Financial data companies such as FactSet, Bloomberg, S&P Global and LSEG have largely dispelled fears that AI tools would replace them—only to face a new question from investors: who pays for the tokens once AI agents start consuming far more data than human analysts.
Financial Data · The Token Question
The AI didn't replace the data firms — now the question is who pays for the tokens
Replacement fears eased as verified data proved essential for grounding AI. But autonomous agents devour data on a scale human analysts never did — and the cost is hard to predict.
~1,000×
Tokens an agentic task consumes vs. standard chat
30×
Cost variation across runs of the same task
24×
Goldman Sachs' projected rise in token use over 4 years
Annual token appetite per user
A basic chatbot vs. an advanced "super agent" — the same job, wildly different consumption.
Roughly 38× the load — and one case burned 1 trillion tokens in six months, adding $6M+ in unplanned annual cost.
Today's data seat — before token charges
Bloomberg Terminal
~$27,660/yr
2-year contract
FactSet
~$12,000/yr
per seat
The open debate: will unpredictable token charges be layered on top of these fees?
The case for incumbents
Verified data grounds AI in trustworthy sources
Bloomberg's ASKB praised as transparent and high-quality
FactSet touts agent interoperability for workflows
The open challenges
Token costs are hard to predict
Frontier models underestimate their own usage
Rebuilding data is costly once verification is priced in
The Bottom Line
Data firms are repositioning as the foundation for AI, not its casualty.
With standards efforts like FDX now targeting agentic AI, folding token economics into pricing models is set to reshape their future revenue structure.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…