The infrastructure play in AI isn't just GPUs and cloud compute — it's the data pipes feeding every trading algorithm and hedge fund model in the world.
The Summary
- LSEG CEO David Schwimmer says AI is driving major growth in market data usage, with over 90% of LSEG's dataset being proprietary data that AI models need
- Financial institutions are consuming more market data than ever as they build AI-powered trading and analytics systems
- The shift positions exchanges and data providers as critical infrastructure for the AI economy, not just traditional finance
The Signal
London Stock Exchange Group sits on something every AI trading system needs: proprietary financial data at scale. CEO David Schwimmer told Bloomberg that AI is transforming how financial institutions consume market data, creating a significant growth opportunity for LSEG despite shares taking a hit in 2025. The company's datasets, more than 90% of which are proprietary, are suddenly more valuable because AI agents don't just need raw prices. They need context, history, and structured data to train on.
This isn't about retail traders checking stock tickers. AI-powered hedge funds and algorithmic trading systems consume data differently than humans. They ingest vast historical datasets to train models, require real-time feeds to execute trades, and need structured metadata to make sense of market moves. Traditional data licensing was built for human analysts pulling quarterly reports. AI agents pull everything, continuously.
"Over 90% of LSEG's dataset is proprietary data that AI models need to function."
The timing matters. As Schwimmer noted, LSEG faced a challenging 2025 for shares, but the AI opportunity offsets that decline. Exchanges and data providers have been steady, slow-growth businesses for decades. Now they're infrastructure plays for the agent economy. Every quant fund building an AI trading system needs clean, reliable, high-frequency data. LSEG owns the pipes.
The broader play here is data moats. Exchanges have natural monopolies on certain datasets. If you want real-time order book data from the London Stock Exchange, you buy it from LSEG. As AI agents proliferate across finance, that proprietary data becomes more valuable, not less. This is the opposite of content businesses getting squeezed by AI scraping. Financial data is structured, regulated, and behind paywalls that actually hold.
The Implication
Watch for data licensing models to evolve fast. The old model charged per terminal or per human user. AI agents don't fit that. Expect exchanges and data providers to roll out tiered pricing based on API calls, model training rights, and real-time versus batch access. LSEG's bet is that the shift from human analysts to AI agents multiplies data consumption, not just usage intensity.
For anyone building financial AI agents, the cost of high-quality data is about to become a major line item. Proprietary datasets from exchanges aren't optional. They're the ground truth your models train on. If you're competing with hedge funds that have direct data feeds, your agents are already behind.