Anthropic is in a close fight with OpenAI, and its path to listing has hit an additional hurdle.

Dario Amodei, the CEO of Anthropic, wrote a 20,000 - word long article in recent days. He wrote, "Humans are about to acquire almost unimaginable power, and it is still very unclear whether our social, political, and technological systems have the maturity to harness this power."

Amodei believes that in just a few years at most, AI will outperform humans in almost all aspects, and this is inevitable.

Meanwhile, he and the co - founder made a commitment that is almost unprecedented in Silicon Valley. They will donate 80% of their personal wealth to charity, and Anthropic's employees also promised to donate their company shares.

Coincidentally, right after making the charitable commitment, Amodei was hit by a combination of products from OpenAI.

In the past, this wouldn't have caused much of a reaction from Anthropic. But now it's different. At this moment, it's a financing race between Anthropic and OpenAI.

Two of the world's top AI companies are simultaneously launching financing rounds, each seeking over tens of billions of dollars. Moreover, for Anthropic, this is the last round of financing before going public, so they can't afford to be careless.

Anthropic's only option is to respond, either by accelerating the financing process or speeding up product iteration.

However, OpenAI's offensive this time is not a large - scale all - out attack but a precise strike on Anthropic's most core business. As a result, the time window for Anthropic to prove itself is rapidly narrowing.

Release of OpenAI's Product Series

On January 27, 2026, OpenAI officially launched Prism, an AI - native scientific research workspace powered by GPT - 5.2.

Prism is a cloud - based scientific research platform that deeply integrates LaTeX editing, literature retrieval, formula management, and collaborative review.

Its core value lies in integrating the previously fragmented toolchain in the scientific research writing process into a single workflow, enabling GPT - 5.2 to work within the complete context of a paper and understand the relationships among formulas, references, charts, and the overall structure.

Those who have written science and engineering papers must have experienced that in the past, you had to write in Word first, then switch to Overleaf to edit LaTeX formulas, open Zotero to manage references, and then spend several hours creating charts.

Now, Prism can handle all the above processes on a single platform and also introduces ChatGPT to obtain information through AI.

Prism is built on the foundation of OpenAI's acquisition of the cloud - based LaTeX platform Crixet.

By deeply integrating Crixet's mature technology with GPT - 5.2's reasoning ability, Prism has achieved several breakthrough functions.

Not only that, it can directly retrieve relevant literature from academic platforms such as arXiv and automatically generate a reference list based on the paper's context. More importantly, it supports an unlimited number of projects and collaborators and is completely free for all ChatGPT personal account users.

Kevin Weil, the vice - president of science at OpenAI, made a bold prediction at the press conference: "I think 2026 will be to AI and science what 2025 was to AI and software engineering."

Science research and AI programming are OpenAI's main focuses this year.

Speaking of AI programming, OpenAI is currently in its Codex release month and will continuously release Codex - related products from January 24 to February 24.

Actually, since the second half of 2025, OpenAI has been continuously strengthening its developer tools ecosystem. Codex has evolved from an initial independent API into a complete developer platform, including the Codex CLI command - line tool, the Codex Web cloud - based autonomous coding agent, and IDE integration plugins.

According to OpenAI's official news, since August 2025, the daily message volume of Codex users has increased tenfold, and it processes trillions of tokens every week.

Meanwhile, the proportion of OpenAI's internal engineers using Codex has risen from 50% in July to 100%, and the number of weekly merged PRs has increased by 70%.

This means that most of the code written with Codex has passed quality verification and has been actually applied in product development.

OpenAI has also made the Codex model integrable into mainstream development tools such as Cursor, Windsurf, Factory, and GitHub by opening its API, forming a relatively complete ecosystem.

OpenAI is transforming from a "comprehensive" general AI tool to in - depth integration in vertical fields.

Codex targets software developers, and Prism targets scientific researchers. OpenAI's strategic intention is very clear: to vertically capture the high - value and professional user market.

This transformation is not accidental but based on a profound understanding of market demand. Although general - purpose chatbots have a large user base, their user stickiness and willingness to pay are relatively low.

In contrast, users in professional fields are more dependent on tools, have stronger payment capabilities, and are more likely to form a stable business model.

Judging from the product release schedule, OpenAI has clearly planned carefully. Prism was launched at the end of January, just when the new academic semester starts, and many scientific research projects are entering the startup phase.

The Codex release month spans the entire February, which happens to cover the peak period for enterprises to determine their annual budgets and make Q1 technology procurement decisions.

More notably, OpenAI did not adopt a high - price strategy when launching these vertical products.

