What Autonomous Vehicle Companies Get Wrong About Freight Sales

Primary Framework: Vendor Economics Framework · Automation Failure Framework Hub: Insights Decision Question: Why aren’t our technology milestones converting to freight contracts, and what does the freight buyer actually need to see before signing? Evidence Window: 2020-2026 Author: Deepak Gupta, Founder & Principal Analyst, Autonomy Bridge

Core Question

One on-road autonomous vehicle company has contracted commercial freight revenue at scale. That company is Gatik, with $600 million in contracted revenue from Walmart, Kroger, and Fortune 50 retailers in Texas, Arkansas, and Arizona. [C4] (Sourced fact)

Every other AV company with freight ambitions has not reached this threshold.

Autonomy Bridge’s sector research characterizes the market precisely: “AV graveyard is massive (Argo, TuSimple, Embark dead). Survivors bifurcated: Gatik ($600M contracts, middle-mile) winning, robotaxi marginal, long-haul pre-revenue.” [C3] (Sourced fact , primary research)

The question this article addresses is not which company will be next. It is what the freight buyer actually requires to sign a contract , and why the technology milestones AV vendors present do not satisfy that requirement.

The answer is in the gap between what vendors demonstrate and what buyers evaluate. AV vendors demonstrate driverless miles, operational design domain expansion, and technical readiness benchmarks. Freight buyers evaluate cost-per-mile against their current asset, safety record on commercial operations, and carrier qualification history. These are not the same metric. They cannot be converted by appending a commercial slide to a technology briefing.

Autonomy Bridge’s primary research across 23 on-road AV companies identifies five commercial failure patterns: sales cycle stall (8 companies), channel constraint (5), pilot-to-scale failure (4), pricing mismatch (4), and qualification failure (2). [C1] Sales cycle stall is the largest category , which means the buyers are engaging, evaluating, and then not signing. The problem is not awareness or interest. It is the content of what vendors present when buyers reach the qualification stage.

Why the Question Matters Now

The AV sector raised the majority of its capital between 2020 and 2022 on technology development timelines. Those timelines have mostly not produced the commercial outcomes projected at the time of investment. The capital cycle is creating urgency: companies that raised $500 million to $1 billion-plus between 2020 and 2022 are approaching or past the point where capital must convert to contracted revenue.

The current evidence shows a widening gap between technology progress and commercial progress.

Waabi. Waabi raised $750 million-plus in January 2026, led by Khosla Ventures, at over $1 billion raised total. The company achieved surface-street autonomy in 2025. The Volvo VNL Autonomous truck platform, through which Waabi plans commercial deployment, requires Volvo validation that has not yet completed. Driverless commercial operations are delayed pending that validation. [C5] (Sourced fact) The capital is real. The commercial operations are not yet.

Kodiak Robotics. Kodiak has 10 driverless trucks hauling frac sand for Atlas Energy Solutions in the Permian Basin, operating around the clock. The company reports autonomous reliability metrics (ARM) at 78% for long-haul highway operations. [C6] (Sourced fact) The Permian Basin oilfield deployment is commercially operating. Long-haul highway trucking , the category the company was founded for , is at 78% readiness with a driverless launch planned for H2 2026. The company has pivoted its near-term revenue strategy to the oilfield environment specifically because that environment removes the open-road carrier qualification barrier. The pivot is commercially rational. It is also an acknowledgment that the qualification barrier for highway freight exists and is not yet cleared.

Udelv. Udelv announced 1,000-plus reservations and a 35,000-unit production plan between 2021 and 2022. The company has no visible commercial deployments as of 2026. No significant news has appeared in public searches in 2025 or 2026. [C10] (Sourced fact , absence of evidence) The reservation-to-deployment gap is 4-plus years with no public progress update.

Conigital. Conigital entered voluntary liquidation in February 2025 after failing to secure sustainable funding. The company had claimed a GBP 500 million investment in 2023 that was later found to be unverified. Twenty-two employees were affected. [C11] (Sourced fact)

Stack AV. Stack AV launched from stealth in September 2023 with SoftBank backing, founded by the leadership team from Argo AI. As of April 2026 , two and a half years after launch , the company has no announced customers, no announced deployments, and no publicly stated business model. [C2] (Sourced fact) SoftBank’s backing provides runway. The absence of any commercial announcement after two and a half years in a sector where technology is not the limiting factor is the relevant commercial signal.

