The Software Factory: 
Volvo’s CTO Anders Bell 
on industrialising the Software-Defined Car

Automotive production is being increasingly punctuated by the convergence of hardware and software. When code fails in a consumer application, the consequence is - perhaps at most - inconvenience. But when it fails in a vehicle travelling at speed on a motorway, the consequences belong to an entirely different order. Yet long before the road, flits the factory. Anders Bell, Volvo Cars' chief technology officer, carries that distinction without apparent drama, and with a clarity that suggests he has been living with its implications for some time.

Speaking at Torslanda, Volvo's manufacturing facility in Gothenburg, where the OEM's new EX60 had just completed its first production run, Bell offers AMS a perspective on software-defined vehicle development that cut against much of the industry's prevailing optimism. His framing was not pessimistic. It was precise. And precision, in this particular field, is worth more than enthusiasm:

"We talk more about architectures and platforms now, and increasingly about the tech stack, rather than just the vehicle architecture itself, because we view the whole car through a technology perspective. Those platforms are service-oriented and built around modularity in both hardware and software."

And that shift in perspective - as any professional intimate with the accelerating developments in carmaking can attest to - is not, by any stretch, cosmetic. That shift reflects a fundamental reorganisation of how Volvo conceives, manufactures and updates its vehicles, and it has implications that extend well beyond the software itself into manufacturing processes, workforce skills, supplier relationships, and the long-term economics of its business.

Volvo's engineering identity, Bell was keen to establish, is thoroughly and deliberately European. "This car (the EX60) was designed, developed and engineered in Sweden, literally just up the road from here - behind those fences where all our development operations are based.

"All the software, CAD and engineering work was done in Sweden, although we work with many partners across Europe. Around 90% of the parts in the EX60 are sourced in Europe." (The remaining 10% is fodder enough, for another article).

But the carmaker's vertical integration is equally striking. It builds its own machines, develops its own battery management systems, manages its own power electronics and now assembles structural batteries in-house. "We are vertically integrated very deeply into the stack: eMachines, batteries, battery management systems, power electronics and more.

"In fact, we are one of the most vertically integrated automotive manufacturers in Europe, and arguably one of the most European carmakers as well. Around 93% of all Volvo's software engineers are based here in Sweden, with the remaining 7% mainly in Poland - and a few in the US."

What "software-defined" production actually means at Volvo

The term "software-defined vehicle" has accumulated considerable freight in recent years, attached variously to infotainment updates, autonomous driving promises and vague assertions about future digital revenue streams. Bell is categorical about what it means in the Volvo context and, more pointedly, about what it does not mean.

"People often misunderstand what 'software-defined vehicle' means. It is not simply about updating infotainment systems," he says.

"This is a genuine ground-up software-defined architecture: advanced zone controllers, core computers, internet backbone, a singular unified operating system, integrated cybersecurity across both vehicle and cloud infrastructure, and more."

The distinction matters because it transforms the production-engineering challenge entirely. Updating a navigation interface is a display problem. Rebuilding the electrical and electronic architecture of a vehicle from first principles, so that every system can be managed, monitored and modified through software, is something of a categorically different magnitude.

Bell's counsel to any chief technology officer who encounters a comparable challenge is at root, philosophical - with a proven practical body of work. "Do not try to do too much too quickly," he says. "Do not put your project on an aggressive timeline to production. First, establish the fundamental proof of concept - then commit to the product." 

That guidance extends to the broader competitive pressure building across the industry. "You should not feel pressured into creating software-defined vehicles simply because the industry is moving in that direction. There are many examples in the media showing that rushing is not a good starting point. When you are creating something that has never existed before, it is impossible to define a detailed milestone timeline too early. You need to get much further ahead before you truly understand what you are trying to create and how to do it."

The failures he alludes to are numerous, well documented and, on at least some occasions, physically consequential. They are also, in his view, instructive. Volvo has learned from them, he says, including from some of its own earlier software difficulties.

Read the article

Volvo built the EX60 without stopping the line

Akhil Krishnan oversaw EX60 development while the XC60 ran at full production beside it. In this interview, he unpacks the industrial logic of SPA3, Volvo's AI-driven sustainability strategy and why flexibility has become the company's defining competitive thesis.

Why being smaller is Volvo's engineering edge

One of Bell's more counterintuitive observations concerns Volvo's size. The carmaker occupies a peculiar niche in the global automotive landscape: substantial enough to sustain the resource commitments that genuine software-defined vehicle development demands, yet agile enough to have escaped the institutional heaviness that afflicts most major volume manufacturers.

