DeepSeek raises prices on its AI models: the era of cheap artificial intelligence is ending

ER10 reported DeepSeek's decision to raise prices on its AI models. A symbolic moment: it was this Chinese developer whose ultra-low tariffs forced the whole market to revise price lists, and now it too admits — time to earn. The reasons are industry-wide: costlier compute, energy scarcity for data centers, and investors demanding payback instead of growth at any cost.

For the market it signals the end of the subsidized era. For years AI services sold below cost — vendors were buying market share. Now tariffs are turning upward across the industry: through direct increases here, through limits and paid "premium tiers" there. The cheap token was a marketing campaign, not a law of nature.

For companies in Kazakhstan already building processes on AI, the conclusions are concrete. First: recalculate the unit economics of your scenarios — what paid off at old prices may not at new ones, especially mass cheap operations like processing every incoming email. Second: budget for AI service price growth annually, the way you budget for rent increases. Third: avoid binding to a single vendor — the ability to switch between models is turning from a technical whim into a negotiating position on price.