The purpose of this strategy is clear: first, quickly capture the market through free or low - price offerings to cultivate user habits, and then achieve commercialization through enterprise editions and advanced functions.

Aiming at Anthropic, Engaging in Close Combat

When we take a broader view, we can see that OpenAI's launch of Prism and the strengthening of the Codex ecosystem are strategically aimed at Anthropic's most core competitive advantage: the in - depth intellectual work scenario.

For a long time, there has been a default consensus in the AI circle. ChatGPT is suitable for daily use and casual chatting by the general public, while Claude is more suitable for writing code, reading papers, and conducting scientific research.

Anthropic has firmly captured the hearts of scientific researchers, programmers, and in - depth content creators with its more rigorous logical reasoning and "to - the - point" style.

When many developers share their usage experiences in the community, they often mention that Claude performs more stably when handling complex tasks, and the output content is more accurate, without the "flattery" or "over - elaboration" that often occurs with ChatGPT.

Anthropic launched "Claude for Life Sciences" in October last year, a model specifically optimized for the medical and scientific research fields.

On January 12, Anthropic brought multiple updates to this product line, deeply integrating platforms such as Medidata, ClinicalTrials.gov, and OpenTargets, enabling automated clinical trial operations and regulatory submission processes.

Several pharmaceutical companies and research institutions have begun using Claude to accelerate the drug R & D process. From literature reviews to experimental protocol design and regulatory document preparation, Claude can provide substantial assistance.

Therefore, Prism is almost a direct challenge, and it's as if OpenAI is just short of explicitly announcing that it's targeting Anthropic.

OpenAI doesn't want to lose to Anthropic in the "AI for Science" area, which is the most important growth point in the future.

Although the market size of the scientific research field is not as large as that of consumer - grade applications, its strategic value cannot be underestimated.

Controlling the discourse power in scientific research tools means being able to influence the way knowledge is produced and, in turn, the innovation ability of society as a whole.

More importantly, scientific research users are often early adopters of technology and opinion leaders, and their choices will have a demonstration effect on other user groups.

Anthropic's previous trump card was the Artifacts function, which transformed AI from a "question - answer" chatbot into a "productivity partner" capable of real - time code generation, document preview, and context management.

Users can directly see the code running results during the conversation, modify and debug in real - time, and the entire workflow becomes very smooth.

This experience advantage has helped Claude build a good reputation among developers. Although OpenAI launched Canvas in response, it has not completely changed the situation.

But this time it's different. Prism integrates scientific research writing, data analysis, and reasoning collaboration, directly breaking down Anthropic's advantage in long - term tasks. And there will be several Codex products launched in the next month.

OpenAI wants to use these products to convey this message to high - value users: "What Claude can do, we can do with ChatGPT, and what Claude can't do, we can also do with ChatGPT."

This all - round competitive stance has brought great pressure to Anthropic. In the past, Anthropic could focus on what it was good at and establish advantages in niche markets. But now, OpenAI is launching targeted products in every niche market, and Anthropic's living space is being continuously compressed.

OpenAI's entry has upgraded this war from the "dialog box" to the "workflow."

In the past, the competition mainly focused on model capabilities, such as whose answers were more accurate, smoother, and more in line with user expectations. Now, the competition has shifted to productization capabilities. Whoever can better integrate AI into users' actual work processes will win the market.

This transformation is a challenge for Anthropic because OpenAI has more resources and experience in productization and ecosystem building.

OpenAI is also challenging Anthropic in the area of "rigor," which Anthropic takes pride in.

Anthropic has always adhered to the "HHH" principle, namely Helpful, Honest, Harmless, and has a good reputation in terms of security and reducing hallucinations, which is crucial for scientific research users.

No one wants to cite a literature fabricated by an AI in a paper, nor does anyone want to develop based on incorrect code generated by an AI.

But OpenAI is trying to surpass Claude in "rigor," at least judging from Prism. This is equivalent to attacking Anthropic's brand image of being "the safest and most trustworthy."

In addition, OpenAI is very likely to start a price war with Anthropic.

One of Anthropic's biggest pain points currently is its high cost. The cost of Claude Code is extremely high, and even for heavy users with the most advanced subscription, they often reach the limit in a single day. As a result, they are forced to open multiple accounts to meet their daily needs.

Financing Amount Breaks Records

The capital market is witnessing the most intense financing race in the history of the AI field.

In January 2026, Anthropic significantly increased its latest round of financing target from $10 billion to $20 billion, and the post - financing valuation is expected to reach $350 billion. This valuation is almost double that of $183 billion in September 2025 in just four months.

