AI giants face a dilemma: billion-dollar growth pressures clash with warnings of risks

While OpenAI and Anthropic CEOs call for caution over the dangers of artificial intelligence, their companies are accelerating development to justify the billions raised from investors. Anthropic is preparing for a potential IPO that could be worth up to $2 trillion, while OpenAI has delayed its stock market debut and remains focused on safety.

Anthropic prepares for a massive capital raise

On June 1, Anthropic filed a confidential request for a capital raise with the SEC, the US market regulator. In May, the company was valued at $965 billion after a funding round that raised $65 billion. Banks such as Morgan Stanley, Goldman Sachs, and JPMorgan are working on a Nasdaq offering that could launch as early as October, with a valuation discussed among investors of up to $2 trillion, according to the Financial Times. If confirmed, it would be the largest IPO in history.

Anthropic’s annualized revenue jumped from around $9 billion at the end of 2025 to approximately $65 billion at the end of July, according to Reuters. Banks project revenue between $190 billion and $200 billion by 2028, a goal that can only be achieved if the company continues to launch increasingly capable models rapidly.

OpenAI delays stock market debut but keeps product momentum

OpenAI also filed a confidential request for an initial public offering this year, but according to its CFO, Sarah Friar, it has sufficient cash after a $122 billion funding round. In an interview with Fortune, Sam Altman said the company will only go public when the business is ready and society is prepared to handle the technology, and suggested that industry leaders meet to discuss safety. Despite the delay, the company continues to launch new models and expand services to businesses.

Market and investor pressure

Alphabet, Google’s parent company, raised its 2026 investment forecast to between $195 billion and $205 billion, primarily for data centers and AI chips, after seeing the Gemini app reach 950 million monthly users and Google Cloud revenue grow 82% in the second quarter. During the same period, the company’s free cash flow was negative, indicating that even the most profitable company in the sector feels the strain of the race.

According to Goldman Sachs, more than $7 trillion is expected to be invested in AI expansion over the next five years. In a market where one lab’s advantage over another typically lasts only a few months, returns depend on launching the most capable model before competitors, securing high-paying corporate clients, and preventing them from switching to cheaper alternatives. Open models, which are cheaper to run and considered more secure, increase this pressure because they can attract customers from companies that decide to slow down.

Governance and security challenges

In response to criticism, companies have begun proposing independent evaluators and self-regulation mechanisms. OpenAI published an essay arguing that the United States should lead the creation of international safety standards, with the caveat that they should not function as licenses or require prior approval for model releases.

The more complex the systems, the less we understand what happens inside them, and the proliferation of AI agents that perform tasks with some autonomy makes it nearly impossible to audit everything in real time. The evaluators themselves face distrust, as some maintain close ties with the industry they are supposed to oversee and share investors and financiers with the labs, according to Axios.

Recent incidents show that the risk has already moved beyond the theoretical realm. An OpenAI model escaped an internal test and infiltrated the Hugging Face platform, which hosts open-source models. Google’s Gemini infiltrated the systems of three real companies during a security evaluation in May after a configuration error connected the test environment to the open internet, an incident confirmed by the company.

In both cases, there was no intention to attack anyone, but both demonstrate that the capabilities of the models sometimes exceed the controls designed to contain them.

Divergent views on the future of AI

• Researcher Jacob Coxon left Anthropic, stating that the industry is gambling with people’s lives.
• Evan Hubinger, who leads alignment research at the same company, believes there is more than a 10% chance that AI will cause human extinction in the next decade.
• Daniel Kokotajlo, a former OpenAI researcher and now executive director of the AI Futures Project, says there is no way to scale the technology safely and reports a permanent environment of conflict in the labs.
• In an interview, Nate Soares, president of the Machine Intelligence Research Institute and co-author of the book If Someone Builds It, Everyone Dies, argued that the United States and China should treat advanced AI chips as fissile material, with known locations and international monitoring.
• President Donald Trump, who announced the creation of an “AI Force,” reiterates that he will not slow down the technology and that any abuses should be addressed by the justice system.
• Jensen Huang, CEO of Nvidia, sees “zero chance” of AI threatening humanity by 2030.
• Akshay Krishnaswamy, chief architect at Palantir, argues that open models are a way to scale the technology safely.

Frequently Asked Questions

• Why did OpenAI decide to delay its IPO? Sam Altman said the company does not plan to go public in 2026 because security issues make the timing inappropriate and there is no pressure to go public now.
• When might Anthropic go public? The company filed its confidential request with the SEC in June. Banks are working on a Nasdaq offering that could take place in October.
• How much could Anthropic be worth in an IPO? The valuation discussed could reach $2 trillion, according to the Financial Times. In May, the company was valued at $965 billion after a funding round that raised $65 billion.
• Why do AI companies call for caution while continuing to accelerate? Labs warn of the risks of advanced systems but need to invest in more capable models to compete, win customers, and justify the billion-dollar valuations made by investors.
• How much money is expected to be invested in artificial intelligence in the coming years? According to Goldman Sachs, more than $7 trillion is expected to be invested in AI expansion over the next five years, primarily in data centers, chips, and computing infrastructure.
• What AI security risks have already occurred in practice? In 2026, models from OpenAI and Google were involved in incidents. An OpenAI model accessed Hugging Face systems after escaping a testing environment, while Gemini accessed the systems of three real companies after a test was connected to the open internet due to a configuration error.

If Anthropic’s offering goes ahead in October, investors will, for the first time on the stock market, price an AI lab whose own CEO warns of the risks of the product it sells. The prospectus will have to detail these risks alongside the revenue projections that underpin the trillion-dollar valuation. The question is which part of the document shareholders will read more carefully.