Sweet burden! Explosive demand growth triggers a shortage of computing power, prompting Anthropic to urgently "sweep up" computing resources.

The CEO of Anthropic, Dario Amodei, stated that the company originally planned for a 10x growth, but the first-quarter revenue and usage increased 80 times on an annualized basis, explaining why the company is struggling to meet demand. Amodei said at a developer conference in San Francisco on Wednesday, "This is the reason we are having difficulties in terms of computing power." He added that the company is "working to provide more computing power as quickly as possible" and will "make this computing power available to users as soon as possible." Amodei said, "Software engineers are the fastest adopters of new technology. This forecasts how the entire economic system will operate in the future, and how artificial intelligence will change the economy." He also stated that the current growth level is "crazy" and "hard to cope with," and hopes that future expansion will be "more normal." Last month, Anthropic stated that the demand for Claude had put "inevitable pressure on the infrastructure", affecting the stability and performance of user experience, especially during peak hours. In the face of exploding demand and tight computing power, Anthropic adjusted the pricing model for its enterprise product Claude Enterprise last month. Under the new model, enterprise customers no longer pay a fixed subscription fee, but pay for additional computing power based on actual usage on top of a base fee of $20 per user per month. In the old model, enterprise customers had to pay up to $200 per authorized user per month for a certain amount of discounted token usage. Regarding the price adjustment, an official explanation provided by Anthropic was that under the old model, some heavy users frequently reached usage limits, causing work interruptions, while others paid but did not use the allocated amount. A company spokesperson stated that the new pricing "better reflects how customers are actually using Claude, as workloads are shifting from fixed seat productivity tools to intelligent body applications." Fredrik Filipsson, co-founder of Redress Compliance, a company that helps negotiate software licensing agreements for enterprises, stated that the new pricing may result in doubling or even tripling costs for heavy users. Explosive growth in demand! Anthropic is gathering computing power everywhere. Anthropic's explosive growth is mainly attributed to the popularity of its Claude artificial intelligence model, especially since the launch of Claude Code last year. Therefore, in addition to partnerships with Amazon.com, Inc. and NVIDIA Corporation, Anthropic is currently in an expansion phase of gathering computing power. Hours before Amodei made the above statements, Anthropic announced a computing power partnership with commercial space giant SpaceX, allowing Anthropic to use the entire computing power capacity of the SpaceX Memphis Colossus 1 data center. Anthropic stated that this partnership will significantly increase the company's computing power levels and will bring in over 300 megawatts of additional capacity in the next month, equivalent to 220,000 NVIDIA Corporation (NVDA.US) GPUs. Additionally, according to recent reports, Anthropic has committed to paying approximately $200 billion to Alphabet Inc. Class C over the next five years for training and deploying Claude's cloud computing power. This investment size reflects multiple previous agreements reached between Anthropic and Alphabet Inc. Class C. In October last year, Anthropic announced an expansion of its usage of Alphabet Inc. Class C's cloud TPU chips, with Alphabet Inc. Class C providing over 1 gigawatt of computing power that will be operational this year. Last month, Anthropic reached a new partnership agreement with Alphabet Inc. Class C and Broadcom Inc. (AVGO.US) to obtain about 3.5 gigawatts of TPU (tensor processing unit) computing power resources, expected to launch from 2027 onwards. Also, Amazon.com, Inc. (AMZN.US) announced further strategic cooperation with Anthropic last month. Anthropic will lock in up to 5 gigawatts of Amazon.com, Inc.'s current and next-generation Trainium chip computing power to train and support its advanced AI models. Sharp increase in annual revenue! Valuation aims at $900 billion The strong demand for Claude not only reflects the inability to keep up with computing resources, but is also directly reflected in Anthropic's revenue. Anthropic announced last month that the company's Annual Recurring Revenue (ARR) had exceeded $30 billion, a significant increase from $9 billion at the end of 2025. Meanwhile, a report from Semi Analysis in early May indicated that Anthropic's ARR had risen to around $44 billion. This growth rate far exceeds that of OpenAI during the same period. In terms of revenue structure, enterprise customers contribute over 80% of the revenue. Eight out of the top 10 Fortune companies are stable customers of Anthropic, and the number of customers paying over a million dollars annually has doubled. Among them, the annual revenue from the standalone product Claude Code is nearing $2.5 billion, over five times higher than the $400 million in mid-2025; enterprises contribute over half of the revenue from Claude Code. This means that Claude Code is not just a model feature that "writes code", but is beginning to be taken seriously as enterprise software that requires careful consideration of budgets, permissions, audits, and organizational management. For upstream clients, the most important change in such businesses is that the demand will be more sustained and reliant on stable delivery, rather than simply spiking when a model is released. Data on enterprise spending in early 2026 shows that among companies purchasing AI tools for the first time, Anthropic already has a 73% market share, surpassing OpenAI. This