the second is computing power benefit: Amazon, Google, Microsoft, NVIDIA, Broadcom, CoreWeave, and more broadly, AI ASIC and Neo-cloud computing power leasing companies are the core of this layer.These aren't necessarily pure Anthropic stock alternatives, but if Claude's enterprise inquiries, code agents, and large-scale model training continue to expand, the first to take on actual orders and infrastructure spending will often be cloud, chip, ASIC, GPU clusters, and computing power leasing platforms. In particular, Amazon's AWS Trainium, Google's TPUs, Microsoft Azure + NVIDIA architecture, Broadcom's position in the Google TPU supply chain, and Neo-cloud platforms like CoreWeave's handling of production-grade AI workloads collectively constitute the computing power landscape behind Claude. The third type is distribution beneficiaries: enterprise software and collaboration platforms like Salesforce, SAP, Snowflake, ServiceNow, and Zoom, as well as Palantir targeting government and defense scenarios, don't really benefit from associating with Claude, but rather from whether they can integrate Claude into their own products. If Claude can help these companies improve customer retention, increase AI revenue streams, and strengthen their agent product narratives, then they are not just Anthropic partners, but rather Claude's partners. This provides a gateway into the corporate world; among these three types of companies, the most noteworthy aspect is valuation asymmetry. For Amazon, Google, Microsoft, and NVIDIA, Anthropic, however important, is merely one part of their vast AI strategies. Anthropic's rising valuation does validate their investments in AI infrastructure and model ecosystems, but it's unlikely to solely determine the stock prices of these trillion-dollar giants. However, the situation is entirely different for companies like Zoom and SK Telecom. Their own market capitalization is relatively limited. If Anthropic's equity value reaches billions of dollars, it could have a more direct impact on their own valuations. What the market is truly interested in is whether this investment is large enough relative to their own market capitalization and whether it hasn't been fully priced in. This also reveals the core logic of so-called "shadow stocks": it's not that whoever has the strongest relationship with Anthropic is the most resilient; rather, it's that whoever has the most asymmetrical value exposure to Anthropic relative to their own market capitalization is the most likely to be rediscovered by capital. Meanwhile, another noteworthy new signal is that Claude's collaboration model is also changing. Initially, most companies' relationship with Anthropic was simply investing and using Claude incidentally. However, by 2025-2026, this collaboration had begun to evolve into "joint products + computing power binding + enterprise workflow embedding"—Snowflake signed a $200 million cooperation agreement, ServiceNow set Claude as the default model for Build Agent, SAP put Claude into Joule and Business AI Platform, and Salesforce combined Claude with enterprise portals such as Slack and Agentforce.This indicates that Claude is transforming from a model provider into a capability layer within enterprise IT infrastructure. This presents a dilemma for many investors: those wanting to participate in this trillion-dollar pricing game must either position themselves early through pre-IPO products or indirectly observe this trend through shadow stocks in public markets like the US and South Korea. However, it's crucial to emphasize that shadow stocks are not the real entity. When Anthropic's valuation rises, the market may re-evaluate the asset value, partnership value, and computing power orders of these companies; conversely, if Anthropic's new round of financing falls short of expectations, the IPO process is delayed, or private market valuations correct, these "Claude concept stocks" may face a reverse correction. The story of the leading AI player is captivating, but position management is always more important than the story itself. In conclusion, over the past two years, the market's AI purchases have primarily focused on NVIDIA, cloud providers, and the "Seven Sisters" (referring to a group of AI companies). However, since 2026, AI transactions have become more segmented: on one hand, cutting-edge model companies like OpenAI and Anthropic are driving up valuations in the primary market; on the other hand, the public market is starting to look for the underlying assets behind them, ranging from computing power, electricity, and cloud services to enterprise software, data platforms, and collaboration portals. Anthropic is unique in that it stands at the intersection of these chains. It has the capital and computing power support of giants like Amazon, Google, Microsoft, and NVIDIA, as well as early-stage investments and enterprise scenario collaborations from companies like Zoom, Salesforce, and SAP. Furthermore, it enters more specific enterprise workflows through software platforms like Snowflake and ServiceNow. While they may not all replicate Anthropic's valuation elasticity, it at least illustrates one thing: the AI pricing game has spread from model companies themselves to a broader ecosystem of capital, computing power, and enterprise software. [MSX Maitong]