Here are five key insights from this week:

1. 1. AI is already here in trading and is revolutionizing markets – XTX Markets, for instance, operates over 25,000 GPUs with 650 petabytes of storage, facilitating $250 billion in daily trading across equities, FX, fixed income, and commodities. According to the StateofAI report that’s up with the likes of Meta, Tesla and DeepSeek

      Read more: stateof.ai

   2. AI-Powered Trading Tools – Market Participants AdaptAccording to Reuters, at least 20 Chinese brokers and fund managers have already integrated DeepSeek’s AI models to improve research, risk management, and client interactions, reflecting a broader industry shift towards AI-driven financial services particularly in the retail space – Tiger Brokers has incorporated the DeepSeek-R1 AI into its TigerGPT chatbot to enhance stock research, sentiment analysis, and earnings summaries. UBS projects a 24% rise in financial IT spending over the next five years due to AI adoption, while Goldman Sachs estimates AI could attract $200 billion in investments.

      Read more: DeepSeek | Tiger Brokers

   3. AI-Driven Index FundsIndex funds are now dynamically adjusted using real-time data, leveraging AI to enhance performance. Qraft Technologies integrates machine learning into ETFs such as AMOM and QRFT, optimizing investment strategies. Indxx designs AI-driven indices tailored to sectors like robotics and big data. Meanwhile, The Voleon Group employs AI for predictive trading, refining asset allocations to maximize returns.

      Data Point: Pictet’s AI-driven fund delivered a 17.3% return over six months, outperforming MSCI World’s 12%.

      Read more:

      Qraft Technologies
      Indxx
      The Voleon GroupSource: The Times

   4. The ability to Fast-Track Research into an EMS?FactSet and Perplexity AI have partnered to integrate FactSet’s financial data into Perplexity’s Enterprise Pro platform, enhancing AI-driven market research and decision-making. FactSet also owns Portware, a leading execution management system (EMS), suggesting potential future integrations where AI-powered insights from Perplexity could enhance trading strategies and execution analytics within institutional trading environments.

      Read more: perplexity.ai

   5. Future of AI Models – Less Data, Smarter Learning?AI model training is evolving. Nathan Benaich’s article on PRIME explores a new method for training Process Reward Models (PRMs) that moves away from traditional data-intensive approaches. Instead of relying on large, annotated datasets, PRIME focuses on continuous feedback at each reasoning step, solving the sparse rewards problem in reinforcement learning. This could lead to more efficient and adaptable AI models.

      Read more: PRIME on arXiv | Nathan Benaich’s Analysis