The AI industry is facing a major financial crisis as dominant providers like OpenAI operate at massive losses, creating an unsustainable ecosystem where many dependent startups risk collapse when AI token prices inevitably rise. The future favors companies that gain control by adopting open models and building their own infrastructure, while those relying on artificially low-priced tokens from dominant providers will likely fail.
The AI industry is facing a significant financial crisis, described as the biggest “rug pull” in tech history. Despite expectations that AI token prices would fall due to technological advances, companies like OpenAI are losing more than a dollar for every dollar earned, with projected losses reaching $115 billion by 2029 and profitability not expected until around 2030. Venture capital has artificially suppressed prices, keeping startups dependent on costly APIs owned by a few dominant providers. This creates a precarious situation where either the AI startups go bankrupt or their customers face steep price hikes.
Many AI startups, often called “wrappers,” build their businesses on top of models they do not own, relying heavily on the pricing set by the underlying model providers. Since these providers are themselves losing money, the current low prices are unsustainable. When prices inevitably rise, thousands of startups risk collapse as they cannot quickly adapt or switch providers due to “hostage contracts” that lock in data and models on external servers. This dependency creates a fragile ecosystem vulnerable to sudden and severe disruptions.
The business model underpinning much of the AI boom is a classic loss leader strategy: companies sell below cost to gain market dominance, hoping scale will eventually bring profits. However, unlike traditional software, AI requires massive physical infrastructure with high power consumption and expensive hardware that depreciates rapidly. These costs mean that even with scale, profit margins remain slim. Providers have resorted to tactics like “sherlocking,” where they replicate successful features of startups within their own platforms, and diluting model quality to cut costs, further squeezing the ecosystem.
OpenAI exemplifies the industry’s challenges, having committed hundreds of billions to cloud infrastructure deals while still burning cash. To bridge the gap, OpenAI began selling ads within ChatGPT, a move previously resisted, highlighting the desperation to cover soaring costs. Analysts estimate that OpenAI will need hundreds of billions in fresh funding by 2030, while revenue targets have been repeatedly revised downward. The pressure to show profitability is mounting, especially with an IPO on the horizon, forcing a reckoning with the unsustainable economics of AI token pricing.
The future of AI startups hinges on control and independence. Companies that rely solely on renting models from dominant providers face existential risks as prices rise and terms tighten. In contrast, a growing alternative involves open models like Meta’s Llama, which can be run on private hardware, giving businesses control over costs, data, and model updates. This shift will likely create a divide: firms that treated cheap AI tokens as a permanent subsidy will struggle or fail, while those that used the low-cost period strategically to build their own infrastructure and expertise will survive and thrive. The time to transition is now, as the era of artificially low AI token prices is ending.