An AI price war could steal the thunder from pending Anthropic, OpenAI IPOs

The high price of high-end artificial intelligence could mean a price war is coming right before two more gigantic initial public offerings. Enthusiasm for the IPOs of Anthropic and OpenAI, expected later this year, could get zapped if they start trying to undercut each other on price in order to hold on to customers.

After the successful IPO of SpaceX earlier this month at a valuation of $1.75 trillion, the Anthropic and OpenAI debuts as public companies are the next milestones in the economy's shift toward AI and automation. So investors are keeping close tabs on the pricing dynamics between the two frontier models.

"OpenAI has been talking about lowering prices. Particularly as [OpenAI and Anthropic] ramp up for IPOs, if that triggers at least a temporary battle on price, then that might change," how their revenues work, Paul Meeks, head of technology research at Freedom Capital Markets, told CNBC earlier this month.

The long-time analyst and investor still sees another two to three years, "well into 2028," for enormous AI infrastructure spending. Anthropic announced a $65 billion capital raise in May, giving the company a $965 billion valuation, and OpenAI closed a $122 billion funding round in March at a valuation of $852 billion.

Both companies have filed documents with the Securities and Exchange Commission to start the IPO process. For now, investors are looking to dodge any potential AI price war by sticking to hardware makers and service suppliers.

Investors are sounding more confident about investments that are further upstream in the AI supply chain, like cloud services, infrastructure and chips. AI 'supercars' or 'workhorses' OpenAI and Anthropic are widely recognized to have the most advanced AI algorithms.

But other cheaper models are looming over the sector too. The cheaper models don't have all the bells and whistles of the high-end, but researchers say they're good enough for most jobs at a fraction of the price.

As AI moves from a cutting-edge technology to a commodity, the popularity of lower cost options could weigh on nosebleed valuations. "For the bulk of everyday tasks (perhaps 90% of them) [China's DeepSeek's V4-Pro] does much the same job at roughly 1.5% of the cost" of Anthropic's leading-edge Claude Fable 5 AI model, Deutsche Bank's Jim Reid wrote in a note to clients last week.

According to an index from analytics website Artificial Analysis , cheaper AI models from DeepSeek, ChatGPT, MiMo and MiniMax all have a weighted average cost below $0.13 per task while the top three versions of Anthropic's Claude cost between $1.80 and $2.75. "Adopters who only need a reliable workhorse – not a supercar – will increasingly ask whether the frontier premium is worth paying," Reid wrote.

Price sensitivity Indeed, AI customers are becoming increasingly sensitive to price. Uber's chief technology officer Praveen Neppalli Naga said in April that the company had blown through its AI budget for all of 2026 in just the first four months.

The company's operations chief Andrew McDonald said last month it was becoming "harder to justify" spending on AI. Microsoft restricted large language model access for employees and canceled Claude Code licenses to restrain costs, according to Forbes .

Executives at Meta Platforms, Salesforce and DoorDash have also scaled back access to AI or discussed the need to make sure it's used productively, the Wall Street Journal reported in May. Chips and infrastructure Ahead of a possible price war among the frontier AI models, the safer AI play for individual investors may simply be found in hardware makers.

Regardless of competition at the algorithmic level, demand for AI is surging overall and unit costs are coming down as more people use AI. That means long term investments in chipmakers and providers of computing power could prove to be more reliable short term.

"Hyperscalers remain remarkably profitable, with early returns on investment positive. We expect operating cash flow from the hyperscaler cohort to be over $900 billion in 2027," Tarek Hamid, a senior analyst at JP Morgan, wrote to clients last week.

JPMorgan expects more than $3 trillion in financings for various kinds of AI-specific chips over the next five years. OpenAI concern Beyond pricing, tech investors have broader concerns about customer segments for the top AI companies.

Dan Niles, founder of Niles Investment Management, said last week he thinks OpenAI is stuck between Alphabet's dominance in consumer AI and Anthropic's positioning in corporate AI. "Are you going to have five different guys win in AI?

No. I think Google wins in consumer … In corporate, you have Anthropic," he said.

"I think OpenAI is stuck between the two of them. So for me, those are the two winners.

I think OpenAI has got the problem because they're jammed between the other two guys."