The recent ups and downs of the markets have been caused by high frequency exchanges decided not by people in the flesh and blood, but by automatic systems. Sophisticated software that send orders to second -sized infinitesimal fractions. Also amplifying the risks.
Like vessels in an increasingly stormy sea, the equity titles undergo strong oscillations, at times violent and often unpredictable. To trigger the waves are the declarations of the American president Donald Trump, now on the commercial duties that inflame relations with China, now on the Federal Reserve, now on the general state of the economy. A fundamental nervousness fueled by the increasingly cautious prognosis of international institutions, which warn a slowdown in world growth.
In this phase of dizzying ups and dizzying, although often invisible to most, it is played by operators as powerful as they are discreet: the so -called “robot traders”. Just look at the recent thud of Wall Street: between 2 and 8 April 2025, the S&P 500 index lost 10.8 percent in a few sessions, triggered by the escalation on the US-China duties. Many analysts agree: the reduction has been fed, accelerated, amplified by high frequency exchanges decided not by trader in flesh and blood, but by automatic systems. These are very sophisticated software, capable of analyzing an immense amount of data, deciding operational strategies and sending purchase and sale orders to infinitesimal fractions of a second.
This silent revolution has redesigned the very nature of the financial markets. It is estimated that between 60 and 75 percent of all negotiations globally are managed by algorithms, a percentage that in some segments, such as American futures, exceeds 80 percent. A transformation that brings efficiency and speed, but also introduces unpublished risks and raises questions about the future stability of finance.
When it comes to automatic trading, we enter a complex universe, dominated by two main technological families. The base is made up of algorithmic trading. In essence, it is the use of computer programs to perform stock exchange orders following predefined rules. An algorithm can be instructed, for example, to buy actions if the price exceeds a certain mobile media or to sell if it goes down under a threshold. The goal is not just speculative. Large institutional investors (pension funds, insurance) use algorithmic trading to perform orders of enormous size efficiently, splitting them not to influence the price too much and obtain better average conditions. It is a question of precision and minimization of large -scale costs. His most extreme version has evolved from algorithmic trading: the high-freequense trading (HFT). If the first is a sporty car, the HFT is the formula1 of the markets, where the only thing that matters is pure speed. These systems operate in milliseconds and even milliones of second. To do this, the HFT companies invest fortunes in technology: very powerful servers placed as close as possible to bags and ultra-welfle connections. Their game is not to focus on large trends, but to accumulate tiny profits (hundredth fractions) on a gigantic number of operations, using micro-artear or acting as a market maker, that is, providing liquidity and gaining from the differential between purchase and sale price.
This algorithmic domain is global. As mentioned, more than half of the world exchanges is automated. Italy is no exception. Already in 2017, Consob estimated an HFT share close to 30 percent on the Italian stock exchange. Most recent data indicate an overall weight of algorithmic trading around 54 percent on the After Hours segment.
The protagonists are the giants of finance: investment banks, hedge fund, institutional and companies specialized in owner trading that invest proper capital. The HFT remains an exclusive club because of the high costs. However, simpler forms of algorithmic trading are also accessible to private traders advanced through online platforms, although with capacity not comparable to institutional ones. The operation of a “robot trader” is a continuous cycle: it receives and processes in real time a huge flow of data (prices, volumes, order books, financial news, macro data, even sentiment on social media). To interpret them, use an arsenal of tools: from classical technical indicators to more advanced techniques. The real qualitative leap took place with the integration of artificial intelligence (IA) and “machine learning”. Techniques of “Natural Language Processing” allow you to “read” and interpret news and tweets to capture the mood of the markets. “Reinforcement Learning” models allow the algorithm to independently learn the best strategies by operating in simulations. The ability to adapt is crucial: systems based on “machine learning” can change the behavior in response to market changes, without human intervention. Once the signal has been generated, the order is sent to the stock exchange in a fraction of a second.
Automatic trading offers many advantages: unsurpassed speed, total elimination of emotion, superhuman analysis capacity. Despite this, the omnipresence of algorithms also introduces new and significant fragility in the financial system. In times of panic, when the market begins to descend quickly, algorithms tend to withdraw their purchase orders almost simultaneously, drying up liquidity just when it would be needed more. In addition to the impact on general volatility, automation brings specific and sometimes catastrophic risks with it: a simple bug, a “Baco” in the code, a hardware problem, an error in the configuration of the parameters or a defective software update can trigger devastating consequences. An algorithm can begin to send incorrect orders, to buy or sell enormous quantities of securities at off -market prices, generating millionaire losses in a few minutes.
The most famous and cited case is that of Knight Capital Group in August 2012: An upgraded algorithm began to buy and wildly sell actions on the New York market, costing the company over $ 440 million in less than 45 minutes and bringing it to the verge of bankruptcy. Another famous episode is the Flash Crash of 6 May 2010, when the Dow Jones index lost almost a thousand points (about 9 percent) in a few minutes, and then recover most of the loss at the end of the day. Although the initial cause can be attributed to a single large sales order (performed through an algorithm), there is ample consensus on the fact that the chain reactions of the HFT algorithms, both through the massive withdrawal of liquidity and through the automatic activation of Stop-Loss orders, have amplified the collapse properly. Similar events, albeit of minor scope, have repeated themselves, such as the mini-crash of the ‘P 500 on 11 May 2022, which lost 5 percent in 10 minutes due to, it would seem, of an algorithmic error.
Not only that. The extreme speed of the HFT can be used to implement manipulative practices, difficult to identify and contrast in real time. For example, insert large purchase or sale orders without the intention of performing them, for the sole purpose of creating a false impression of demand or demand and thus pushing the other participants (humans or more slow algorithms) to move in a certain direction, and then quickly erase the order-Esca and take advantage of the induced price.
Global regulators of the bags (such as sec and consob) are aware of the risks And they try to mitigate them with tools such as “circuit breakers” (automatic negotiation interruptions in case of excessive movements), more stringent rules on the tests of algorithms and increasingly sophisticated market surveillance systems. However, technology runs very fast and keeping the step is complex. The traders robots are now an integral part of the financial ecosystem. They offer efficiency and liquidity (in normal times), but their speed and interconnection make them powerful risk amplifiers, especially during the stress phases. They are not the ultimate cause of turbulence, which have cheap and geopolitical roots, but can transform an abnormal wave into a tsunami. And make the sea of finance much more dangerous.