This round of financing is led by Singapore's sovereign wealth fund GIC and Coatue Management, with participation from a star - studded lineup including Sequoia Capital, Microsoft, and NVIDIA.

Sequoia Capital is also an investor in OpenAI. This "two - horse - riding" investment strategy is not uncommon in Silicon Valley but also reflects the capital's cautious attitude towards the AI track, not wanting to put all eggs in one basket.

Although Anthropic has raised a large amount of funds, it also faces significant survival pressure.

The company has raised its revenue target for 2026 from $15 billion to $18 billion, nearly a four - fold increase compared to the $4.7 billion in revenue in 2025. It expects to achieve about $55 billion in revenue in 2027, a growth rate that is considered crazy in the entire technology industry.

However, Anthropic has postponed its profit - making time from 2027 to 2028, one year later than expected. This means that the company will still need to burn money continuously in the next two years, mainly for Claude 5 model training and the expansion of global computing power infrastructure, which are bottomless pits for burning money.

As the model scale expands and the training data increases, the required computing power grows exponentially, which leads to a rapid increase in the cost of training large - language models.

Anthropic previously stated that it would invest $50 billion in building data centers in the United States. This figure sounds astonishing, but in the current AI race, it may only be the starting price.

Building data centers not only requires a huge amount of capital but also a large amount of power supply and cooling systems. The construction period of these infrastructures often takes several years.

During this process, Anthropic needs to maintain sufficient financial support at all times; otherwise, it may face the risk of a broken capital chain.

Actually, OpenAI's financing scale is even more astonishing.

On January 28, news broke that SoftBank is adding $30 billion in investment to OpenAI, and this is only a phased target; it will invest more in the long run.

Previously, SoftBank invested $41 billion in OpenAI in December 2025.

However, this financing is still far from enough for OpenAI at present. They are seeking a total financing of up to $100 billion, and the valuation may reach $830 billion.

Amazon and NVIDIA are also participating in the discussions for this round of financing.

If this financing scale is successful, it will set a new historical record in the technology industry.

The reason why both Anthropic and OpenAI are expanding their financing is simple: both companies hope to go public in 2026 or 2027.

Anthropic, which was originally on a smooth path, suddenly collides with OpenAI, which will greatly hinder its financing and even the progress of going public because it will directly reduce its market share and expected revenue.

When investors evaluate a company's value, they not only look at its current performance but also its future growth potential.

If the market believes that Anthropic is at a disadvantage in the competition with OpenAI, its valuation will be affected. This impact is not only reflected in the financing valuation but also in the future IPO pricing.

Anthropic has hired the well - known law firm Wilson Sonsini to start IPO preparation work and has had preliminary communications with several investment banks. It may go public as early as the second half of 2026.

Wilson Sonsini is one of the most prestigious law firms in Silicon Valley and has assisted technology giants such as Google and LinkedIn in going public.

There are only two paths ahead for Anthropic.

The first path is to go public before OpenAI occupies these vertical markets to lock in the valuation; otherwise, it may face the risk of a significant valuation shrinkage.

Going public can bring a large amount of capital to the company and provide an exit channel for early investors. However, it also means that the company will face stricter regulatory and information disclosure requirements, and its quarterly financial performance will be scrutinized by the market.

The second path is to defeat OpenAI head - on with its products.

This path is more difficult, but if successful, the rewards will be greater.

Anthropic needs to comprehensively surpass OpenAI in multiple aspects such as product features, user experience, and ecosystem building to gain an advantage in the competition.

This requires a large amount of R & D investment and time, but time is exactly the resource that Anthropic lacks the most. In the rapidly changing AI market, every day of delay may mean a loss of market share.

Judging from the financing rhythm, Anthropic is accelerating. From the $3.5 billion financing in March 2024, to the $13 billion financing in September, to the current financing target of $20 billion, the financing scale is constantly expanding, and the financing intervals are constantly shortening.

This acceleration reflects the company's urgent need for funds and the market's continuous optimism about the AI track.

However, financing is only a means, not an end. Ultimately, the products and technology will decide the winner. No matter how much money is raised, if the products cannot meet user needs, the market share will be lost.

In the current AI race, the technology iteration speed is extremely fast, and today's leading advantage may be overtaken tomorrow. Both Anthropic and OpenAI understand this.

The result of this race will also largely determine the future landscape of the AI industry.

This article is from the WeChat official account "Zimu AI", author: Miao Zheng. Republished by 36Kr with permission.