AI Pushes Trading Into Thinner Markets
Cheaper research → more trades. Plus: runaway AI-agent costs and the race to trade GPU capacity.
Hey, I’m Matt. I’m a former Bloomberg News reporter, and you’re reading AI Street, where I report on how Wall Street uses AI.
AI to Push Trading Into Thin Markets
I have a running theory that AI is going to push more trading into markets that have historically been thinly traded.
As we’ve covered a lot around here, AI is making investment research cheaper. Investors can track more companies and test more ideas. Tasks that used to require more people, time and money are getting easier to run at scale. I’ve heard versions of this in multiple conversations with investors using AI.
This Bloomberg story has more reporting on this trend:
Advances in artificial intelligence are leveling the field for fund managers, making it easier for boutique firms to compete with big macro and bond investors.
From digesting speeches in multiple languages and crunching global inflation numbers, to tracking company filings and the tone of investment committee discussions, AI is picking up much of the work once carried out by teams of analysts, according to five executives who have recently set up their own shops.
“Technology has changed the economics of building an investment firm,” said Dharmesh Maniyar, a machine-learning PhD who founded his second fund, MQT Asset Management, late last year.
Now if everyone is getting access to cheaper investment tools, this should lead to more trading. And recent research supports this trend.
When Italy banned ChatGPT in March 2023, retail investors made fewer trades in unfamiliar assets, shifted back toward popular names, and held portfolios that looked more concentrated and alike.
And AI-driven portfolios trade much more often than their human active-manager counterparts. According to a 2024 IMF report, AI-driven ETFs already turn over holdings about once a month, versus less than once a year for typical active equity ETFs.
Volume has been above historical norms since COVID. The pandemic reset U.S. stock-trading volume higher. A quiet pre-2020 day was closer to 7 billion shares. Since then, average daily volume has generally stayed above 10 billion shares, rising to 12.2 billion in 2024, according to the Cboe.
TL;DR
I’m having a hard time seeing trading not marching higher and permeating into more markets given current AI trading trends. It didn’t make a lot of economic sense to spend limited human time researching smaller markets. Now it does.
The Cost of Runaway AI Agents
One caveat about the BBG article above: I think it underplays how hard it is to get AI working accurately at scale. It’s one thing to summarize a central bank speech, but quite another to pull accurate numbers out of hundreds of pages of regulatory docs. This is also a topic we’ve discussed a lot.
And even when you can receive accurate, helpful responses, AI research isn’t cheap. I’ve heard a lot recently about the exploding costs of compute from financial services firms. Uber reportedly blew through its annual AI coding-tools budget in a few months.
I’ve had the same experience trying to create my own agents with not a lot success.
Milos Maricic, a consultant who advises sovereign wealth funds on AI, said he had a hedge fund client spending roughly $24 million a year on LLMs, not realizing that about a third of that spend was just wasteful.
One of the fund’s agents kept calling a third-party sourced that went dark, but the agent kept getting timeouts or 429 errors and treating them as a reason to try again. It retried for weeks.
Maricic said his team cut those costs by a third without hurting research quality, mostly by tracing spend to specific teams and workflows, stopping runaway agents and caching repeated prompts.
As I mentioned last week, I think managing compute expenses, or maybe we can call it something fun like “token master,” is an emerging job.
Computing costs are exploding across companies. Who manages that? Each individual department head? How do you decide which use cases to spend those resources on? Those seem like hard questions to me.
The Growing Market for Making Markets in Compute
A year ago, trading “compute” sounded weird. `Like, you’re going to turn GPU capacity into something you can trade, like oil or electricity?’
Today not so weird. Multiple firms are trying to build the financial infrastructure of the emerging commodity. DRW’s Don Wilson is backing Silicon Data’s GPU pricing index, which feeds CME’s planned futures market.
This week, a16z announced its first bet on this emerging asset class by leading a $33 million seed round for Ornn, which publishes a rival index that powers ICE's planned futures market.
As a16z’s Ali Yahya put it to Bloomberg: “This asset class is very immature at the moment, but will professionalize over time.”
This Week in AI Street
AI Finds What Markets Miss in News: Study
A new paper takes that same framework — embeddings of text — and points it at financial news.
The researchers, including AQR’s Bryan Kelly, turn news articles into embeddings and test whether those representations contain information the market has not fully priced.
...
Arman Khaledian, PhD, a former quant at Millennium and now CEO of Zanista AI, put it this way in an email: “It’s like a factor model, but instead of prices you’re running it on vectorised news, stripping out the predictable part to see what’s actually moving things.”
ROUNDUP
What Else I’m Reading
Secretive Wall Street Powerhouse Jane Street Seizes the AI Spotlight | WSJ
XTX Markets looks for AI gems after striking gold with Anthropic | FN
Alphabet Shares Drop After Second AI Star Departs for Rival | BBG
AI Can Model but Can’t Make the Next Rainmaker | WSJ
Starling rolls out AI-powered romance scam detection feature | Finextra
CIOs managing billions in assets read AI Street.
Paid subscribers access original reporting, data analysis, and interviews with the executives and researchers leading Wall Street’s AI buildout.
SPONSORSHIPS
Reach Wall Street’s AI Decision-Makers
AI Street reaches institutional investors, C-suite executives, and Big Law attorneys at firms including JPMorgan, Citadel, BlackRock, Skadden, McKinsey, and more. Sponsorships are reserved for companies in AI, markets, and finance. Email sponsors@ai-street.co for more details.
CALENDAR
Upcoming AI + Finance Conferences
32nd Computing in Economics and Finance – June 29 • Venice
Academic conference on computational methods in economics/finance (numerical methods, simulation, macro/asset pricing, econometrics).
Risk Live Europe 2026 – June 29–July 1 • London
Risk.net conference for banks, buy-side firms and regulators, with sessions on AI, automation, model risk and operational risk.
The Fin AI Symposium 2026 – July 3 • San Diego
Industry symposium focused on applied AI in finance, with practitioners sharing tooling, workflows, and deployment lessons.
ARPM Quant Bootcamp – July 13–16 • New York
Four-day quant workshop from ARPM, focused on portfolio construction, risk, machine learning and the plumbing behind systematic investing.
Artificial Intelligence for Financial Services – July 23–24 • Cambridge, MA
MIT executive-education workshop on AI/ML applications in financial services, emphasizing strategy, use cases, and implementation considerations.
Thanks for reading!
I’m always happy to receive comments, questions, and feedback.
connect with me on LinkedIn, or
send an email to matt [at] ai-street.co