John Hayes is the founder and CEO of Ghost Autonomy, the self-driving AI software platform for the mainstream consumer automotive market.
When Apple introduced the iPhone, it never dreamed of an app like Uber emerging. What will the killer apps for the future car look like, and who is going to make them?
As the software-defined car takes shape before our eyes, it is clear that, beyond autonomy, the automotive software opportunity is wide open and largely undefined in terms of scope and potential applications. While copying existing iPhone apps for automobiles may not be the right answer, copying Apple’s open software development model may well be the right way to build the next trillion-dollar car company.
Software is easy to talk about but hard to build. Harder still is figuring out what to build, especially for consumer applications. Consumers come with capricious whims, and with that comes risk. Like Hollywood movies, software can be expensive to produce, and it is hard to predict exactly what people want.
Auto companies have struggled to develop software over the past decade, both in terms of execution and customer adoption. A July 2023 J.D. Power study found that consumers are largely disappointed with the new technology features in their cars, contributing to an overall decline in owner satisfaction in consecutive years. Internal efforts have been mixed at best, with a number of high-profile failures. Tesla is still years—if not decades—ahead of legacy automakers.
While it’s easy to point to the fact that software is outside of automotive’s historical expertise, software is actually also hard for software companies to develop—even the giants of Silicon Valley. Look no further than Meta, which built a disruptive consumer product itself 20 years ago on Facebook. Today, the business significantly relies on products and features it has bought (Instagram, WhatsApp) or copied (short-form video a la Snapchat and TikTok). The fact is that most large companies struggle to innovate, especially in software.
Silicon Valley has developed two solutions to this problem, both relying on external developers for innovation: venture capital (VC) and APIs.
• Venture capital. The largest companies outsource the failure to develop new products into the VC/startup system. This allows them to avoid painful strategy changes and reorganizations that would slow their growth.
Venture capitalists fund new software companies to build the next generation of applications. Some startups ultimately become standalone companies, but many are bought by large incumbents and sold as new product lines or features. YouTube, Instagram, WhatsApp, Beats, Android, Nest and Ring are all household names that started as venture-backed standalone companies and eventually became product lines at larger companies.
• APIs. In this model, like Apple’s App Store, the large incumbent corporations build an open platform that allows new applications to access their user base. Critically, this model typically pushes the go-to-market burden on the app developers, pushing them to market and service their own apps.
Practically, and financially, this operates more like a tax. The platform/corporation has much less involvement in the ownership and delivery of each app’s specifics, allowing many more different kinds of apps to exist. APIs also serve to activate VCs. New APIs (with the promise of stability) have proved to create a flurry of new companies.
Re-Architecting Cars For Open Software Development
To unlock a developer ecosystem, cars will need to adapt to a new open architecture—with a few significant technical changes.
Open API
Car companies must commit to building and servicing a real API in the mold of major Silicon Valley platforms with these core tenets.
• Self-service. Open development that does not require entering into a contract to develop an app.
• Open documentation. An API that makes specific promises about functionality, access, data structures and requirements. (This is hard and requires sophisticated thinking.)
• Long-term stability. Once documentation and apps are published, car companies must be willing to support apps built this way for at least a decade.
Open Access To The Car
Car companies must make the entire car accessible by software—and by software developers. This requires enabling direct access to core components like data feeds from sensors. This also requires reducing or eliminating entirely specialized components or modules that are walled off from developer input and innovation. After all, the surface area available to outside developers is proportional to the size of the overall software opportunity.
Connectivity
Car companies must ensure that the car has connectivity that will enable interoperability with the cloud and access to the next generation of data center applications. Cars are expected to be on the road for more than 10 years. While processing capabilities in the car might be fixed for any given model year, the cloud will continue to evolve.
Large language models (LLMs) and visual language models (VLMs) are the latest data center applications that can produce extraordinary results, but only with extraordinary compute resources. While the specific use cases of these models for cars are still being developed, it’s clear there will always be more powerful applications available in the data center. The software-defined car will need to access the data center to run the next generation of applications, which makes connectivity essential.
The Fast Lane To Innovation And Profits
Software is clearly the new chassis on which future vehicles will be built. Traditional automakers need to figure out software to compete with technology-forward upstarts across the globe. By embracing an open software development model with APIs, they should be able to fuel innovation at low effort and cost while bringing a dynamic ecosystem of valuable, interconnected applications to market—inspiring consumers to buy into a new vehicle experience.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Read the full article here
