Financial trading instructions have gone through significant changes over the past two decades, with the evolution of electronic trading, algorithms, smart order routers, the rise of HFT, and ever more complex strategies in search of alpha. 

The Trading Latency Race 

One of the significant challenges has been the need to achieve ever lower latency to ensure your algorithm can be first, achieve the highest profit, and minimize slippage. The latency race has been evolving for years, and depending on the strategy, sits within different time horizons: from simple arbitrage in the single-digit nanosecond range to highly complex strategies in 10’s of Microseconds. Whichever algorithm you use or write, there is continued drive for reducing latency and maintaining performance while increasing the complexity and accuracy of your algorithms. After all, the competition never sleeps.

As with most industries, technology is pushed to the limits by the few at the extreme end before filtering down into the wider market. Take automotive as an example. Formula One racing drives huge innovation in the car market, as do the manufacturers of supercars or high-end electric vehicles. These niches drive the future of automotive which eventually filters down into the majority of cars. 

The same is true for financial trading. The high-frequency trading (HFT) markets lead the race in technology, investing heavily into the bleeding edge applications and taking calculated risks on new investments and strategies. Many large banks are also investing into cutting-edge technologies for risk and the search for alpha. 

The Future of FPGAs in Trading 

One such technology is FPGA’s. FPGAs provide a highly deterministic, highly parallel, and low latency platform to implement trading strategies (and more recently AI) to ensure the fastest systems. Because of the need for custom hardware programming and the cost and time involved in this, FPGA technology has proved challenging for the wider market and often out of the reach of the majority, much like we don’t see many F1 cars on the road. 

Since inventing the technology 36 years ago, Xilinx has been in the forefront of FPGA technology. Now we’ve created the Accelerated Algorithmic Trading (AAT) Platform, a full reference architecture which is democratizing the technology for use to the wider market. AAT is built on HLS, a C/C++ style programming language designed to make FPGA’s accessible to all, and is powered by Xilinx Alveo, the industry's first enterprise-ready FPGA cards. Xilinx is enabling the next generation of trading platforms based on industry-leading technologies at a low cost of entry. AAT allows the electronic trading software community to benefit from FPGA’s in a way not possible before.  

The Details of Trading Democratization 

AAT is an illustration of our wider vision of making the benefits of FPGAs available to the broader community of software developers. A cornerstone of this vision is Xilinx Vitis, our unified software development platform. Vitis provides a lot of flexibility for developers to seamlessly build accelerated applications on FPGAs. As part of Vitis unified software environment, Xilinx offers Vitis libraries, an open-source, performance-optimized libraries that offer out-of-the-box acceleration with minimal to zero-code changes to existing applications, written in C, C++ or Python.  

The integration of Xilinx Vitis libraries provides quants with a new platform to accelerate their trading strategies using already accelerated libraries, many of which can be run from python. 

The AAT reference design supports the CME market data platform (MDP) exchange feed for Market by Price (MBP) and can be used to create a broad range of low latency algorithmic trading applications for brokers, exchanges, market data vendors, sell-side vendors, and proprietary traders while minimizing losses to HFTs.

Looking To The Future 

Xilinx is also preparing users for the future AI race. FPGAs deliver the fastest AI inferencing, combined with highly parallel processing capabilities and low networking latency.  As AI becomes an integral part of trading strategy, this positions Xilinx Alveo trading deployments to easily extend into AI without requiring significant additional hardware, along with no custom hardware development.  

Want to learn more? Join us for a live webinar and get insights on how to customize AAT for your own trading platform.  Register today at https://pages.xilinx.com/EN-WB-2021-06-30-AAT_LP-Registration.html

For more information about AAT visit https://www.xilinx.com/applications/data-center/financial-technology/accelerated-algorithmic-trading.html