Google TPU chips and Gemini 3 Pro models began running autonomous cryptocurrency trading systems across global exchanges in February 2025. BitMart and nof1.ai deployed AI-powered trading assistants that execute trades without human intervention, with nof1.ai launching competitions where AI models trade real capital against each other.
Flow Traders, a Dutch market maker operating across European and Asian exchanges, established dedicated deep learning units for algorithmic trading. TPU infrastructure processes millions of data points per second from multiple exchanges simultaneously—a task conventional computing cannot handle at the required speed.
The deployments coincide with divergent regulatory environments worldwide. China reaffirmed its cryptocurrency ban while Bitcoin reached all-time highs in dollar terms. USDT faced regulatory downgrades in multiple jurisdictions, creating volatility that AI systems are designed to exploit.
Gemini 3 Pro's multimodal capabilities allow trading systems to process financial documents, social media sentiment across languages, and price data through unified models. The systems analyze market sentiment and adjust positions in real-time across global time zones without human oversight.
At nof1.ai's autonomous trading competitions, AI models from international participants compete on risk-adjusted returns rather than raw profit. Algorithms run on cloud TPU infrastructure, with performance measured against market benchmarks from multiple regions.
Infrastructure costs create barriers to entry. TPU clusters required for real-time processing across global markets cost hundreds of thousands monthly, accessible only to well-capitalized platforms. This concentrates autonomous trading capabilities among major exchanges and institutional players.
Market microstructure is changing as AI systems interact globally. When multiple autonomous traders across different time zones respond to the same signals simultaneously, they create feedback loops that human traders cannot anticipate. Regulators in the US, EU, and Asia have not yet addressed these dynamics.
The convergence of advanced AI chips, frontier language models, and 24/7 global cryptocurrency markets marks a shift where infrastructure determines competitive position in algorithmic trading.