By James Calloway

London Stock Exchange Group said Wednesday that years of its Tick History data can now be analyzed in seconds on NVIDIA chips, down from hours. The claim rests on one benchmark: NVIDIA's cuDF software running with the Polars data library. LSEG published no third-party verification of the figure.

What does the setup actually change?

Tick History holds 110 trillion rows covering 100 million instruments from more than 580 exchanges, venues and contributors, with history going back over 30 years. Under the new configuration, LSEG said, customers can run the archive on NVIDIA graphics processors without making another full copy of the data, and existing licenses and access controls stay in place.

That last detail matters for desks weighing storage cost against compliance exposure. Firms want historical data "without the burden and risk of creating another full copy," said Ron Lefferts, who has co-led LSEG's Data & Analytics division since April 2025.

LSEG gave no launch date and no pricing for the GPU setup. Both gaps sit at the center of any buy decision.

Where does demand sit?

Usage of the archive is growing faster than LSEG's data revenue. Customer API requests for Tick History rose 48% in the first half year-on-year. Traffic on LSEG's real-time network grew 29% and peaked at 26 million messages per second in June.

LSEG lists three use cases:

• Strategy testing over longer histories and more venues
• Execution analysis, including order book reconstruction and execution-quality checks
• Risk work, including market replay and AI research

Execution-quality checks carry regulatory weight: brokers use them to demonstrate best execution to regulators. LSEG already sells cross-venue surveillance built on a consolidated orderbook of more than 40 UK and EU venues, launched in January.

Retail brokers are already in the customer base. Capital.com signed up for LSEG news, market data and research feeds in February 2025 to serve more than 750,000 traders.

What is measured and what is asserted?

The seconds-not-hours figure is a benchmark result, not a production metric. The 48% API growth and the 26-million-messages-per-second peak are measured. The claim that full copies become unnecessary is asserted by the vendor and remains untested by customers until the product ships.

Researchers can work "without downsampling away the detail that makes the data valuable," said Ioana Boier, global head of capital markets strategy at NVIDIA. Downsampling — cutting tick data into coarser intervals to ease processing — is standard practice today, and its removal would change how TCA and replay studies are built.

How crowded is the field?

Data & Analytics, the division selling Tick History, grew 5.1% organically in the first half — the slowest of LSEG's four divisions. Markets grew 11.9% and Risk Intelligence 9.7% on the same basis. The GPU push reads partly as an answer to that lag.

Competitors are moving along parallel tracks:

• KX, which merged with tick data provider OneTick in 2025, released two AI blueprints built with NVIDIA in March, including a trading signal agent
• CME Group put its live futures and options feeds on Google Cloud in 2019
• LSEG itself began delivering datasets through Databricks in September 2025 and named Tick History among those to follow

The FCA has also worked with the same chipmaker: it opened a sandbox with NVIDIA in 2025 where financial firms can test AI tools on NVIDIA's computing platform.

What is still missing?

The work sits under LSEG Everywhere, the group's strategy for delivering data into AI tools, which connected its feeds to ChatGPT for licensed users in December.

Until LSEG names a launch date, a price and an independent benchmark, desks can file this one under promising but unpriced.