Hey, I’m Matt. You’re reading AI Street, where I report on how Wall Street uses AI. On Thursdays, I curate the week’s biggest AI-in-finance stories and share my own reporting and analysis.
Can AI Help a Small Team Run a Big Fund?
TL;DR — Wall Street vets start a new AI fund to scale their investing judgment.
Alex Popa, a former portfolio manager at Capital Group, and Charles Elkan, who used to run machine learning at Goldman Sachs, founded the AI investment firm Poesis. This week, the firm said it plans to open its first strategy, AI Clarity, to outside investors in early 2027. Poesis has been trading U.S. large-cap stocks with its own money since August.
AI agents research stocks and propose trades. Popa and his team set the risk limits and approve every trade. The strategy invests in 80 to 100 stocks, and the firm has 10 employees, including data scientists and engineers, per Bloomberg.
Poesis could manage billions of dollars with fewer than 100 people, Bloomberg reported, citing people familiar with the firm’s plans.
Most of what’s happening in finance is that hedge funds are folding AI into their existing investment processes. There aren’t many funds you can invest in that are built around AI from the start. One that comes to mind is Bridgewater’s AIA Macro fund, which the firm runs as a separate strategy. And which I covered in detail in August.
There’s also Man Group’s AlphaGPT, which comes up with investment ideas, writes the code and runs the backtests. Man partner Ziang Fang told me its ideas are on par with those of the firm’s human researchers.
PRESENTED BY PORTRAIT
Find Your Next Best Investment Idea. Stay Ahead of What Matters.
Apply your investment framework across thousands of companies, identify & rank the best-fit opportunities, and continuously track the developments that matter to your investment thesis.
Portrait helps fundamental investors with two powerful AI-driven research workflows:
Idea Generation at Scale: Go beyond traditional stock screens. Portrait combines quantitative filters with qualitative investment criteria, deploying AI research agents to evaluate thousands of companies and rank opportunities based on how closely they fit your investment framework, with source-grounded research behind every result.
Always-On Thesis Monitoring: Continuously track key thesis signposts, KPIs, risks, and catalysts across your portfolio. Portrait identifies developments that strengthen or challenge your investment case, helping analysts stay ahead of material changes without manually tracking every update.
Built for fundamental investors. Designed to scale your investment process.
I’m Moderating at ICAIF in Milan
I’ll be at the ACM International Conference on AI in Finance (ICAIF) at Bocconi University in Milan moderating a panel on private capital and AI investing on Saturday. If you’re going, please come say hi!
Databento Valued at $1.1 Billion
Four months after raising $97 million in a Series B round, financial data startup Databento announced a $100 million Series C today, taking its valuation to $1.1 billion.
Since the last raise, the company’s revenue has more than doubled. Salt Lake City-based Databento counts five of the six largest frontier AI labs as clients. Unlike many legacy data providers, which mostly sell data in bulk, Databento lets investors pick and choose the data they want. For more on the company, check out my July interview with CEO Christina Qi.
Money Manager Says AI Uncovered $22 Billion in Client Assets
TL;DR
The investment firm is paying about $35 million a year for the agents, which helped flag assets its advisers hadn’t known about, including old 401(k)s.
Investment firm Mariner, which is paying about $35 million a year to deploy AI agents, said the digital employees helped uncover $22 billion in client assets in its current book of business. (The firm and its affiliates advise on more than $632 billion in assets.)
To be clear, the money wasn’t lost. It’s money that Mariner’s clients already had and that their advisers didn’t know about. That’s my understanding from this InvestmentNews story, which covered the comments from Mariner CEO Marty Bicknell.
The story doesn’t say exactly how the AI agents found the assets, but Bicknell mentioned 401(k) plans that clients had left behind at previous employers as one example.
This is a great example of AI doing something humans likely wouldn’t do at scale. It can sift through data across different systems to piece together a fuller picture of a client’s finances.
REGULATION
CFTC Delays Decision on CME’s GPU Futures
The regulator pushed back the deadline for its review of proposed compute futures tied to the cost of renting Nvidia H100 and B200 chips, according to a Nov. 4 filing.
SEC, PCAOB Set Up AI Enforcement Units
SEC Enforcement Director David Woodcock said Nov. 10 that the agency’s new AI unit will help investigators analyze documents, trace financial flows and respond to AI-enabled fraud and market manipulation.
The Public Company Accounting Oversight Board announced a unit the same day to investigate potential AI-related misconduct in public-company audits and develop tools for its enforcement staff.
BoE’s Hall Floats Using AI to Block Harmful Trades
Jonathan Hall, an outgoing member of the Bank of England’s Financial Policy Committee, floated a way to police AI trading agents: use a separate model to estimate the harm a proposed trade could cause and block the trade if the risk is too high. For now, Hall urged firms to limit agents’ access, monitor their activity and assign them unique IDs so trades can be traced back to whoever is responsible.