"We are in a uniquely fortunate position,” he says. “We are large enough to commit the resources required and to understand the scale needed across an entire vehicle portfolio. You cannot build something like this for only one car. It needs to scale across multiple platforms and powertrain configurations. But at the same time, we are small enough not to become institutionalised or trapped in committees."

The result is a form of productive hybridity that sets Volvo apart in global automotive manufacturing. "That combination gives us the best aspects of a startup culture - freedom to move quickly, think differently and build - while also understanding scale, mass production and long-term industrialisation,” says Bell. “Most startups are focused on survival - and so do not think deeply enough about what happens if they succeed at scale over the next ten years. We can think that way."

He pushes the argument further still. "We are probably the smallest fully independent end-to-end car development organisation in the world - from tyre to charge port. That small size is one of our strengths. People ask how we can achieve this when larger companies struggle. The answer is precisely because we are smaller."

This flexibility carries an inevitable corollary. Volvo has had to invent much of what it uses - because no one else had done it before at the required depth or specificity. "We are nimbler, faster and more adaptive. We have also made many mistakes, because there is nobody to copy. We are going into unknown territory, so we have had to invent a great deal ourselves: the processes, the ideas, the tool chain, the testing systems, and how software is configured across multiple generations of silicon and multiple factories. There was no supplier solution we could simply buy. So we created it ourselves."

That admission of mistakes, made without apparent discomfort, is among the more revealing aspects of Bell's account. It suggests a culture that has normalised learning from failure rather than concealing it, a disposition that is rare in large organisations and essential for the kind of long-horizon engineering that Volvo has undertaken. Fail-fast, Learn-fast, Iterate-fast. That epistrophe seems to capture the Volvo philosophy in its approach to technology.

The test drivers who never sit in a car

The EX60 is not Volvo's first software-defined vehicle. It is the third, following the EX90 and ES90, all built upon the same software master. That lineage materially changes the risk profile of the EX60's software foundation in ways that comparisons with first-generation attempts at software-defined architecture typically do not acknowledge.

Asked directly whether EX60 customers would be serving, in effect, as test drivers for software still finding its maturity, Bell was frank. "No, the customers will not be the test drivers. I am the test driver. The machines are the test drivers as well. I am currently driving the EX60 myself."

He went on to explain the architectural continuity underpinning that confidence. Volvo are starting production with a very solid software foundation and enough time to refine everything before customer deliveries. It is also important to understand, he points out, that the software itself is not entirely new. It is fundamentally the same software platform used in the EX90 and ES90. “The EX60 is the third car built from the same software master," he says.

Read the article

Megacasting: How Volvo Torslanda casts an EX60 floor in two minutes

At Torslanda, a single aluminium casting replaces hundreds of stamped and welded components on the EX60. Peter Schüler, Megacasting Shop Manager, explains the physics, the process, and the industrial ambition behind Europe's most advanced high-pressure die-casting cell.

Yet the development model Volvo uses operates across, rather than within, individual model lines. "That does not mean the code is identical, but it comes from the same core software repository. For example, if an engineer works on the door-locking system, they do not release code specifically for the EX90. They release it to the software master, and from there the software is configured appropriately for each model."

The infotainment system itself is a case in point. "It is effectively the same software in the EX90 and EX60,” says Bell. “It is already proven and hardened. Some of the difficulties we experienced previously are now behind us because this is the third vehicle developed from the same software master."

Yet Bell is equally clear that no software-defined vehicle is finished at the point of delivery. The real world is a more inventive generator of edge cases than any test environment - a motif that keeps resurfacing across Volvo’s expert insights. "Absolutely,” he says. “It has been a difficult learning experience for us. The number of corner cases the real world throws at you is immense." Even this is an insight. A corner case, as compared to an edge case, is generally a rare, extreme scenario that occurs only when multiple operating parameters or conditions are pushed to their absolute limits simultaneously. Volvo is in the process of mastering these too.

Three factories, one carmaker

The word "factory" has a specific meaning in Bell's lexicon that ripples out beyond the physical buildings on the Torslanda site. He divides Volvo's engineering operation into three conceptually distinct entities, each with its own logic, rhythm and output.

