For most of the 2010s, the smartest algorithmic trading code in US equities lived inside the sell-side market makers at Citadel Securities, Virtu, Two Sigma Securities, Jane Street and IMC. By 2026 that has shifted. The most aggressive recent improvements in execution quality and cost have come from buy-side firms, particularly the systematic credit and equity desks at the largest US asset managers, who now run their own internal smart order routers and increasingly bypass traditional sell-side intermediation for a meaningful share of their flow. The change is quiet, but it is large enough to show up in published execution quality reports.
What algorithmic trading actually covers in US markets today
Algorithmic trading in US markets is now an umbrella term for at least five distinct activities. The first is automated execution: VWAP, TWAP, implementation shortfall and the newer market-impact-aware schedules that break up institutional orders. The second is electronic market making at exchanges and dark pools, dominated by the sell-side principal trading firms. The third is statistical arbitrage and quantitative directional trading at hedge funds and proprietary trading firms.
The fourth is the systematic asset management end, where firms like Two Sigma, AQR, DE Shaw and Renaissance run model-driven portfolios. The fifth is the increasingly software-driven buy-side execution, where traditional active managers (Fidelity, BlackRock, Capital Group) have rebuilt their trading desks around in-house algorithms that can route directly to venues, IOIs and broker dark pools.
The technology stack underneath all five activities has converged. Co-located servers at Nasdaq and NYSE data centres, FPGA-accelerated parsers for the SIP and proprietary direct feeds, kernel bypass networking, custom timing infrastructure, and increasingly hardware-accelerated machine learning inference. The cost to set up a competitive execution stack is now lower than it was a decade ago, which has lowered the barrier for buy-side firms to bring execution in-house. The US payment rails fintechs sit on are downstream of the settlement infrastructure that handles the eventual T+1 clearing.
Co-location economics also matter. A rack inside the Nasdaq Carteret data centre or the NYSE Mahwah facility costs tens of thousands of dollars a month and is essentially mandatory for any firm trying to compete at the front of the book. The economics favour scale, but the marginal cost of adding a second strategy on an existing co-located rack is low, which is why the firms that already have the footprint expand more easily than new entrants.
How buy-side desks have taken back execution from the sell side
Three forces drove the buy-side rebalance. The first is access to better venue connectivity. The post-Reg NMS proliferation of US trading venues (now sixteen registered equity exchanges plus several dozen dark pools) created an arbitrage opportunity for any firm that could route intelligently across them. The sell-side smart order routers exploited that for years. Buy-side firms have started writing their own to keep the spread.
The second is data. The buy side now has access to the same direct feeds, the same alternative data sets and the same machine learning libraries that the sell side uses. The asymmetry that historically favoured sell-side firms has narrowed to specific niches (latency-critical market making, certain options books) rather than across the board.
The third is the FIX-protocol-mediated trading workflow. The buy-side trading desks at the largest US asset managers run order management systems (Charles River, BlackRock Aladdin, SimCorp Dimension) that increasingly include execution algorithm libraries written in-house. The OMS is no longer just a router. It is the actual trading platform. Settlement plumbing on the US side now mostly happens in the background relative to the order management surface that traders see.
The buy-side execution capability has also reshaped recruiting. The strongest quantitative researchers and execution engineers, who used to default to the sell-side prop trading firms, increasingly accept offers at the major US asset managers. The compensation gap has narrowed enough that the choice now turns on the breadth of problems each side offers rather than purely on pay.
A scoreboard for US algorithmic trading activity in 2025
The composite figures below come from SIFMA, the SEC Market Information Data Analytics System (MIDAS), CBOE and Nasdaq market data, and the public Rule 605 and Rule 606 reports that brokers and venues file.
The number that has moved most is the share of US institutional equity volume executed through buy-side-controlled algorithms rather than sell-side high-touch desks. That share has crossed roughly seventy percent at the largest US asset managers, up from below forty percent a decade ago. The implications for sell-side commission revenue have been substantial, and the major US broker-dealers have responded by repositioning around prime brokerage, capital introduction and other non-execution services.
Options trading deserves a longer look. The US options market has seen the most aggressive recent growth in retail participation, particularly in zero-day-to-expiry contracts. The interaction between retail flow and market-maker delta hedging has become a recurring topic at SEC roundtables and academic finance papers. Market structure around that interaction is still being worked out.
The risks and trade-offs that still matter
Three risks recur in supervisory and post-mortem reports on US algorithmic trading. The first is market structure fragility. The May 2010 flash crash, the August 2015 ETF dislocation, the May 2022 mini-crash and several smaller recent episodes have all involved interactions between algorithmic strategies that nobody fully anticipated. The Limit Up-Limit Down regime and the consolidated audit trail have improved oversight, but the underlying coupling of automated participants remains a tail risk.
The second is model risk. A buy-side desk that runs its own execution algorithms inherits the model risk burden that the SEC and FINRA have long applied to sell-side firms. Documentation, backtesting, monitoring and governance practices that were comfortable inside a sell-side compliance organisation are now table stakes for any buy-side firm that runs its own execution.
The third is cyber and operational risk. The 2023 ransomware incident at ION Group disrupted derivatives trading across multiple US firms for days. The country’s algorithmic trading infrastructure depends on a small number of critical vendors and infrastructure providers, and a single outage at any of them can ripple through the market. Banking innovation that scales globally in the trading space increasingly involves co-investment in shared resilience infrastructure rather than purely competitive builds.
Treasuries are where the most interesting structural change is happening. The shift toward all-to-all electronic trading on platforms like OpenDoor, BrokerTec and Fenics has reduced the role of inter-dealer brokers and given buy-side firms direct access to liquidity that used to require sell-side intermediation. The 2025 Treasury market reform proposals from the SEC and the Federal Reserve are accelerating this transition.
What US fintech founders should understand about algorithmic trading now
For a US fintech founder considering an algorithmic trading product, three lessons hold up across recent launches. The first is that execution quality is the actual product. Retail brokerages compete on commission-free trading at the front, but the customer experience is shaped by the price improvement, the speed of fills and the reliability of the routing. Robinhood’s payment-for-order-flow controversy made this visible to a broader audience.
The second is that infrastructure choices matter more than headcount. A small US fintech that has invested in co-location, direct feeds and a credible smart order router will outperform a much larger fintech that has not, in measurable execution quality terms. The capital requirement for the infrastructure has dropped, which has democratised access but raised the bar on operational discipline.
The third is that the regulatory perimeter is unforgiving. The SEC’s MIDAS surveillance, FINRA’s market regulation function, and the consolidated audit trail have made every algorithmic trading firm in the US visible to regulators in near real time. Firms that operate inside that perimeter as if it were not watching tend to learn the hard way.
The shift of execution from sell side to buy side is not a story about one camp winning. It is a story about the entire US market structure moving toward a configuration where intelligence and capital can sit on either side of the trade. The interesting question for the next three years is whether retail-facing fintech execution converges on the same buy-side-dominant pattern.
For US market-structure context informing buy-side execution practice, see SEC market structure data.
