Getting your
Trinity Audioplayer ready...If Bay Area artificial intelligence companies OpenAI and Anthropic go public at eye-popping valuations in coming months, many Americans could soon own stakes in the firms without ever choosing to buy their stock.
That offers the prospect of a substantial reward, as investors bank on healthy returns from a technology that has taken the world by storm since OpenAI released its pioneering ChatGPT less than four years ago.
But it also poses risk if retirement portfolios become increasingly tied to companies whose trillion-dollar valuations depend on AI living up to its hype. Critics point to warning signs: OpenAI and Anthropic both anticipate going public at valuations approaching $1 trillion, yet the San Francisco companies remain unprofitable and are spending billions of dollars a year on computer processing.
Major AI players have complicated, sometimes-circular financial arrangements with each other, many of them built on debt.
When companies go public, their stock becomes available for purchase in the stock market. People often end up holding a particular company’s stock because their retirement account — such as a 401(k) or an IRA — includes index funds that may add newly public companies after a waiting period or once they meet eligibility rules. Many pension funds also feature stock index holdings.
SpaceX, which includes AI company xAI, went public June 12 and as of this week was valued at $2.5 trillion despite losing $5 billion last year. Anthropic and OpenAI have each confidentially submitted IPO paperwork to the SEC, though neither has set a public offering date. Many analysts expect their IPOs could occur this year.
Meanwhile, the future utility and profitability of generative AI technology — which produces writing, imagery, computer code or audio in response to user prompts — is unclear.
Anthropic and OpenAI, said Omri Even-Tov, a professor of accounting at UC Berkeley’s Haas business school, are “priced for a future that hasn’t happened yet.”
But the future looks bright, according to Wedbush Securities analyst Dan Ives, who in a note to investors Tuesday hailed SpaceX’s “solid” public offering as “a good sign for OpenAI and Anthropic as both companies likely head down the IPO path.”
Public-market money flowing into the AI giants would fuel the next phase of AI growth, Ives said.
“As tech stalwarts like SpaceX, OpenAI and Anthropic get more capital and go public, this will just further drive more investments and (capital expenditure) into the AI Revolution flywheel,” Ives said.
For regular people with retirement accounts, the uncertainty over AI’s prospects “puts everybody between a rock and a hard place,” said Paulo Carvão, a senior fellow at the Harvard Kennedy School who is studying AI regulation in the U.S. and the technology’s societal effects. “The opportunity is that this is transformational technology and has all the characteristics of what economists call ‘general-purpose technologies’ that diffuse quickly across the economy,” Carvão said.
Should AI become general purpose, ordinary Americans, through retirement accounts and other investments, could be “investing in what is foundational for the future.”
But excessive concentration of a person’s investments in a few companies poses risk. Many major index funds base stock purchases on companies’ stock market valuations, known as market caps. That means the biggest companies take up the largest share of investors’ portfolios — and if OpenAI, Anthropic and SpaceX enter indexes at enormous valuations, retirement savers could end up with more exposure to them than they might expect.
Should the companies’ stock price fall, or if AI’s promises don’t materialize, retirement funds could take a hit.
“All of a sudden, you have a lot of your savings concentrated in a small number of very large-cap stocks,” Carvão said. “Diversification is the best way to avoid risk — this is a movement in the opposite direction.”
Another fundamental risk, Carvão said, is whether demand for AI can grow fast enough to cover the enormous cost of developing and operating “frontier models” — the huge AI systems that power products such as chatbots, coding assistants and image generators — in power-hungry data centers, and turn a profit.
J.P. Morgan on Tuesday projected global AI capital spending of $5.5 trillion through 2030, with $4.1 trillion of it financed by debt.
For Anthropic and OpenAI, “radical change in revenue generation will have to happen so that you can sustain this level of advancements,” Carvão said. Anthropic has focused on selling to businesses, and while OpenAI is also seeking to boost its presence in that “enterprise” market, it’s largely focused on consumers, he said.
“It seems that Anthropic has a better path to profitability,” Carvão said, “and it’s operating in a market that tends to have more stable demand, which is the enterprise market.”
For businesses, AI has the potential to be as revolutionary as the personal computer and the internet. “These technologies start to find their ways into business processes, and once they’re embedded in the business processes, they’re here to stay,” Carvão said.
