Three Foreign High-Frequency Quantitative Trading Giants Under Investigation by Chinese Customs

Deep News
Oct 10, 2025

Since July, the Chinese subsidiaries of three major foreign high-frequency quantitative trading giants - Tower Research Capital, Jump Trading, and Optiver - have successively been investigated by the Shanghai Special Commissioner's Office of China General Administration of Customs. The investigation focuses on the compliance of import customs declaration processes for specialized communication equipment used by these three companies within China.

A person familiar with the investigation revealed that over the past three months, customs officials have conducted investigations at the Shanghai offices of the aforementioned three foreign high-frequency quantitative trading giants, with the first two companies being required to implement restrictive management measures for such specialized communication equipment.

According to the aforementioned insider, the investigation found that Tower Research Capital had undeclared or falsely declared overseas customized equipment in cabinets leased at Tonglian data center, including servers, FPGA/ASIC chips, network cards, and switches, with some equipment lacking 3C certification. The investigation progress of Jump Trading and Optiver remains unclear.

When contacted for verification, a relevant official from the Shanghai Special Commissioner's Office of China General Administration of Customs responded: "This matter belongs to internal work and is not convenient for external publicity."

**The "Hardware Speed Race" of Foreign High-Frequency Trading**

In the field of foreign high-frequency quantitative trading, hardware equipment is a critical component for achieving microsecond-level response speeds, which is crucial for the profitability of high-frequency quantitative institutions. It is understood that after 2010, foreign high-frequency traders began comprehensive "speed competition" in software, hardware, and network architecture, including hardware integration with customized servers integrating core components such as SSDs, GPUs, FPGA/ASIC chips, and optimized cooling systems; network customization through modification of network cards and switches, and development of specialized routing equipment optimized for trading business characteristics; strategy hardware implementation through FPGA and ASIC chips directly implementing strategy logic to avoid memory latency issues in software computing.

Regarding Tower Research Capital's equipment at the Tonglian data center, the aforementioned insider indicated that its customized FPGA/ASIC chips have speed advantages in parsing exchange native protocols faster than standard equipment, while consuming less power. Some brands of its customized servers have no public purchasing channels, and certain server brands with public information require specific conditions (such as providing purchaser's target exchange, project timeline, and referrals) to obtain pricing inquiries. Additionally, most switches cannot be found by brand and model information on public networks, but all provide FPGA or ASIC chip packaging methods to achieve full-process trading. Furthermore, most network card brands also have no public information available online. "The above equipment involves undeclared or falsely declared behavior," the aforementioned insider revealed.

**The High-Speed Corridor from Chongming Island to Zhangjiang**

Shanghai Tonglian data center is located in Shanghai Zhangjiang, directly across from the exchange data center. Market participants indicate that as a high-specification managed data center outside the financial system, Tonglian data center's core node function is to distribute trading signals from exchanges like CME transmitted through the Go West line to foreign high-frequency quantitative institutions. This core node requires high-specification switches to achieve efficient message forwarding, with trading signals such as CME gold futures forwarded from Tonglian data center leading market competitors by an average of 15-20 milliseconds. Shanghai Tonglian data center's fiber optic connection to the exchange data center naturally makes it the preferred choice for foreign high-frequency quantitative trading giants.

An industry insider pointed out that when trading signals from Tonglian data center enter the exchange data center, foreign high-frequency institutions very likely adopt "whole cabinet" business methods in exchange colocation facilities. To pursue ultimate trading speed, whole cabinets can deploy their own switches directly connected to exchange market data front-end machines and trading front-end machines, and can also deploy FPGA/ASIC chips on switches to directly parse market data and place orders directly to trading front-end machines, then replicate a complete agreed-format message to the trading system for record-keeping based on counter protocols. A more extreme deployment method is using one switch to connect exchange market data front-end machines and trading front-end machines, implementing unified functionality with FPGA or ASIC chips within that switch. This method not only raises fairness concerns but may also pose threats to exchange front-end machines, as the near-black-box whole cabinet deployment solution is not disclosed to futures companies, exchanges, or regulatory authorities.

**Regulatory Enhancement: New Programmatic Trading Rules Implementation**

The "Futures Company Supervision and Management Measures" clearly require futures companies to directly accept customer trading instructions through information systems with their own management authority. Industry insiders point out that since the Yishidun company futures market manipulation case exposed that the involved futures company used trading systems not purchased or developed by the company itself to transmit customer trading instructions without conducting trading customer account fund and position verification on that system, regulators have strengthened institutional requirements for futures companies in technical compliance, programmatic trading reporting, and fund and position verification. In March this year, multiple futures exchanges jointly issued documents clearly prohibiting futures companies from leasing cabinets to third-party institutions and requiring exchange data centers to be limited to futures companies' own use.

The "Futures Market Programmatic Trading Management Regulations (Trial)" implemented on October 9 this year defines high-frequency trading for the first time and establishes a dynamic regulatory framework. It clearly states that futures exchanges implement key management of high-frequency trading, focusing on verifying information such as trading strategy types, maximum order/cancellation frequency, and daily maximum order/cancellation volumes reported by high-frequency traders, while closely monitoring high-frequency trading behavior changes and timely assessing market impact. Futures exchanges can also establish order/cancellation fee systems and trading limit systems, timely adjust order/cancellation fee standards and trading limit standards, and implement differentiated management of high-frequency trading fees.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10