Hou Xiaodi's Latest Post: CPM Is the Key to Separating Real Autonomous Trucking from Hype | Linear Voice

After a decade in autonomous driving, Xiaodi Hou has gone from CTO to CEO, and from "technology-first" to "operations-first" — searching for the balance between technical capability and business model after living through the industry's highs and lows.

Recently, Hou published an essay sharing Bot Auto's latest thinking on industry practice, arguing that CPM (Cost Per Mile) is the key metric separating hype from real commercialization in autonomous trucking. He emphasizes that simply achieving "driverless" is nowhere near enough — autonomous trucks must also be cheaper, more efficient, and more scalable than existing models. As long as CPM remains higher than traditional fleets, autonomous driving will remain nothing more than an expensive science experiment.

Linear Capital was a co-lead investor in Bot Auto's Pre-A round. We've translated and compiled the essay below.

"Eating hot pot and singing songs, while a 30-ton semi-trailer cruises day and night between coasts under AI command, revolutionizing logistics efficiency." — This rosy vision, proposed a decade ago, remains unrealized.

Instead, the road to autonomy has been paved with missed deadlines, runaway R&D spending, and pilot projects too costly to scale.

What's the root cause? Few companies have truly grasped that CPM (Cost Per Mile) is the industry's key metric. In my view, CPM is what separates industry hype from genuine commercialization. For autonomous driving, simply achieving "no driver in the vehicle" is far from sufficient — it must also be cheaper, more efficient, and more scalable than existing models. As long as CPM remains above that of traditional fleets, autonomous driving will remain an expensive science experiment. In this industry, CPM is not an optional footnote — it is the scoreboard for the entire game.

The point of exploring autonomous truck technology is not to replace drivers, but to fundamentally reshape the economics of traditional freight. BCG notes in its latest white paper: "Only what gets measured can improve. If autonomous trucks are to compete with conventional methods at scale, the industry needs a standardized way to measure true costs and operational impact."

The Cost Per Mile Framework offers a meaningful way to evaluate different technologies and business models, creating a relatively standardized, comprehensive metric that covers nearly all relevant operating costs.

As someone in this industry, I've come to deeply appreciate how we used to charge ahead pursuing technical success while ignoring the logic of how business actually works. The autonomous driving industry keeps falling into the "resources-first" trap, making promises it can't keep. This mindset assumes that simply pouring in more capital, engineers, and compute power will inevitably drive the industry forward.

Yet proceeding down this path long enough inevitably leads to diminishing marginal returns. Investment keeps climbing while output steadily declines. Meanwhile, we keep deluding ourselves into treating tiny, incremental improvements as justified returns on massive spending.

I've stepped into plenty of these pitfalls myself. But after setbacks and constant reflection, I believe it's time for the autonomous driving industry to stop packaging grandiose concepts and return to the fundamental starting point of "product value."

BCG's report emphasizes that a comprehensive CPM calculation must include the following elements:

• Driver cost
• Fuel cost
• Conventional operating cost
• Autonomous operating cost
• Vehicle hardware cost
• Infrastructure / facility cost
• Other overhead cost

The conversation around CPM has already begun. We believe the only viable path forward for this industry is to bravely confront all the real-world challenges of autonomous freight. For companies, the goal shouldn't be peddling simplified, fuzzy numbers, but doing the complex math behind the economics for our customers.

BCG's framework aligns closely with my own thinking. The report notes that "the cost impact of autonomous trucks is far from a simple trade-off between saved driver wages and increased vehicle costs. Rather, its impact permeates every aspect of operations." This includes autonomous tech maintenance, remote assistance operations, last-mile delivery costs, and roadside breakdown rescue.

Honest accounting must capture all operating costs, not just the visible ones. This includes all direct and indirect costs stemming from the complexity of autonomous systems. Whitewashing this complexity doesn't help the industry solve problems — it merely postpones challenges, embedding seemingly minor operational obstacles deep into system design that eventually become intractable cost burdens down the line. We call on the entire industry to reach consensus on calculation methods that reflect economic reality, so that any accounting firm analyzing the same operating entity would arrive at the same net profit figure.

After a decade in this industry, I've come to understand that the complex business of autonomous trucks boils down to a simple formula:

• Product value = Driver Cost (saved) - Operating Cost

There are mainly two types of players in this market: those selling software and hardware technology (SaaS/HaaS), and those selling transportation services (TaaS). Technology vendors tend to emphasize only the "driver cost saved" while offloading the real "operating costs" onto customers. This easily leads to abandoned products and dead ends for companies.

As a TaaS provider, Bot Auto's model aims to minimize operational disruption for customers by taking on the associated challenges ourselves: doing the full economic math and relentlessly driving down our own CPM to deliver complete services with genuine economic value.

At Bot Auto, our driverless program isn't for show. Today, we're already moving real cargo with real trucks, because throughout this process we keep asking ourselves two questions:

1. Will customers actually use our product? The only way to answer this is through real-world practice — no simulation can provide the answer.

2. Can we achieve profitability? Only real road-mileage data can fully reveal CPM's true impact.

We're using real-world evidence to answer these questions and continuously build customer trust. Customers aren't looking to buy technology itself — they want products that balance system safety with economic advantage. In trucking, customer confidence is built on CPM. Every shipment we move is optimizing this number. We believe every cent of cost reduction pushes autonomous logistics transport from novelty toward industry standard.

A common misconception in the autonomous trucking industry is that profitability requires a massive, rapidly expanding fleet. Our data proves otherwise.

Based on CPM modeling, we've identified two milestone figures: a fleet of just 30 autonomous trucks on specific high-value freight routes can reach break-even in just two years; to surpass the average human-driven fleet CPM of $2.25/mile requires roughly 100 autonomous trucks, not thousands. This is the path from "technology vaporware" to "commercial reality" for autonomous trucks.

The autonomous trucking industry should adopt standardized CPM calculation methods to enable fair comparison, reduce resource misallocation, and set reasonable expectations for the public. As BCG's report states, "Industry-wide adoption of this CPM framework would increase transparency, encourage fair comparison between technologies and business models, and drive continuous improvement in cost and performance."

I strongly agree. This isn't just about transparency — it's about whether the autonomous driving industry can break free from the shackles of "resources-first" thinking. That rigid dogma has trapped the industry in a vicious cycle of diminishing marginal returns.

The best way to solve the autonomous driving industry's chronic problem of overpromising and underdelivering is for all players to commit to being accountable for results together. The name of that commitment is Cost Per Mile.