These cases are not a pattern of technology failure. In most of them, the technology is progressing. They are a pattern of commercial model failure , the inability to translate technology progress into contracted buyer commitments at a pace that sustains the business.

What the Evidence Shows

Autonomy Bridge’s primary research covers 23 on-road AV companies. The problem code distribution is: [C1] (Sourced fact , primary research)

(Autonomy Bridge proprietary analysis, 2024-2026) [C1]

Sales cycle stall is the dominant pattern. Eight companies are engaging buyers, running evaluation processes, and not closing. This is different from the channel constraint pattern (where buyers are not reached) or the qualification failure pattern (where the wrong buyer is targeted). These companies are in the room with the right buyers and not converting.

Autonomy Bridge’s sector research identifies three specific failure patterns for the on-road sector that explain the stall: [C3] (Sourced fact , primary research)

1. Freight customers qualify carriers on safety history. New AV carriers have none.
2. Fixed-route success doesn’t generalize. Gatik’s Walmart playbook hasn’t replicated.
3. AV founders speak technology. Freight buyers speak cost-per-mile.

These three patterns are not separate problems. They are different expressions of the same underlying gap: the evidence package AV vendors present is structured around technology readiness, while the evidence package freight buyers require is structured around operational qualification.

Gatik’s commercial position is the sector’s primary benchmark. Gatik operates fully driverless middle-mile autonomous trucks for Walmart, Kroger, and Fortune 50 retailers in Texas, Arkansas, and Arizona. The company has completed more than 60,000 driverless orders, accumulated 10,000-plus driverless hours and 1,000-plus driverless miles, and contracted $600 million in total revenue , including $400 million in new contracted revenue in the second half of 2025. The fleet is scaling from 10 to 60 trucks. [C4] (Sourced fact)

The operating characteristics that produced those contracts are specific: fixed routes, short-haul, B2B shipper relationships, and geofenced operations that define a bounded operational envelope. The freight buyers at Walmart and Kroger could evaluate a cost-per-mile comparison against their current carrier costs on those specific routes. The geofenced operation produced a safety record on commercial operations , not test miles, but orders delivered. [C4] (Sourced fact; inference on evaluation criteria is reasoned)

The companies with working commercial models share the fixed-environment characteristic.

EasyMile pivoted away from open-road passenger shuttle operations to autonomous industrial towing and airport logistics , TractEasy vehicles deployed at BMW Group’s Dingolfing plant, Daimler Truck, John Deere, Munich Airport, Narita Airport, and more than 30 commercial sites total, with 15-plus additional sites planned for 2026. [C7] (Sourced fact) EasyMile achieved the first Level 4 autonomous operations in Japan at Narita Airport. The pivot from open-road to fixed-environment is the commercial decision that produced deployments.

Oxa raised $103 million in a Series D in March 2026, including $50 million from the UK National Wealth Fund. The company’s research entry is explicit: it “deliberately pivoted away from open-road to industrial AV where deals close faster.” Port of Tyne quayside operations and a 14-vehicle deployment in Jacksonville are the commercial outcomes. [C8] (Sourced fact)

Embotech deployed autonomous vehicle marshalling systems across BMW Group’s Dingolfing, Leipzig, Regensburg, and Spartanburg factories , handling hundreds of cars per day, with six plants contracted by end 2025 , and completed a Port of Rotterdam 30-unit deployment. [C15] (Sourced fact) Factory and port operations are fixed-environment deployments where the buyer can evaluate operational reliability within a controlled boundary.

The pattern is consistent. Fixed-environment deployments close. Open-road deployments do not, yet. (Reasoned inference from named evidence)

Where the Market Is Commonly Misread

The standard AV commercial presentation leads with technology milestones. The number of driverless miles accumulated. The expansion of the operational design domain. The autonomous reliability metric percentage. The stack architecture advantages. None of these is the metric a freight buyer uses to evaluate a new carrier.