QUOTABLE
If everything stopped and there was no more model progress, we would still just have years and years of work to do here.
John Crepezzi of Jane Street on the Signals and Threads podcast.
Point72 Vet Builds Firm to Invest in AI Compute
Michael Ashby, who led Point72’s push into digital assets, said Wednesday that his new firm, Khumbu Advisory, is building an investment platform focused on AI compute. He sees opportunities in credit, equity and trades that take advantage of differences in how computing power is priced, per his LinkedIn post.
Banks Seek to Rein in Soaring AI Costs, McKinsey Says
Some banks are setting up teams to control AI spending. Token costs could reach 20% of banks’ total labor costs, Reuters reported, citing McKinsey.
Point72 Unit Sues Former Analyst Over Alleged Code Theft
Point72’s quant arm, Cubist, sued former research analyst Xin Jin, alleging he used Anthropic’s Claude Code and OpenAI’s Codex to steal proprietary source code. Jin’s lawyer told Bloomberg the claims were without merit.
ROUNDUP
What Else I’m Reading
SEC, CFTC Pressed by Hedge Funds to Delay Full 24/7 Market Trading | BBG
Stifel Working with Anthropic to Reimagine Its Dashboard | Press Release
JPMorganChase expands Claude Code to more employees | Anthropic
Most Large RIAs Lack an AI Leader, Data Shows | WealthManagement.com
Banks ban AI from accounts over fears of rogue agents | The Telegraph
Claude gets personal finance tool | Finextra
JPMorgan Chase signs cybersecurity and coding deal with OpenAI | Finextra
Goldman elevates record number of engineers to coveted partner ranks | FT
How Hackers Steal Millions in AI Computing Power | WSJ
Our software engineering interviews are changing (a little) | Jane Street
This Week in AI Street
Jump Trading on What It Takes to Trade Compute
Renting GPUs is like booking a vacation in the 1980s. You call a few travel agents, ask what’s available, compare prices and hope for the best.
And there’s a mismatch between buyers and sellers. GPU owners want to sign long, steady contracts with established counterparties. GPU renters often want short-term access and don’t have much of a track record.
This disconnect is what Affaan Mustafa and Alejandro Soumah set out to solve at Itô Markets. The team joined Jump Trading in October to lead a new compute trading desk. Mustafa said the desk is bringing together GPU capacity and offering it to customers “in smaller blocks with more flexible terms.”
We also talked about how Itô built an agent to match GPU buyers and sellers on Slack and WhatsApp, what a standard GPU contract might look like and whether AI tokens could become a market of their own.
SPONSORSHIPS
Reach Wall Street’s AI Decision-Makers
November sponsorships are SOLD OUT. Limited December placements are available starting Dec. 3.
Reach more than 5,000 institutional investors and decision-makers across finance, AI, and law. To discuss available dates and sponsorship fit, email sponsors@ai-street.co or fill out this form here.
CALENDAR
Upcoming AI + Finance Conferences
Big Data, Artificial Intelligence, and Financial Economics — November 13 • Cambridge, MA - NBER research conference on big data, AI and financial economics, also available on YouTube.
ICAIF ‘26 — November 14–17 • Milan ← I’ll be attending
7th ACM International Conference on AI in Finance, bringing together researchers and practitioners on generative AI, trading, risk, fraud detection and responsible AI, hosted at Bocconi University.QuantMinds International — November 16–19 • London
Quantitative finance conference bringing together practitioners, academics and asset managers to discuss models and research.
Momentum AI Finance — November 16–17 • New York
Reuters Events forum for financial-services leaders on scaling AI adoption, governance and measurable business impact.
AI in Finance 2026 — November 17 • London
Informa event for banking and technology leaders on deploying AI at scale, governance and AI-powered customer experiences.
From Data to Decisions: AI in Asset Management — November 17 • Zurich
Swiss AI Summit satellite event on AI in investment research and decision-making, with speakers from Aviva Investors, Robeco, UBS Asset Mgt and others.
AI in Financial Services Conference USA — November 18 • New York
GlobalData conference on generative and agentic AI deployment across banking, insurance and wealth management.FinRegLab AI Symposium — November 18–19 • Washington, DC
FinRegLab forum on AI’s economic impact, agentic payments, credit and risk models, and financial policy. Registration requires an invitation request.
Singapore FinTech Festival — November 18–20 • Singapore
Fintech festival featuring sessions on agentic AI, financial infrastructure, payments and the future of financial services.
Neudata New York Winter Data Summit — December 3 • New York
Investment data summit covering alternative data, AI agents, data infrastructure and their uses in investment research.
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