Financial Times Article on AI GPU Market Contained "Factor of 8" Math Error, Says Investor Gavin Baker

October 20, 2025 – Prominent investor Gavin Baker, founder of Atreides Management, has publicly highlighted a significant mathematical error in a recent Financial Times article concerning the artificial intelligence (AI) graphics processing unit (GPU) rental market. Baker stated that the error led to calculations being "off by a factor of 8," fundamentally misrepresenting the market's outlook.

According to Baker's social media post on October 19, 2025, the Financial Times piece "writes an article about how current GPU rental prices suggest that AI a bubble, but makes a math error such that all of their calculations are off by a factor of 8 by conflating per server and per GPU economics." He further elaborated that correcting this error revealed a "reasonably bullish" scenario for "subscalers" and "super bullish" prospects for "hyperscalers" in the AI industry.

The investor, known for his focus on technology investments, indicated that the corrected analysis ironically served as "another positive indicator for GPU residual values." This suggests that the underlying demand and value of GPUs, critical components for AI development, might be stronger than the Financial Times article initially portrayed. Baker credited @ShanuMathew93, @zephyr_z9, and @The_Colonel__ for assisting in identifying the discrepancy.

Gavin Baker's firm, Atreides Management, specializes in global technology investments, often taking concentrated positions based on deep fundamental research. His critique underscores the importance of precise data analysis in financial journalism, particularly in rapidly evolving and high-stakes sectors like AI and its infrastructure. The incident highlights how nuanced technical details, such as the distinction between per-server and per-GPU economics, can significantly alter market perceptions and investment theses.