Autonomy Bridge’s sector research states the gap directly: “AV founders speak technology, freight buyers speak cost-per-mile.” [C3] (Sourced fact , primary research)

Misread 1: Technology readiness equals commercial readiness. A freight shipper evaluating an AV carrier does not ask how many driverless miles have been accumulated. They ask: what is the cost-per-mile on this lane, and how does it compare to my current carrier? They ask: what is your safety record on commercial operations, and how does it compare to the carrier I have today? If the AV vendor cannot answer both questions with operational evidence , not test data, not projected economics , the evaluation does not proceed to contract. (Sourced fact , primary research finding; application to individual companies is reasoned inference)

Misread 2: The safety history qualification barrier is bypassable. It is not. Autonomy Bridge’s sector research identifies the qualification failure pattern precisely: “Freight customers qualify carriers on safety history. New AV carriers have none.” [C3] (Sourced fact)

Freight carriers are qualified through the FMCSA Safety Measurement System and Carrier Safety Fitness Determinations in the US. A new carrier has no safety rating history. A carrier with a safety violation history is disqualified from shipper approved-carrier lists. An AV company entering the freight market as a new carrier has the same zero-history status as any new carrier , and must accumulate commercial operational history before major shippers will place it on preferred carrier lists.

This barrier cannot be addressed through a safety briefing about the technology’s design. It requires time on commercial routes, at commercial volume, generating safety record data in the FMCSA system. Gatik has accumulated that record. Companies that have only operated in test or demonstration environments have not. (Sourced fact on qualification system; carrier status of named companies is reasoned inference)

Misread 3: Technology bundling is a commercial advantage. Einride bundles autonomous capability with electric freight on a single platform. The research entry identifies the commercial constraint specifically: “selling autonomous-electric as one package means long cycles educating buyers on both technologies simultaneously.” [C9] (Sourced fact) A freight buyer evaluating Einride must simultaneously evaluate autonomous operations, electric powertrain economics, charging infrastructure requirements, and range limitations , each of which is an unfamiliar cost structure for a conventional diesel fleet operator. Every additional evaluation dimension extends the sales cycle.

Misread 4: Licensing the technology removes the freight qualification barrier. Nuro pivoted its entire business model from hardware to technology licensing in September 2024 after burning through delivery robot capital. The company raised $203 million at a $6 billion valuation in August 2025, with a partnership to deploy 20,000 Uber/Lucid robotaxis over six years. [C12] (Sourced fact)

Technology licensing transfers the product and the commercial model simultaneously. The licensee , Uber, Lucid, an automotive OEM , inherits the commercial responsibility for deployment. Whether the licensee’s go-to-market, cost structure, and carrier qualifications satisfy the end buyer’s requirements is now outside Nuro’s direct control. The licensing model is commercially rational as a capital recovery path. It does not solve the freight qualification problem , it transfers it. (Reasoned inference from Nuro model structure)

Misread 5: TIER IV open-source creates commercial scale. TIER IV’s Autoware platform is widely adopted in research institutions and by AV developers globally. The company raised $400 million-plus total, with a Series C, and is testing in Tokyo, Pittsburgh, and Munich. The research entry identifies the revenue constraint: “Companies use Autoware for free and only pay for enterprise support, making it hard to capture value from the platform’s growing ecosystem.” [C2] (Sourced fact) Platform adoption and platform revenue are different metrics. TIER IV’s Newmo partnership for Japan robotaxi is the commercial application that generates direct revenue. The open-source ecosystem generates influence, not contracted freight revenue.

Market Structure and Buyer Reality

On-road autonomous vehicles is not one commercial market. It is four structurally separate buyer segments with different qualification criteria, different procurement paths, and different revenue structures. A commercial strategy that does not specify which segment it is addressing will fail against each segment’s requirements simultaneously.

Buyer segment structure:

(Autonomy Bridge proprietary analysis, 2026) [C1][C2][C3]

The B2B fixed-route segment is the only closed commercial model. Gatik’s Walmart playbook , short-haul, geofenced, defined lane, B2B shipper , removed every open-ended qualification variable from the freight buyer’s evaluation. Walmart could evaluate a specific set of routes, compare against its current carrier cost on those routes, and observe commercial operational history on those routes before committing to fleet expansion. [C4] (Sourced fact; evaluation inference is reasoned)

The sector research notes that “Gatik’s Walmart playbook hasn’t replicated.” [C3] The commercial model is proven. The replication challenge is that each new shipper relationship requires a new route-specific safety record, a new route-specific cost-per-mile comparison, and a new carrier qualification process. The model does not generalize from Walmart to all shippers automatically. It replicates through a shipper-by-shipper commercial motion that takes time.