advantage in the enterprise market, along with its "security-first" product positioning, directly correlates with the fact that Anthropic is one of the leading large model manufacturers in AI safety and alignment research. The compliance and accuracy of its model outputs are significantly better than similar products, leading to a rapid increase in penetration in industries with strict regulations such as finance, law, and healthcare. At the same time, according to sources, Anthropic has initiated its latest round of financing, requesting investors to submit placement schemes promptly. The financing round is expected to be around $50 billion and is set to be completed by mid-May. Anthropic aims for a target valuation of around $900 billion, but given the increased demand for the company's shares from investors, the final valuation may likely exceed this figure. The growth trajectory of Anthropic's valuation is also remarkable. With a valuation of $183 billion in the F round in September 2025, the G round in February 2026 saw a valuation of $380 billion, representing an increase of over 107%. Now, the target valuation is set at $900 billion. If the final terms progress as planned, the valuation will likely double from February, surpassing or matching its main competitor OpenAI, which completed a record funding of $122 billion in March 2026, reaching a post-investment valuation of $852 billion. However, some analysts point out that based on the information disclosed during the preparation of the prospectus, Anthropic still faces three major obvious risks in its growth logic. Firstly, there is still structural risk in customer concentration. Despite having over 300,000 enterprise customers, the revenue is highly concentrated among top customers. Besides Amazon.com, Inc. and Alphabet Inc. Class C, which contribute over 20% of the revenue, the business model is deeply tied to cloud platforms. Anthropic has committed to purchasing up to $30 billion of Azure computing power from Microsoft Corporation, over $100 billion of AWS computing power purchases from Amazon.com, Inc. over the next decade, and up to $40 billion investments and 5 gigawatts of TPU computing power from Alphabet Inc. Class Cthese long-term agreements, despite being accompanied by substantial equity investments and equivalent commercial interests, may still impact its gross margin levels if core cloud providers raise prices or face supply fluctuations. On the revenue side, if top customers reduce purchases, shift to self-developed large models, or support other competitors, it will also directly impact revenue stability. Secondly, Anthropic's unique governance structure may raise concerns among investors. Anthropic employs a Public Benefit Corporation (PBC) structure, with the core governance mechanism being the Long-Term Benefit Trust (LTBT), holding special T-class shares with the right to elect a majority of the board members, meaning that even after going public, strategic decision-making will prioritize serving "long-term human interests" over shareholder returns. While this unique design provides policy advantages for Anthropic amidst increasing AI regulation, it may also be viewed as a "mission discount" by public market investorstrust institutions having veto power over certain business decisions may lead to profit being sacrificed for ethical considerations, limiting the voting rights and profits of investors. Balancing the trust mission with shareholder interests will be a core challenge for Anthropic. Lastly, the high valuation premium puts Anthropic under immense pressure to deliver on performance. With a target valuation of $900 billion, based on an ARR of around $44 billion, its price-to-sales ratio is around 20 times, still well above the industry average of 8-12 times for SaaS companies. To support the current valuation, Anthropic needs to achieve a revenue goal of $70 billion and cash flow of $17 billion by 2028, meaning it needs to maintain at least a 50% growth rate annually for the next three years while continuously improving its gross margins. Although the Semi Analysis report shows that the margin of Anthropic's inference infrastructure has increased from around 38% twelve months ago to over 70%, and the unit economic model is improving, the costs of training and inference of large models remain high, coupled with the price war pressures from competitors such as OpenAI, Alphabet Inc. Class C Gemini, achieving performance realization still poses a considerable challenge. It is worth noting that Anthropic's IPO process directly competes with OpenAI. After completing a $122 billion funding in March 2026, OpenAI's valuation already stood at $852 billion, with plans to launch an IPO in the second half of 2026. If Anthropic goes public first, it will absorb a significant amount of investor demand for AI stocks, potentially weakening OpenAI's market response to the IPO. This competitive pressure has forced both parties to accelerate their IPO preparations. Presently, Anthropic has hired Goldman Sachs Group, Inc., JPMorgan, and Morgan Stanley as the main underwriters, with long-time legal firm Wilson Sonsini leading the IPO compliance work, pacing slightly ahead of OpenAI. To transition from a challenger to establishing a market position that can compete with OpenAI, Anthropic still needs to overcome multiple hurdles in ecosystem building, cost control, and expanding the consumer market. The success of Anthropic's IPO and its acceptance in the capital markets will largely depend on whether investors are willing to pay a premium for its differentiated positioning of "security-first" and unique governance structure. Even if it successfully goes public, in an industry where the pace of large model technological iterations is less than twelve months, whether Anthropic can maintain technological dominance and continually increase market share will continue to be the focus of the market.