Rivian Invests Hundreds of Millions in Custom AI Chips

Rivian is pouring 'hundreds of millions of dollars' into developing its own custom silicon for autonomous driving, CEO RJ Scaringe revealed in a recent interview. The electric vehicle maker's first in-house chip — internally known as the Rivian Autonomous Processor or RAP-1 — is set to debut later this year, marking a bold entry into the increasingly competitive custom chip race among automakers.

The move positions Rivian alongside Tesla, which has long championed its own in-house chip development, and signals a broader industry trend where EV companies seek to control their entire technology stack from software to silicon.

Key Takeaways

• Rivian is investing 'hundreds of millions of dollars' in a dedicated custom chip program
• The first chip, RAP-1, will launch in 2025 and is built on TSMC's 5nm process
• Successor chips RAP-2 and RAP-3 are already in the roadmap with more advanced fabrication technology
• New chip generations are planned 'every few years'
• Rivian has assembled a dedicated team for ongoing chip development
• The investment is described as an ongoing commitment, not a one-time expenditure

RAP-1 Arrives as Rivian's First Custom Autonomous Driving Processor

The RAP-1 chip represents Rivian's inaugural foray into custom silicon design. Built on TSMC's 5nm process node, the processor is specifically engineered for autonomous driving workloads — a domain where computational demands are immense and growing rapidly.

TSMC's 5nm technology is the same fabrication process used in Apple's A15 Bionic and Qualcomm's Snapdragon 8 Gen 1 chips, offering an excellent balance of performance and power efficiency. For an automotive application, this matters enormously: autonomous driving processors must deliver massive computational throughput while operating within the thermal and power constraints of a vehicle.

Scaringe did not disclose the specific performance benchmarks of RAP-1 or how it compares to existing solutions from NVIDIA, whose Drive Orin and Drive Thor platforms currently dominate the autonomous driving chip market. However, the decision to go custom suggests Rivian believes off-the-shelf solutions cannot fully meet its software and hardware integration requirements.

Why Automakers Are Racing to Build Their Own Chips

Rivian's chip investment reflects a seismic shift in how automakers think about technology. Traditionally, car companies relied on Tier 1 suppliers like Bosch, Continental, and Mobileye for their electronics and computing needs. That model is rapidly breaking down.

Tesla pioneered the automaker-as-chipmaker approach with its Full Self-Driving (FSD) computer, first deployed in 2019. The chip, designed in-house and manufactured by Samsung, gave Tesla tight control over its autonomous driving hardware-software stack. The results have been compelling enough to inspire competitors.

Several factors are driving this trend:

• Performance optimization: Custom chips can be tailored to specific AI models and sensor fusion algorithms, eliminating wasted silicon area
• Cost control: At scale, in-house chips can be significantly cheaper than buying from NVIDIA or Qualcomm
• Supply chain independence: The global chip shortage of 2021-2023 exposed the risks of relying entirely on external suppliers
• Competitive differentiation: Proprietary hardware creates a technology moat that competitors cannot easily replicate
• Software-hardware co-design: Tighter integration between chip architecture and software stack can unlock performance gains impossible with general-purpose processors

Beyond Tesla, companies like NIO in China and even traditional automakers like BMW and Mercedes-Benz have explored custom or semi-custom chip programs. Rivian's commitment of 'hundreds of millions' underscores just how seriously the company views this as a strategic imperative.

A Long-Term Commitment, Not a One-Time Bet

Scaringe was careful to frame Rivian's chip program as an ongoing investment rather than a finite project. 'This is not like investing a few hundred million dollars and calling it a day,' the CEO said. 'We have assembled a team. That team will continue to develop future versions of this platform.'

This language is significant. Chip development is notoriously expensive and requires sustained investment over many years. A single generation of custom silicon can take 3-4 years from initial design to mass production, and the technology moves so quickly that companies must begin work on the next generation before the current one even ships.