The transit and fixed-environment industrial segment is closing deals. EasyMile, Oxa, and Embotech have all converged on the same commercial finding: fixed-environment deployments produce contracts; open-road deployments do not yet. The qualification criteria in fixed environments are different from highway freight , airport safety certification, factory acceptance testing, port operational protocols , and they are achievable without FMCSA carrier history. [C7][C8][C15] (Sourced fact)

Sensible 4 demonstrates what happens when the technology exists but the commercial translation does not. The company’s DAWN autonomous driving platform has been proven at −40 degrees Celsius, accumulated 5,669 kilometers and 560 hours in a Tampere, Finland winter pilot with 1,663 passengers transported, and has planning pilots in Norway, Switzerland, Japan, and Germany. [C16] (Sourced fact) The research entry identifies the constraint: “stuck in perpetual pilot mode across 4+ countries , needs to convert one pilot into a commercial contract to prove the business model beyond R&D demonstrations.” The technology is validated. The commercial contract is not closed. The gap is the transition from demonstration evidence to operational qualification evidence.

The robotaxi segment operates on different economics. May Mobility is deploying driverless service in Peachtree Corners, partnering with Uber in Arlington TX, and partnering with Lyft in Atlanta. The research entry identifies the margin constraint: “municipal transit and campus deployments are low-margin, subsidy-dependent markets , Uber/Lyft partnerships offer commercial scale but pricing pressure from ride-hail platforms may squeeze margins further.” [C13] (Sourced fact)

Vay operates a teledriving service in Las Vegas at approximately 2,000 trips per month across 30 vehicles. The pricing is $0.35 per minute. The research entry identifies the structural limit: “teledriving requires remote operators (labor cost), so margins may not improve with scale the way autonomous vehicles do , unclear path to profitability without reducing operator-to-vehicle ratio.” [C14] (Sourced fact) The service works. The economics at scale are not yet demonstrated.

Economics and Competitive Implications

The cost-per-mile translation gap is the primary commercial constraint for freight-facing AV companies. The freight buyer’s evaluation is simple: what is the fully loaded cost per mile under the AV model, and how does that compare to my current operation? The AV vendor’s typical presentation does not answer this question. It presents technology capability and projected economics under assumptions that the buyer cannot verify independently.

The freight buyer’s evaluation metric. A freight shipper operating a private fleet calculates total cost per mile including driver wages, fuel, maintenance, insurance, and equipment amortization. A shipper using contract carriers evaluates the carrier’s cost-per-mile rate against spot and contract market rates. Either way, the buyer has a current cost baseline. The AV vendor must demonstrate that its cost per mile , at commercial volume, on the buyer’s specific lanes, under the buyer’s carrier qualification requirements , beats that baseline. (Sourced fact , primary research finding on buyer metric; specific cost structures are reasoned inference)

The Gatik benchmark. Gatik has contracted $600 million in total revenue, with $400 million added in the second half of 2025, scaling from 10 to 60 trucks. [C4] (Sourced fact) The implied per-truck contracted value at 10 trucks is $60 million , which, distributed across the contract life, represents the per-mile economics the Walmart and Kroger procurement teams approved. The specific per-mile rate is not publicly disclosed. The fact that Fortune 50 shippers signed at that value provides the sector’s first validated commercial data point. (Reasoned inference from stated data)

The oilfield advantage. Kodiak’s pivot to Permian Basin oilfield hauling is commercially rational precisely because the incumbent cost structure in that environment is higher than highway freight. Hauling frac sand in the Permian Basin currently requires human drivers in a physically demanding environment with high labor turnover. The cost-per-mile comparison for autonomous operation in the Permian Basin is more favorable for AV than the cost-per-mile comparison for highway long-haul, where established carriers have efficient operations. [C6] (Sourced fact on operations; cost-per-mile comparison is reasoned inference)

Kodiak reports 5,200-plus paid driverless hours in the oilfield. “Paid” is the operative qualifier , the buyer is paying commercial rates. Commercial history is accumulating. That history will be the carrier qualification evidence for subsequent freight buyer evaluations. (Sourced fact; inference on qualification value is reasoned)

