The price of artificial intelligence has dropped to zero over 30 months — an absolute record of technological deflation.
The artificial intelligence market is experiencing a unique phenomenon: the cost of neural network computations has dropped to nearly zero in just 30 months. This is the most rapid price deflation of any technological cycle in history—even the era of personal computers did not see such a pace.
Why Cheap Intelligence Changes the Game
For comparison, PC prices fell by 90% over 15 years. Language models have gone through the same journey in less than three years. In my observation, this deflation is the key driver of all current changes in the software industry. When the cost of a technology collapses, the premium for its scarcity disappears along with it.
The economics of software companies were built on entirely different calculations. Investors valued SaaS businesses based on the assumption that software development would remain expensive. But this calculation is crumbling five times faster than it did in the PC era.
Such aggressive deflation is splitting the market in two. On one hand, it destroys the pricing power of model developers; on the other, it creates enormous benefits for end consumers. Companies that sell intelligence are losing market power, while those that apply it are receiving a gift. Every business just needs to figure out which side of the graph it stands on.
How Much the US and China Spend—and Why the Numbers Are Deceptive
Deflation is accelerating amid record investments. US private investment in AI reached $285.9 billion in 2025, while China's stood at just $12.4 billion. The formal 23-fold gap looks alarming, but it is deceptive.
China calculates expenses differently: the $12.4 billion figure includes only private investments, whereas Chinese AI companies have received about $184 billion through government channels since 2000. Analysts at Societe Generale also doubt the 23-fold gap—China's actual spending, in their view, is much closer to that of the US, but authorities conceal it from conventional measurement methods.
US statistics distort the picture for another reason. The impressive $285.9 billion does not fuel a competitive industry as a whole but instead concentrates among a few companies: the number of mega-rounds over $1 billion jumped from 15 to 28. OpenAI's $40 billion round is the most striking example, while Google has spent over $150 billion on AI infrastructure.
My conclusion: The US and China are pouring hundreds of billions into AI, but the technology itself is becoming cheaper so quickly that the beneficiaries are not the creators but the users. Money accumulates among a few model developers, while cheap intelligence strengthens the rest of the economy—this is the main paradox of the current cycle that we are observing in real time.