Automatic software recalls approach record sixth year in a row

For possibly a record sixth consecutive year, the number of software recalls will exceed last year’s record of uselessness. So far in 2025, 166 recalls have affected various automakers. Looking at the 2024-2025 period combined, Tesla has the most vehicles affected (5,777,605), Ford has the largest number of recalls (73), and Stellantis suffers the highest cost per vehicle sold ($743.22) due to the software recalls.

“We are looking at a pivotal moment in history,” claims Todd Warren, an adjunct professor of computer science at Northwestern University, who tracks NHTSA-reported software recalls on a website, along with several metrics, including the percentage of wireless solutions by manufacturer. “Automakers will either collapse under the weight of their own technical debt or emerge with a significant competitive advantage, given better quality and methods to address problems in the field.”

That competitive gap may be more astonishing than the rising number of recalls. In a recent WIRED article entitled “Why Are Car Software Updates Still So Bad,” Carlton Reid suggests that the average cost of a recall is $500 and, according to Harman Automotive, over-the-air software reflashing costs $66.50 per vehicle. By applying these costs to each manufacturer and their varying ability to resolve their over-the-air security issues by 2024-2025, Telsa will save billions of dollars. Meanwhile, Ford has the highest total costs at over $2.5 billion, while Stellantis trails just behind at $1.8 billion. Comparing these egregious costs to their domestic rival, General Motors ($443,000 over 2024-2025), Ford is at a $1 billion annual disadvantage just from software recall costs, and each Stellantis vehicle sold carries a $661 burden on each GM vehicle. However, these staggering numbers ignore the associated warranty bills, the impact of customer dissatisfaction, and jury damages, such as the recent $200 million verdict against Tesla.

The causes

The multivariate equation of contributing causes and associated effects is incredibly complex and beyond even the imagination of any analyst or consultant, despite the claims of some. It is worth describing the underlying themes of the root causes in detail to understand possible ways forward.

INCREASING COMPLEXITY

In the same WIRED article, the author quotes McKinsey that “… since 2021, the complexity of the latest software platforms has increased by approximately 40 percent per year.” In part, the pressure to add new technology is forcing this, with almost half (48%) of car buyers prioritizing technology in vehicles over brand or styling, and China’s 49.9% year-on-year increase in the installation of advanced driver assistance systems (ADAS). Standing still means falling behind globally.

BIDIRECTIONAL IGNORANCE

Traditional automotive development from the likes of Detroit and Munich has emphasized deep, deliberate engineering accuracy. Several major consulting firms weighed in on the “talent gap” between Silicon Valley and the automotive industry, leading several OEMs to try the quick fix by injecting “fail fast” methodologies used for websites and apps. The strong problem: neither culture understood the necessary elements of each other’s way of working and how to integrate the required quality aspects with the necessary speed elements. As Harley Shaiken, professor emeritus at UC Berkeley, puts it well: “Silicon Valley is not the future, Detroit is not the past. They are both part of the winning strategy for the future.”

TECHNICAL DEBT

According to Planview, more than 20% of development budgets spent on new products are “…diverted to address technical debt issues,” which is the cost of old code waiting for completion, updates, or fixes. In other words, teams want to win the next competition, but have injuries or unfulfilled obligations from the last match. For companies that have built vehicles since the dawn of software, averaging about 2,000 semiconductor chips per vehicle and potentially thirty or more platforms, that historical burden can weigh significantly on strategic plans.

Author’s note

One of my five predictions at the start of 2025 was that the cost of recalls would double, and that Ford or GM would be acquired because neither had the fundamentals and economic advantages to “…continue unaided” in a costly series of unrealized gambles (e.g. autonomy, electric vehicles, Ford’s infamous FNV4 architecture). While neither prediction has been completely true, the trends suggest I wasn’t completely wrong either, so Ford will need at least some aggressive course corrections to avoid this fate.

Sometimes I hate being right.