The fixed-environment commercial advantage. EasyMile, Oxa, and Embotech all made commercial progress by targeting environments where the cost-per-mile comparison is specific and the qualification requirements are defined by the facility operator, not the FMCSA. A BMW Group factory manager evaluating autonomous vehicle marshalling can calculate the labor cost of manual car movement, compare it against the EasyMile service contract, and make a closed economic evaluation without any carrier safety history requirement. [C7][C15] (Sourced fact; inference on procurement logic is reasoned)

The competitive implication is direct: AV companies that build commercial operational history , in any fixed or defined environment where the buyer can evaluate a real cost-per-mile comparison , are accumulating the qualification evidence that later enables the harder freight buyer conversations. Companies that accumulate test miles without commercial operations are not building the same evidence base. (Reasoned inference from named evidence)

What Decision-Makers Should Conclude

The diagnostic for AV commercial teams begins with translating the technology roadmap into the freight buyer’s commercial evaluation framework. The sequence is:

Step 1: Identify your target buyer segment precisely. B2B fixed-route middle-mile, long-haul highway freight, robotaxi, and transit/industrial are different commercial markets with different qualification criteria. Einride is selling an autonomous-electric bundle to freight shippers, which requires buyers to evaluate two unfamiliar cost structures simultaneously. May Mobility is selling autonomous transit to municipalities that fund operations through subsidy programs. Gatik is selling a cost-per-mile freight service to enterprise shippers on defined lanes. These are not variations of the same sales motion , they are different commercial products requiring different buyer qualification evidence.

Step 2: Map the qualification criteria your target buyer uses. For B2B freight shippers: cost-per-mile on specific lanes versus current carrier rates, carrier safety record in the FMCSA system, equipment compliance, and carrier insurance. For industrial/airport/factory operators: safety certification for the specific facility, operational reliability metrics in a controlled environment, and service contract economics. These are the evidence requirements the buyer applies before signing. Building the commercial presentation around technology specs instead of these criteria produces sales cycle stall.

Step 3: Build the qualification evidence before pursuing the contract. For freight shippers: the carrier safety history required by FMCSA and by shipper approved-carrier programs cannot be built in a demonstration or test environment. It requires commercial operations generating real freight, on real commercial routes, reported through the carrier compliance infrastructure. For industrial operators: the evidence is a reference deployment at an equivalent facility, demonstrating operational reliability under production conditions, with a verified cost reduction. The reference deployment is the qualification document.

Step 4: Sequence the fixed-environment deployment as the credentialing path , not as a retreat from the core market. Kodiak’s Permian Basin oilfield deployment, Oxa’s industrial pivot, and EasyMile’s industrial/airport pivot are all commercially rational as credentialing strategies. Each builds operational history and cost-per-mile data in an environment where the qualification barrier is achievable. That history is then portable to more competitive freight buyer conversations. Companies that skip the fixed-environment phase in order to pursue open-road freight directly are attempting the hard qualification conversation without the evidence it requires.

Step 5: Translate cost-per-mile, not technology specs, in the commercial presentation. The freight buyer’s question is: what does this cost per mile on my lanes, and how does it compare to my current operation? The AV vendor’s technology brief does not answer this question. The commercial team must provide a lane-specific cost-per-mile comparison, built on actual operational data from comparable deployments, that the buyer can verify against their own cost baseline. Gatik closes contracts because its buyers can make this comparison on their specific routes. Companies that cannot provide this translation are presenting to buyers who have not yet reached the evaluation stage the contract requires.

For investors evaluating the on-road AV sector:

The commercial signal that matters is not driverless miles, ODD expansion, or autonomous reliability percentage. It is contracted commercial revenue at cost-per-mile economics that replicate or improve on the incumbent. Gatik’s $600 million contracted is the current sector benchmark. Every other company’s commercial path should be evaluated against whether its evidence base , carrier qualification history, cost-per-mile data from commercial operations, route-specific operational record , supports the contracts it is claiming to pursue.