Rivian has already mapped out its roadmap with RAP-2 and RAP-3 as planned successors to the first-generation chip. These future processors will use fabrication technologies 'more powerful' than the 5nm process used for RAP-1 — likely TSMC's 3nm or 2nm nodes, which promise significant improvements in transistor density, performance, and energy efficiency.

The 'every few years' cadence Scaringe described aligns roughly with the semiconductor industry's process node advancement timeline. TSMC's 3nm process is already in volume production, and 2nm is expected to ramp up in 2025-2026.

The Competitive Landscape: NVIDIA, Qualcomm, and Mobileye

Rivian's custom chip push comes at a time when the autonomous driving chip market is intensely competitive. Here is how the major players stack up:

• NVIDIA dominates with its Drive platform, offering the Drive Orin (254 TOPS) and the upcoming Drive Thor (2,000 TOPS), used by companies like Mercedes-Benz, BYD, and Volvo
• Qualcomm has entered the automotive space aggressively with its Snapdragon Ride platform, securing deals with General Motors and BMW
• Mobileye (an Intel subsidiary) continues to serve as the go-to solution for many traditional automakers with its EyeQ line of vision processors
• Tesla remains the benchmark for in-house automotive chip design, with its next-generation AI5 chip reportedly in development
• Huawei has emerged as a major player in China's autonomous driving chip market with its Ascend series

By developing RAP-1, Rivian is essentially betting that it can match or exceed the performance of these established players — at least for its own specific use case. The advantage of a custom chip is that it does not need to be a general-purpose solution. It only needs to run Rivian's software stack exceptionally well.

Financial Implications for a Cash-Conscious EV Startup

Spending 'hundreds of millions' on chip development is a significant commitment for Rivian, which reported approximately $9.4 billion in cash reserves at the end of 2024 but continues to burn through capital as it scales production. The company delivered around 51,579 vehicles in 2024 and is working to achieve profitability.

However, the long-term economics of custom silicon could work in Rivian's favor. NVIDIA's automotive-grade processors are not cheap, and licensing fees for autonomous driving platforms can add $1,000 or more per vehicle. At scale — if Rivian reaches its production targets of hundreds of thousands of vehicles annually — the per-unit cost savings of an in-house chip could be substantial.

There is also the Volkswagen factor. VW invested $5.8 billion in a joint venture with Rivian in 2024 to share the EV maker's electrical architecture and software platform. If Rivian's custom chips become part of this shared platform, the cost could be amortized across a much larger vehicle fleet, dramatically improving the return on investment.

What This Means for the Autonomous Driving Industry

Rivian's chip announcement carries broader implications for the autonomous driving ecosystem. It signals that the era of one-size-fits-all computing platforms for self-driving cars may be ending. As AI models become more sophisticated and vehicle architectures more complex, the pressure to optimize at the silicon level will only intensify.

For NVIDIA and Qualcomm, this trend represents a potential long-term threat. Every automaker that goes custom is a customer lost — not just for one chip sale, but for an entire generational roadmap of products and services.

For consumers, the implications are more nuanced. Custom chips could lead to better-performing autonomous driving systems, faster over-the-air updates, and more efficient vehicles. But they could also create proprietary lock-in, making it harder for third-party developers to build compatible applications.

Looking Ahead: Rivian's Silicon Roadmap

With RAP-1 set to arrive later this year, the next major milestones to watch will be its integration into Rivian's production vehicles and any performance benchmarks the company chooses to share. The transition from prototype to mass production is where many chip programs stumble, and Rivian will need flawless execution to justify the investment.

The development of RAP-2 and RAP-3 on more advanced process nodes will be equally critical. Moving to 3nm or 2nm fabrication is not trivial — it requires deep partnerships with TSMC and significant design expertise. Rivian's ability to attract and retain top semiconductor talent will be a key factor in the program's success.

One thing is clear: Rivian is no longer content to be just an automaker. By investing hundreds of millions in custom AI chips, the company is declaring its ambition to be a full-stack technology company — one that designs everything from the atoms in its processors to the algorithms running on them. Whether that ambition pays off will depend on execution, timing, and the relentless pace of innovation in both the EV and semiconductor industries.