OpenAI slashes prices 75% as AI price war heats up in July

OpenAI, Meta, and xAI have slashed model prices by as much as 75% in 10 days, while Anthropic — the most expensive provider — is on track to post its first quarterly profit.

A wave of price cuts sweeping across the AI industry has exposed a fundamental divide: companies burning cash to win market share versus a profitable rival that refuses to compete on token pricing.

"Every enterprise now is thinking about spend and the value they're getting in exchange for AI," OpenAI CEO Sam Altman told CNBC on Thursday.

Between June 29 and July 9, Anthropic, xAI, OpenAI, and Meta released new models — Sonnet 5, Grok 4.5, GPT-5.6, and Muse Spark 1.1 — each priced below the last. Meta's Muse Spark 1.1 costs $1.25 per million input tokens and $4.25 per million output tokens, undercutting Grok 4.5 at $2 and $6 and OpenAI's GPT-5.6 Terra at $2.50 and $15. Anthropic's Claude Opus 4.8 remains the most expensive at $5 and $25. On July 15, Altman went further, posting on X that OpenAI's Sol model is already half the price of Anthropic's Fable 5 and that he is "happy to deliver at a quarter of the price" — a 75% cut from current levels.

The divergence in strategy reflects a deeper structural gap. OpenAI posted a $9.3 billion operating loss in Q1 2026, losing $1.60 for every dollar of revenue. Anthropic, by contrast, expects Q2 revenue of $10.9 billion — more than double Q1's $4.8 billion — and its first operating profit of $559 million. The company's path to profitability rests on an enterprise customer base that values security and compliance over token pricing. About 85% of Anthropic's revenue comes from enterprise clients, with more than 1,000 customers paying over $1 million annually. OpenAI generates more than 60% of its revenue from consumer subscriptions, a price-sensitive base that responds directly to rate changes.

The Cost Advantage That Isn't

Altman has framed OpenAI's price cuts as a function of genuine efficiency gains, calling a 50% reduction in inference costs "core intellectual property" that the company guards as a trade secret. The narrative mirrors Amazon Web Services' strategy of passing efficiency savings to customers — AWS cut prices more than 100 times between 2006 and 2018 while remaining profitable throughout.

But the comparison breaks down on OpenAI's income statement. AWS never posted a $1.60 operating loss per dollar of revenue while cutting prices. OpenAI's $122 billion March funding round and $73 billion cash balance are funding the gap, but the company also carries $665 billion in committed compute spending through 2030 — obligations that must be paid regardless of demand.

"The pricing from some of the other labs is very extreme and has very high margins," Meta CEO Mark Zuckerberg said in an interview, announcing Muse Spark 1.1 at $1.25 per million input tokens. Meta's advertising business subsidizes its AI push, giving it more room to be aggressive.

Anthropic's Bet on Enterprise Stickiness

Anthropic has not matched the price cuts. Its Sonnet 5 limited-time offer at $2 and $10 per million tokens expires Aug. 31, after which pricing reverts to $3 and $15. The company's leadership has publicly urged enterprises not to reduce AI usage because of cost concerns — a message that assumes customers will pay a premium for reliability and safety.

The bet will face its first test in October, when Anthropic targets an IPO at a valuation of about $965 billion, according to people familiar with the matter. If enterprise customers renew contracts at stable pricing through Q3, it would validate the strategy. If a Fortune 500 client defects to OpenAI or Meta on price, the premium narrative weakens.

OpenAI's IPO has been pushed to 2027 at the recommendation of advisers, who warned that going public sooner would require accepting a lower valuation. Altman has rejected any adjustment to the company's $1 trillion valuation target, according to people familiar with the discussions.

For investors, the price war creates a clear framework. Anthropic's profitability and IPO timeline offer a benchmark for the sector's economics. OpenAI's $9.3 billion quarterly loss and $665 billion in committed compute spending represent the downside risk if demand growth slows. Meta's willingness to sell below cost, backed by its advertising cash flow, adds further pressure on standalone model providers. The next milestone is Aug. 31: if Anthropic lets its Sonnet 5 discount expire without extending it, the company will have signaled that it sees no need to compete on price.

This article is for informational purposes only and does not constitute investment advice.