Remaining Unknowns

Gatik replication beyond Walmart and Kroger. Gatik’s contracted $600 million is concentrated in its initial Fortune 50 shipper relationships. Whether the B2B fixed-route middle-mile model extends to additional shippers at comparable contract value , and whether the fleet can scale from 60 trucks to the size required to generate the revenue base that justifies the company’s growth trajectory , is not determinable from public data. (Open question)

Waabi driverless commercial timeline. Waabi achieved surface-street autonomy in 2025. Driverless commercial operations depend on Volvo VNL Autonomous platform validation. Whether that validation completes in 2026 and produces commercial freight operations, or whether the delay extends into 2027, is not publicly determinable. The $750 million-plus raised in January 2026 extends runway. It does not change the Volvo validation dependency. [C5] (Open question)

Kodiak long-haul timeline at 78% ARM. Kodiak reports autonomous reliability metrics at 78% for long-haul highway operations, with a driverless long-haul launch planned for H2 2026. [C6] Whether the remaining performance gap represents 6 months or 24 months of development is not publicly estimable. The oilfield commercial operations provide near-term revenue while this question resolves. (Open question)

Stack AV commercial viability. Stack AV has operated for 2.5 years with SoftBank backing and no public commercial traction. SoftBank’s investment thesis, the runway duration, and whether commercial announcements precede the next capital event are all unknown. [C2] (Open question)

Sensible 4 pilot-to-contract conversion. Sensible 4 has completed a validated winter-conditions pilot in Tampere with demonstrated all-weather capability at −40 degrees Celsius, and has pilots planned in Norway, Switzerland, Japan, and Germany. Whether any of these pilots converts to a commercial operating contract is the company’s defining commercial question. [C16] (Open question)

Aurora and Torc Robotics current commercial status. Aurora and Torc Robotics are referenced in the sector’s broader context but are not in the primary research dataset for this analysis. Their current commercial status , contracted revenue, carrier qualification history, cost-per-mile economics , is not covered in the evidence base used in this article and is not addressed. (Open question , evidence gap)

Frequently Asked Questions

Why aren’t driverless miles converting to freight contracts? Freight buyers do not evaluate new carrier relationships based on accumulated driverless miles. They evaluate based on cost-per-mile on specific lanes versus current carrier rates, safety history recorded in the FMCSA carrier compliance system, and operational record on commercial freight routes. Driverless miles accumulated in test or demonstration operations do not build FMCSA carrier history and do not demonstrate route-specific cost-per-mile economics. The metric AV vendors present and the metric freight buyers use to qualify new carriers are structurally different. That gap is why sales cycles stall after technology briefings.

What makes Gatik’s commercial model work when others haven’t? Gatik operates fully driverless middle-mile trucks on fixed routes for B2B enterprise shippers in Texas, Arkansas, and Arizona. The fixed-route, short-haul, geofenced structure allowed Walmart and Kroger to evaluate a specific cost-per-mile comparison on defined lanes and observe commercial operational history , 60,000-plus driverless orders , before committing to fleet expansion. The model removed open-ended qualification variables: route variability, unexpected operational conditions, and unverifiable safety projections. Freight buyers approved what they could evaluate specifically, not what was projected generally.

What is the freight carrier qualification barrier for AV companies? Freight carriers in the US are qualified through the FMCSA Safety Measurement System and Carrier Safety Fitness Determination process. Major shippers maintain approved-carrier lists based on carriers’ safety records, compliance history, and performance data in the FMCSA system. A new carrier , AV or otherwise , has no history in that system and must accumulate commercial operational record before qualifying for preferred carrier programs. A technology demonstration does not build this record. Commercial operations generating real freight on commercial routes, reported through the carrier compliance infrastructure, build this record. The barrier requires time on commercial routes, not time on test courses.

Why are fixed-environment deployments (factories, airports, ports) closing faster than highway freight? Fixed-environment buyers , factory managers, airport operators, port authorities , qualify autonomous vehicles against the specific operational requirements of their facility, not against the FMCSA carrier system. The cost-per-mile comparison in a factory is labor cost of manual vehicle movement versus autonomous service contract economics. A facility manager can make that comparison with a reference deployment at an equivalent facility. The qualification evidence is a reference site with verified operational reliability and cost reduction data. This is achievable without highway commercial operations history. Highway freight buyers require both route-specific cost-per-mile data and carrier safety history in the FMCSA system , a higher evidence threshold that takes longer to build.