"I divide our operation into three elements,” he outlines. “First, there is the vehicle creation factory, which develops new vehicles on the tech stack. Creating a new car on the SPA platform takes around two and a half years from concept to launch. Second, there is the technology factory, which develops next-generation technologies across the stack: drive units, battery chemistry, battery management system, power electronics, airbags, and so on. Those technologies are then introduced into future vehicles. Third, there is the software factory. This is unique because it delivers software to everything simultaneously. The same release goes out to factories, customer vehicles, pilot vehicles, and even future model years."

The third of those ‘factories’ is, in operational terms, the most novel. It produces output that, unlike a physical vehicle or a new battery chemistry, can be distributed instantly and at near-zero marginal cost to vehicles already in customers' hands. This of course, creates an entirely new category of product liability, quality management and release engineering.

"This creates a major shift for the automotive industry. Traditionally, a vehicle passed through the factory, quality gates and industrial validation before reaching customers. Now we also deliver digital products directly to customers through software updates."

Read the article

Torslanda: Inside Volvo’s most ambitious factory transformation

Magnus Olsson has launched seven Volvo models. As Torslanda's plant manager, he now oversees the factory's most complex transformation, and makes the case that people, not machines, will determine whether Western manufacturing prevails.

The automation underpinning that delivery is itself a significant engineering achievement. It also collapses the traditional separation between vehicle development and post-sale operations, forcing engineering teams to treat deployment velocity, systems validation and lifecycle monitoring as core competencies rather than downstream support functions.

"Our automated testing evolves continuously,” he says. “Before any code reaches a physical vehicle inside our walls, it must pass through rigorous testing before it can enter the software master. Everything sits within a single software repository covering SPA2 and SPA3."

He continues, "we create 20 complete vehicle software builds every day - almost one per hour. Every four hours we select the best build and run a complete automated vehicle test to ensure there are no regressions. This process runs continuously, 24/7."

Customers, says Bell, will typically receive four meaningful over-the-air updates per year, roughly one every quarter. Meanwhile, internally, Volvo is moving extremely fast around the clock.

Over-the-air at scale and the 99.9% imperative

The gap between those two tempos - four updates per year visible to customers against one build per hour running internally - is where Volvo's quality engineering does most of its work. The tolerance for failure at the point of delivery to customers is vanishingly small.

"We are currently rolling out our One HMI updates to more than 2 million vehicles built as far back as 2020. The success rate for over-the-air updates must be above 99.9%, and the remaining failures must be benign - for example, a failed update that simply requires a retry," he says.

And the consequences of falling short of that threshold are far from abstract. Bell says you obviously cannot have failures that immobilise a vehicle. “Even 0.1% is too many when you are dealing with millions of vehicles. Some cars may have damaged wiring, rodent damage, weak 12-volt batteries or previous crash damage. We monitor deployments extremely carefully, rolling them out incrementally and constantly analysing the data."

Yet the number of people responsible for this operation would surprise most observers. Bell declined to give a precise figure, but offered a revealing indication of scale. "I will not give an exact number, but it is smaller than most people think. It is a four-digit number - but not a very high one. It also depends on where you draw the line. There are the people writing code, but also those involved in quality assurance, integration and coordination."

The test driver's verdict

Bell is not a passive observer of the vehicle he has helped to build. He drives the EX60 personally, has done so through the streets of central Stockholm and speaks about the experience with an enthusiasm that is, for a senior executive, markedly unguarded.

"It is fantastic,” he says. “The day after the reveal on January 22, I drove the car through central Stockholm without camouflage for the first time. That was a very special experience, especially given the importance of this launch for the company and the number of Volvo employees seeing the car in public for the first time."

On its driving qualities, he is specific rather than promotional. "The vehicle itself is excellent. It has near-perfect 50:50 weight distribution. I have driven both the P6 and P10 variants extensively. The P6, which is the lower-spec version, is already extremely good - almost surprisingly good. Rear-wheel-drive EVs are genuinely fun to drive."

The reference to the ES90 as a regular driver, paired with the assertion that he does not miss all-wheel drive, is the kind of statement that carries weight only when the person saying it has a clear incentive to say the opposite. "I also drive the ES90 rear-wheel-drive and honestly do not miss the all-wheel drive. The P10 takes performance to another level, while the P12 is completely over the top."

The verdict, when it finally arrives, is delivered with the same economy that characterises Bell's approach to engineering timelines.  a valuable reformulation: Establish the proof of concept. Understand what you have built. Then, and only then, commit to the product.

"It is a fantastic car. And yes, I really do like it."