Wall Street’s biggest high-frequency trading (HFT) firms have long relied on expensive, proprietary trading systems for their competitive edge. But a new challenger could be emerging from an unexpected direction: open-source artificial intelligence (AI). While traditional financial powerhouses invest millions in proprietary algorithms, platforms like Chinese AI startup DeepSeek could make sophisticated trading technology free—or nearly free—to anyone. This shift raises an important question: Could cheap and more accessible AI reshape Wall Street, or will the traditional barriers of infrastructure and expertise maintain the status quo?
Harry Mamaysky, director of financial studies at Columbia Business School and an expert on using AI in finance, points out that DeepSeek is the culmination of many developments. “A lot of AI is already open-source,” he told Investopedia, referring to Llama, the AI model launched by Meta (META), and the company Hugging Face.
“The tricky part is getting the hardware to run this on, getting the data to feed to it, and then customizing the generic models to specific use cases,” Mamaysky said.
Below, we’ll take you through how this would work for employing open-source AI in finance.
Key Takeaways
- Open-source projects benefit from a community of developers who continually improve the technology.
- The pace of advances are often faster than the slower processes of large financial institutions.
- By eliminating expensive licensing fees, open-source AI could significantly lower the financial barriers for many, including smaller firms and independent investors.
- DeepSeek can be tailored to specific needs or preferences without requiring extensive technical management.
- But AI is just one part of very expensive processes, so it’s not exactly opening up such trading to everyone.
The Evolution of AI Trading
Wall Street trading has been dominated by elite firms wielding proprietary AI systems—expensive algorithms developed behind closed doors with massive resources. These institutions have historically maintained their edge by combining their deep pockets, specialized talent, and sophisticated computing infrastructure. A recent industry analysis indicates that developing advanced AI trading models requires investments ranging from $500,000 to well over $1 million, before factoring in ongoing costs for talent retention and infrastructure maintenance.
The use of AI in trading dates to the 1980s, when firms first began using simple rule-based systems for automated trading. The real transformation came in the late 1990s and early 2000s when machine learning algorithms advanced the era’s quantitative trading strategies. Major firms like Renaissance Technologies and D.E. Shaw pioneered using sophisticated AI models to identify market patterns and execute trades at unprecedented speeds. By the 2010s, high-frequency trading (HFT) powered by AI had become a cornerstone of market operations, with the largest firms investing hundreds of millions in computational infrastructure and talent to maintain their competitive advantage.
Algorithmic high-frequency trading is estimated to account for about half of Wall Street’s trading volume.
DeepSeek and similar open-source AI initiatives are challenging this traditional model with their collaborative approach to development. Rather than keeping algorithms locked away, these platforms harness the collective expertise of a global developer community that continuously refines and improves the technology.
However, getting on board isn’t as simple as downloading open-source code. While these new tools cut down on certain barriers to entry, they don’t automatically level the playing field. Traditional trading systems are deeply embedded in market operations and backed by years of real-world testing. The challenge for open-source alternatives lies not just in matching the sophisticated capabilities of established systems, but in proving they can reliably perform within the demanding constraints of live trading.
In addition, firms adopting open-source AI systems must still develop the right operational frameworks, ensure regulatory compliance, and build the needed infrastructure to deploy these tools effectively. So, while open-source AI may lower the costs of sophisticated trading technology, it’s unlikely that you’ll soon download open-source AI trading platforms like you would an open-source note-taking app.
Cost and Accessibility
One of the most appealing aspects of open-source AI is its potential to reduce upfront costs. Traditional proprietary systems demand hefty licensing fees and investments in custom software. Citadel LLC’s ongoing partnership with Alphabet Inc. (GOOGL), for instance, harnesses over a million virtual processors to cut complex calculation times from hours to seconds, but it requires massive ongoing infrastructure investments.
DeepSeek’s open-source approach offers a stark contrast. Its V3 and R1 models are freely available, and it uses an MIT license, meaning it can be modified and used for commercial purposes. While the software may be free, implementing it effectively requires significant investments in the following, Mamaysky noted:
- Computing infrastructure and hardware
- High-quality market data acquisition
- Security measures and compliance systems
- Ongoing maintenance and updates
- Specialized expertise for deployment and optimization
While you can certainly access DeepSeek’s latest model and download the code for free, successfully deploying it in an HFT environment demands far more than that.
How Much Does HFT Cost Average Investors?
A contested area around algorithmic HFT is how much it costs typical investors. Estimates vary widely, especially since much of the trading happens in dark pools and off-exchanges. Estimates in the 2010s put the figure in the tens of billions, but it’s likely far lower. A 2021 study estimated it costs $5 to $7 billion, but that was in the stock market alone—not including derivatives, currency, and other forms of trading.
Transparency and Accountability
A frequently touted advantage of open-source AI is its inherent transparency. With the source code available for public scrutiny, stakeholders can audit algorithms, verify their decision processes, and modify them to follow regulations or meet specific needs. A good example is International Business Machines Corporation’s (IBM) AI Fairness 360, a set of open-source tools to audit and mitigate biases in AI models. In addition, the architectural details and training data for Meta’s Llama 3 and 3.1 models are publicly available, developers can assess compliance with copyright and other regulatory and ethical standards. This level of openness contrasts with the “black box” nature of proprietary systems, where internal workings are hidden and sometimes lead to opaque decisions that even a system’s creators might be unable to unfurl.
However, it would be misleading to characterize all proprietary trading systems as opaque black boxes. Major financial institutions have made significant strides in improving the transparency of their AI models, driven both by regulatory pressure (initiatives like the European Union’s AI Act and evolving U.S. guidelines) and internal risk management needs. The key difference is that while proprietary systems build their transparency tools internally, open-source models benefit from community-driven auditing and validation, often speeding up problem-solving.
The Innovation Gap
DeepSeek’s R1 model breakthrough caught the attention of industry leaders—even OpenAI’s Sam Altman acknowledged in early 2025 being “on the wrong side of history” regarding open-source models, suggesting a potential shift in how the industry views collaborative development.
Nevertheless, Mamaysky said the real challenge of realizing the promise of a shift to open-source AI lies in three critical areas: scaling the hardware infrastructure, securing high-quality financial data, and adapting generic models for specific trading applications. As such, he doesn’t think the advantages of well-sourced firms will go away anytime soon. “Open-source AI, in and of itself, does not pose a risk [for competitors] in my view. The revenue model is the data centers, the data, the training, and the process robustness,” he said.
The AI race is further complicated by geopolitics. Former Google CEO Eric Schmidt has warned that the U.S. and Europe need to focus more on building open-source AI models or risk losing ground to China on this front. This suggests that the future of financial AI may depend not just on technical capabilities, but on broader strategic decisions about how trading technology is developed and shared.
The Bottom Line
The emergence of open-source AI platforms like DeepSeek could represent a significant shift in financial technology, but they aren’t yet a threat to Wall Street’s established order. While these tools dramatically reduce software licensing costs and increase transparency, Mamaysky warned that “making the models open source or not is probably not a first-order issue” for these firms.
What we’re likely to see is a hybrid future with combined open-source and proprietary systems. So, the question isn’t whether open-source AI will replace traditional Wall Street systems but how it will be integrated into them.
:max_bytes(150000):strip_icc()/GettyImages-2198847002-cc8108a7220b49a08ba3c15af197daff.jpg?w=1200&resize=1200,0&ssl=1&description=How+More+Accessible+AI+Could+Revolutionize+Wall+Street+Trading){kind=link}