Bot Auto CEO Xiaodi Hou on why cost-per-mile is the key to autonomous trucking

Following our recent coverage of the challenges facing autonomous trucking, Robotics and Automation News now interviews Xiaodi Hou, co-founder and CEO of Bot Auto, to explore a different approach – one focused less on flashy demos and more on cost-per-mile performance.

Hou, a veteran in autonomous driving and AI, is taking aim at the industry’s reputation for over-promising and under-delivering.

With a PhD in Computation and Neural Systems from Caltech and a background in machine learning and computer vision, he brings both technical depth and commercial realism to the table.

Bot Auto isn’t pitching autonomy as a distant dream – it’s already running regular autonomous freight trips between Houston and San Antonio, with a safety driver on board.

The company plans to go fully driverless in 2025 and is betting on a Transportation-as-a-Service (TaaS) model, operating its own fleet rather than licensing its software. The reason? Freight customers want capacity, not complexity.

In this interview, Hou explains why cost-per-mile is the ultimate success metric, how Bot Auto uses transformer-based AI models, and what it will take to scale. His message is clear: execution – not just innovation—is what will make autonomous trucking real.

Robotics and Automation News: Your career has been deeply rooted in autonomous vehicle technology. What was the pivotal moment or realization that led you to shift your focus from online advertising to autonomous trucking?

Xiaodi Hou: It is a misconception that my transition was from online advertising. My career has consistently focused on computer vision and AI, from my PhD in Computation and Neural Systems at Caltech to co-founding another autonomous trucking company in 2015.

The pivotal moment was realizing that autonomous trucking could solve real-world problems: driver shortages, supply chain inefficiencies, and safety issues.

Unlike consumer applications, trucking has clear economic incentives and operational constraints that make autonomous technology not just possible, but necessary.

R&AN: Bot Auto is positioned as a “trucking company that uses autonomous technology” rather than a technology vendor. What is the strategic thinking behind this “Transportation as a Service” (TaaS) model, and how do you believe it will give you an edge in the current market?

XH: The “Transportation as a Service” (TaaS) model addresses a fundamental market reality that freight customers want transportation capacity, not technology complexity. By operating our own fleet, we control the entire value chain – from software development to operational efficiency.

This allows us to iterate rapidly based on real-world feedback and capture the full economic value of our innovations. We’re not transferring operational challenges to customers; we’re solving them directly and delivering reliable transportation capacity.

R&AN: You’ve emphasized the importance of “cost-per-mile” (CPM) as a key metric for Bot Auto. Could you elaborate on why this is a more critical indicator of success than other metrics like “miles per intervention,” and how it shapes your technology and business decisions?

XH: CPM is the ultimate measure because it captures the only thing that customers actually care about: total transportation cost. “Miles per intervention” is a technology metric that doesn’t translate to business value.

A truck with perfect technical performance that costs $3 per mile is commercially useless if human drivers cost $2. CPM forces us to optimize the entire system – technology, operations, maintenance, and efficiency – for real economic viability rather than impressive technical demonstrations.

R&AN: The autonomous trucking industry has faced skepticism due to what you’ve described as “over-promising and under-delivering.” How is Bot Auto working to rebuild credibility, and what concrete steps are you taking to ensure you meet your stated goals and timelines?

XH: The hype cycle needs to end. Bot Auto is driving innovation and rebuilding industry credibility through transparent execution and realistic timelines. Instead of grandiose promises, we demonstrate tangible progress: from stealth to hub-to-hub demos in three months, now moving to driver-out operations in late summer 2025.

We publish our safety approach, engage with first responders, and focus on solving operational challenges rather than chasing headlines. Our goal is to under-promise and over-deliver, proving autonomous trucking works through consistent execution.

R&AN: Bot Auto utilizes transformer-based neural networks. Can you explain in layman’s terms for our audience at RoboticsAndAutomationNews.com how this AI approach differs from and potentially improves upon other technologies used in autonomous driving?

XH: Think of traditional autonomous driving as following a rigid recipe with each step programmed separately. Our transformer-based approach is more like having an experienced chef who understands ingredients holistically.

The system processes all sensor data simultaneously, creating a unified “world model” that can handle unexpected situations. Instead of separate modules for perception, planning, and control, we have one foundation model that learns patterns across all driving tasks, making it more adaptable and reliable.

R&AN: You’re starting with a single truck route between Houston and San Antonio. What are the key challenges you’re focusing on solving in this initial phase, and what is your roadmap for scaling up your operations?

XH: Our Houston-San Antonio route focuses on three critical challenges: system reliability in real traffic conditions, operational efficiency to achieve competitive costs, and safety validation with real cargo.

We’re building our operational playbook – how to maintain trucks, handle edge cases, and integrate with existing logistics networks. The roadmap scales by replicating this proven operational model to a larger fleet then new routes, rather than reinventing technology for each expansion.

R&AN: You’ve mentioned that an “ultimate guideline” tied to profitability is crucial. Beyond CPM, what other key performance indicators are you tracking to ensure Bot Auto’s long-term success?

XH: Beyond CPM, we track operational availability (uptime percentage), safety metrics (incidents per mile), customer satisfaction scores, and operational efficiency improvements over time. We also monitor our capital efficiency which tells us how quickly we can deploy new trucks profitably.

The ultimate guideline is simple: can we operate autonomous trucks more cost-effectively than human drivers while maintaining superior safety and reliability? Every KPI ladders up to this fundamental business viability question.

R&AN: The relationship between autonomous trucks and human drivers is a topic of much discussion. What is your vision for how the trucking workforce will evolve with the adoption of autonomous technology, and what role will human drivers play in the future of logistics?

XH: Human drivers won’t disappear, their role will evolve. Autonomous trucks excel at long-haul highway driving, while humans remain superior for complex urban deliveries and customer interaction.

I envision a hybrid model where autonomous trucks handle the middle-mile efficiently, while human drivers focus on higher-value first and last-mile services. This creates more specialized, better-paid driving jobs while solving the long-haul driver shortage that’s constraining our supply chain.

R&AN: With your experience, you have a unique perspective on the future of mobility. What do you believe will be the most significant bottleneck to the widespread adoption of autonomous trucking in the next five to ten years – technology, regulation, public perception, or something else entirely?

XH: The biggest bottleneck isn’t technology or regulation; it is economic viability at scale. Many companies are trapped in resource-intensive approaches that require massive capital without clear paths to profitability.

The real challenge is proving that autonomous trucks can operate more cost-effectively than human drivers while maintaining safety.

Once we demonstrate sustained profitable operations, adoption will accelerate rapidly because the economic incentives are undeniable. It’s about execution, not innovation.