ChatGPT Analysis Matches Best-Performing Hedge Funds: Study

Hey, it's Matt. Here’s what’s up in AI + Wall Street. Was this email forwarded to you? Sign up below:

RESEARCH

ChatGPT Signals Linked to Higher Fund Returns: Study

Hedge funds whose trading patterns aligned with ChatGPT's forward-looking analysis of earnings calls earned 3-5% higher annual returns, according to new academic research.

The study examined how 633 hedge funds changed their stock holdings each quarter from 2016 to 2023. Researchers then compared these trading decisions to ChatGPT’s signals that were extracted from the same conference call transcripts that were available to the funds.

Researchers analyzed fund performance before and after ChatGPT's March 2022 base-model release by measuring how closely each fund's investment decisions aligned with these AI-generated signals. About 19% of hedge funds showed significant alignment by late 2022.

The largest gains appeared in bigger hedge funds, although the study doesn’t confirm whether they were actively using generative AI tools in‐house. ChatGPT and the best-performing funds were most likely to reach the same conclusions when analyzing specific companies rather than overall market trends.

The study doesn’t confirm whether funds were actively using AI, only that their investment patterns coincided with ChatGPT’s analysis.

I asked lead author Jinfei Sheng for his takeaway:

𝘊𝘩𝘢𝘵𝘎𝘗𝘛 𝘪𝘴 𝘱𝘰𝘸𝘦𝘳𝘧𝘶𝘭 𝘧𝘰𝘳 𝘱𝘳𝘰𝘤𝘦𝘴𝘴𝘪𝘯𝘨 𝘭𝘢𝘳𝘨𝘦 𝘢𝘮𝘰𝘶𝘯𝘵𝘴 𝘰𝘧 𝘵𝘦𝘹𝘵 𝘥𝘢𝘵𝘢 𝘰𝘧 𝘧𝘪𝘳𝘮𝘴, 𝘸𝘩𝘪𝘤𝘩 𝘣𝘦𝘯𝘦𝘧𝘪𝘵𝘴 𝘪𝘯𝘷𝘦𝘴𝘵𝘮𝘦𝘯𝘵 𝘱𝘦𝘳𝘧𝘰𝘳𝘮𝘢𝘯𝘤𝘦. 𝘛𝘩𝘪𝘴 𝘪𝘴 𝘦𝘴𝘱𝘦𝘤𝘪𝘢𝘭𝘭𝘺 𝘵𝘳𝘶𝘦 𝘪𝘧 𝘺𝘰𝘶 𝘩𝘢𝘷𝘦 𝘮𝘰𝘳𝘦 𝘬𝘯𝘰𝘸𝘭𝘦𝘥𝘨𝘦 𝘢𝘣𝘰𝘶𝘵 𝘈𝘐 𝘢𝘯𝘥 𝘸𝘪𝘵𝘩 𝘮𝘰𝘳𝘦 𝘳𝘦𝘴𝘰𝘶𝘳𝘤𝘦𝘴 (𝘥𝘢𝘵𝘢, 𝘵𝘦𝘢𝘮).

Authors: Jinfei Sheng, Zheng Sun, Baozhong Yang, Alan L. Zhang
Paper: Generative AI and Asset Management

COMPUTE

How Much Power Are Data Centers Consuming?

Ever since DeepSeek's new open-source model sparked Nvidia's massive selloff last month, I've been trying to see whether AI computing demand is actually decreasing. So far, I've not seen any evidence of a slowdown in demand — and I've written before about Big Tech's plans to spend billions more on building AI infrastructure.

• Google reported a striking statistic in its Q4 earnings: Cloud customers consume more than 8x (!) the compute capacity for training and inferencing compared to 18 months ago— which is only mid-2023!

• Meta is bringing almost a gigawatt of capacity online this year alone, and is building what Zuckerberg called "a 2+ gigawatt AI data center that would cover a significant part of Manhattan if it were placed there."

• Meanwhile, Amazon acknowledges they're hitting the limits: "It is true that we could be growing faster, if not for some of the constraints on capacity," CEO Andrew Jassy noted on their Q4 earnings call.

This week, I was trying to see if there's a datapoint that could serve as a proxy for AI computing demand—sort of like how Kastle Systems tracks office badge swipes as a barometer for return-to-office trends.

AI Street has sophisticated readers, so if you have any ideas on how to capture this, I'd love to hear them.