Still, a high-profile AI mishap in a prominent company could spook businesses and slow adoption, or government actions could take an enterprise AI product off the market, he said.
Consumer-focused OpenAI “seems to be extremely challenged,” Carvão said. “They’re operating not only at a loss but with negative cash flow.”
The consumer market is more price-sensitive than the enterprise side, he said, adding that competition from AI companies like China’s DeepSeek adds to the pressure.
However, AI is “starting to be transformative” among consumers, for example, by starting to replace internet search.
“We need to see how some of the potential harms or problems are going to be mitigated, from hallucinations all the way to the harms like mental health,” Carvão said. Looming regulations could create “guardrails that are beneficial for consumer adoption,” he added.
Prominent AI industry critic Ed Zitron, owner of San Francisco public relations firm EZPR, argues ordinary folks, through their retirement funds, “will be unwittingly fueling these companies that are bad.”
Zitron said he believes fundamental and probably insurmountable obstacles confront Anthropic and OpenAI.
Anthropic, OpenAI and SpaceX did not respond to requests for comment.
“The people, as ever, who will suffer from these things will be regular people, rather than people who made this large con inflate,” Zitron said. “It is morally reprehensible that these companies are being allowed to go public. I think SpaceX is probably the best of them, but only because of how bad OpenAI and Anthropic are.”
It takes up to three years to build a data center, thanks to steel and labor shortages, growing localized opposition and slow provision of electrical power infrastructure to run them, Zitron said. But AI companies are buying and storing pricey AI-processing chips to stock data centers once they open. With new, cutting-edge chips released every year, he said, the stockpiled processors risk becoming obsolete before they can become operational.
Those infrastructure costs matter to retirement savers because they shape whether the companies can grow into their valuations.
The industry’s torturous financing raises alarms for Zitron. As an example, he cites Nvidia’s investment in data center company CoreWeave, which buys Nvidia chips and then uses them as collateral to borrow money to buy more chips. CoreWeave then gets AI processing contracts with companies like OpenAI and Meta, and takes those contracts to banks to borrow money to buy more chips.
CoreWeave, of New Jersey, told this news organization Wednesday that references to “circular financing” don’t “stack up with the facts.” Its deal with Nvidia “is a strategic investment in the broader AI ecosystem, helping accelerate the buildout of infrastructure that customers are already demanding.”
In September, OpenAI and Nvidia announced a partnership under which Nvidia said it intended to invest up to $100 billion in OpenAI progressively as new AI infrastructure is deployed.
“It’s companies effectively handing money to other companies so they can use it to buy their stuff,” Zitron said. “You don’t have to do this if you have a real industry.”
Close financial ties among AI players — even among rivals — highlight the risk to Americans’ retirement funds should AI falter, Zitron said. For example, Google, which markets its Gemini chatbot to consumers and businesses and has spent hundreds of billions of dollars on AI, said in April it would invest up to $40 billion in its competitor Anthropic, according to news reports.
Pricing of AI products also sends up a red flag, Zitron said. After OpenAI and Anthropic moved business customers to usage-based pricing earlier this year, “instead of getting this all-you-can-eat Bacchanalia, you had to pay for everything you used,” Zitron said.
Uber in June capped employees’ use of AI tools in the face of burgeoning costs, Bloomberg reported. Last month, Uber chief operating officer Andrew Macdonald said on the Rapid Response podcast that a link between the company’s use of AI and productivity gains was “not there yet.”
Americans wary of having their retirement savings dependent on the performance of AI companies going public have more to consider, Berkeley’s Even-Tov said. Many index funds invest heavily in companies that have gone all in on AI, like Nvidia, Google, Meta, Microsoft and Amazon, with such firms making up more than 40% of the S&P 500 index, Even-Tov said.
Strong stock market performance by those companies in recent years has padded retirement savings for many Americans.
But, Even-Tov said, “just buying the market isn’t as diversified as it sounds.”
If AI is a bubble that pops, holding stock for the long term instead of panic-selling makes sense, he said. “We’ll see indexes go back to what they were, and go higher,” Even-Tov said.
Even if AI proves transformative, the rise of giant AI companies may force ordinary investors to reconsider how diversified their “diversified” retirement accounts really are.