Has AI product pricing come full circle back to where it started?

A wave of technology companies are rolling out AI discounts and free perks. Amazon.com (AMZN) announced that platform sellers can use its Quick Plus AI assistant for free for a full year. Meanwhile, Microsoft (MSFT) is slashing subscription prices for its Copilot AI dramatically. The promotional moves by these two companies are just part of a broader wave of price cuts sweeping across software vendors and AI labs including OpenAI.

We keep hearing tech executives tout the transformative power of AI. But it is clear that enterprise customers are not willing to pay full price. And building AI infrastructure at scale is enormously expensive, creating a real contradiction that must be resolved somehow. Torsten Slok, chief economist at Apollo Global Management, noted in his daily research email on Monday that credit investors are now buying bonds from hyperscale cloud providers on the assumption that these companies' operating cash flows will triple between 2025 and 2030. Slok wrote: "If that expectation falls short, the AI investment boom will cool, credit spreads will widen, and capital expenditure plans will be cut."

The current pricing environment is reminiscent of the early days of the AI revolution two or three years ago, when AI companies desperately tried to persuade users to try their products, with revenue taking a back seat. OpenAI, Anthropic, and Google once cut prices for their flagship large models by more than 90%. But last year the tide turned, with companies placing greater emphasis on recouping their investment costs. Enterprise software vendors followed suit, shifting from charging per account to billing based on actual usage. The previous per-seat pricing model had become loss-making after employees' AI usage surged. But as related reports noted, customers then complained that usage-based billing actually resulted in higher bills, so the industry once again returned to a phase of aggressive price wars.

The same phenomenon is appearing in the consumer market. Former Apple and Nest executive Tony Fadell said on social platform X on Wednesday: "What consumers want from AI is something practical, easy to use, and preferably free, or bundled into services they already pay for." Of course, some of the current discounts serve a different purpose, hoping to drive total customer spending through other means. Take Microsoft (MSFT) as an example: it still retains per-seat subscriptions, layered with on-demand usage-based billing. Now it is lowering seat fees, betting that after enterprises purchase more accounts, the bills generated by usage will lift overall revenue. It is not hard to imagine that customers will eventually see through this model.

One way out of the current predicament is to eliminate a batch of companies that have rushed into the AI race. Today, established software companies, startups, and AI vendors are all making homogeneous products. Publicly listed traditional software companies must deliver revenue growth, even if at a low rate, but that goal may not be realistic. What enterprise customers likely want is for AI to replace some of their existing software spending, rather than seeing their total bills grow ever larger. An industry shakeout is coming. Continuously cutting prices on costly AI products cannot be sustained for long.