FUNDRAISING

Harvey AI Hits $3B Value with $300M Sequoia Round

Legal AI startup Harvey raised $300M led by Sequoia Capital, with new investors Coatue and LexisNexis joining existing backers, as the company reports 4x revenue growth

Startup Tines Hits $1.1 Billion in Goldman-Led Round

AI startup Tines raised $125M led by Goldman Sachs, reaching unicorn status as it expands its business automation software used by Canva, Databricks, and Oak Ridge National Laboratory. (Bloomberg) $

Sardine Secures $70M for AI-Driven Fraud Prevention

Fraud prevention startup Sardine raised $70 million in funding led by Activant Capital to expand its AI-powered compliance and risk management tools, bringing its valuation to $660 million as it serves major clients like X money and develops new features to combat deepfake-based fraud. (Bloomberg) $

All three founders are building tools to solve problems they faced in their previous jobs.

• Harvey CEO Winston Weinberg, a former BigLaw associate

• Tines founder Eoin Hinchy, former cybersecurity exec at DocuSign

• Sardine CEO Soups Ranjan, former head of financial crime risk at Revolut

Knowing how the industry thinks is a big help as is being able to sell to your old colleagues.

INVESTING

Reddit Partners with NYSE Parent on Sentiment Data

NYSE Parent ICE will develop trading data products from Reddit forum content to help hedge funds identify real-time market sentiment shifts and power trading algorithms. (WSJ) $

I imagine traders were scraping Reddit previously but after the company prohibited automated bots, they now have to pay for this valuable data. I asked Dexin Zhou, a professor at Baruch College who’s studied the effects of retail traders, for his thoughts:

1. Increased retail activity of retail investors and given that they engage heavily in social media activities. Sophisticated investors can learn a lot from social media data about retail traders

2. These textual data are helpful in training large language models

See Zhou’s study here: $1 Invested in Retail's Bullish Stocks Shrinks to 60¢: Study

MDOTM, Noonum Partner on AI & Thematic Investing

Summary: MDOTM Ltd. has partnered with Noonum to integrate Large Language Models and Knowledge-Graph technology into its AI investment platform, Sphere, enabling asset and wealth managers to leverage advanced thematic investing tools for greater personalization and efficiency. (Press Release)

• Five Minutes with MDOTM’s Peter Zangari, PhD

COMPANIES

AI to Impact Nearly All of Wells Fargo: CFO

Wells Fargo’s CFO said AI will soon impact nearly every aspect of the bank’s operations, with numerous AI initiatives currently being piloted across the company. (MSN)

IBM CEO: DeepSeek to Drive AI Growth

IBM's CEO Arvind Krishna believes AI costs will drop after Chinese startup DeepSeek demonstrated cheaper model development. Speaking at Dubai's World Government Summit, he predicted this would boost AI adoption and validated IBM's view that effective AI models don't require massive spending. (Bloomberg) $

Wall Street Opens Up to Open Source

Financial firms are increasingly sharing their technology through open-source platforms, with JPMorgan, BlackRock, and others making internal tools public as they look to cut costs and speed up AI development. (Business Insider) $

POLICY

US, UK Opt Out of AI Declaration at Paris Summit

Earlier this week, government officials, AI executives, and academics met in Paris to discuss the future of AI.

• US and UK Decline AI Declaration: Both countries chose not to sign a global declaration advocating for “inclusive and sustainable” AI, citing concerns over clarity on governance and national security. (BBC)

• US Warns Against Overregulation: Vice President JD Vance emphasized that excessive AI regulation could impede innovation, highlighting the administration’s focus on rapid technological advancement. (AP)

• France Announces Major AI Investments: President Emmanuel Macron unveiled plans for €109 billion in private sector investments to strengthen France’s AI infrastructure and competitiveness. (CNBC)

ADOPTION

AI Adoption by the AI Chat Numbers

If you’re reading this newsletter, you’re an early AI adopter. Most people are not experimenting with it yet.

A new study from Anthropic analyzed millions of conversations with its AI chatbot, Claude, mapping them to occupations using the U.S. Department of Labor’s database.

• Two-thirds of occupations haven’t even tried AI in a quarter of their tasks.

• Even among those experimenting, only 4% use AI for more than three-quarters of their work.

While software development dominates AI usage at 37.2%, finance-related activities make up just 5.9% of that. But unlike other tech trends (like VR headsets), I believe this pattern will change. Once you get familiar with how best to use AI, it’s pretty easy to see its benefits. My own AI usage keeps increasing.

AI Investments Haven’t Boosted Productivity Yet: Fed

One-Sentence Summary: Despite heavy AI spending and market hype in 2024, companies have yet to see clear productivity or job gains. (Fed)

WHAT ELSE I’M READING

Can AI Improve Investment Decisions in Systematic Trading?

An analysis of how multiple AI models can enhance quantitative trading through automated research and execution, while maintaining human oversight and risk controls. (Quant Journey with Code)

AI Agentic Systems in Financial Markets

The evolution of AI in financial markets, from basic LLM use to automated workflows and multi-agent systems, with applications in research, portfolio monitoring, and trading. (Rajesh T. Krishnamachari)