How should AV investors evaluate commercial progress? The commercial signal that matters is contracted revenue at verified cost-per-mile economics , not driverless miles, ODD expansion, or autonomous reliability metrics. Gatik’s $600 million contracted across a Fortune 50 shipper base is the sector benchmark. The questions for any other company are: does the company have commercial operations generating freight revenue (not test or demonstration operations); does the company have an FMCSA carrier safety record (for freight-facing companies); and can the company provide a route-specific cost-per-mile comparison that a freight buyer can verify against their current carrier costs? Companies that cannot answer all three questions affirmatively have not yet built the commercial qualification evidence their target buyers require.

Evidence Base

Sources used in this article:

1. Problem_Proof_Matrix , On-Road Filter , 23 companies: SCS(8), CC(5), PTS(4), PM(4), QF(2). Autonomy Bridge primary research, 2024-2026. [C1]
2. On-Road AV Company Research (32 companies) , Individual company entries for Gatik, Waabi, Kodiak, EasyMile, Oxa, Einride, Sensible 4, May Mobility, Nuro, Vay, Embotech, Stack AV, Udelv, Conigital, TIER IV. Autonomy Bridge primary research, 2026. [C2]
3. Sector Research , On-Road AV Section , “AV graveyard massive (Argo, TuSimple, Embark dead). Survivors bifurcated: Gatik winning, robotaxi marginal, long-haul pre-revenue.” Three documented sector problems. Autonomy Bridge primary research, 2026. [C3]
4. Gatik contracted revenue and fleet data , $600M contracted; $400M H2 2025; 10 trucks scaling to 60; 60,000+ driverless orders; Walmart/Kroger/Fortune 50. Public disclosure, 2025-2026. [C4]
5. Waabi raise and Volvo partnership , $750M+ raise January 2026; surface-street autonomy 2025; driverless delayed pending Volvo validation. Public disclosure, January 2026. [C5]
6. Kodiak Permian Basin and ARM data , 10 driverless trucks; Atlas Energy Solutions; 5,200+ paid driverless hours; ARM 78% long-haul; H2 2026 driverless launch planned. Public disclosure, 2025-2026. [C6]
7. EasyMile pivot and deployments , 30+ commercial deployments; BMW/Daimler/John Deere/Munich Airport/Narita Airport; 15+ sites planned 2026; Level 4 Japan. Public disclosure, 2025-2026. [C7]
8. Oxa Series D and pivot , $103M Series D March 2026; industrial pivot; Port of Tyne; 14-vehicle Jacksonville. Public disclosure, March 2026. [C8]
9. Einride funding and IPO , $113M February 2026; SPAC IPO $1.8B; autonomous fraction small; Mars 300-truck deal. Public disclosure, February 2026. [C9]
10. Udelv reservation status , 1,000+ reservations; 35,000-unit plan 2021-2022; no visible deployments 2025-2026. Public disclosure history; absence of 2025-2026 evidence. [C10]
11. Conigital liquidation , Voluntary liquidation February 2025; fraudulent funding claim. Public record. [C11]
12. Nuro pivot and funding , Business model pivot to licensing September 2024; $203M Series E at $6B valuation August 2025; Uber/Lucid 20,000-vehicle program. Public disclosure. [C12]
13. May Mobility partnerships , Uber/Arlington; Lyft/Atlanta; subsidy-dependent municipal model. Public disclosure, 2025. [C13]
14. Vay service data , 2,000 trips/month; $0.35/minute; 30 vehicles scaling to 100; remote operator model. Public disclosure, 2025. [C14]
15. Embotech BMW and Port of Rotterdam , 6 BMW plants; Port of Rotterdam 30 units; $27M Series B; TUV SUD CE marking. Public disclosure, 2025. [C15]
16. Sensible 4 Tampere pilot , −40C capability; 5,669 km; 560 hours; 1,663 passengers; multi-country pilots planned. Public disclosure, 2025. [C16]
17. Vendor Economics Framework , Autonomy Bridge. [C17]

Highest-confidence conclusions (sourced fact):

• Argo AI, TuSimple, Embark dead; Conigital liquidated February 2025
• Gatik has $600M contracted revenue; 10 trucks scaling to 60; Walmart/Kroger/Fortune 50
• Waabi raised $750M+; no driverless commercial operations; Volvo validation pending
• Kodiak: 10 driverless oilfield trucks; 78% ARM for long-haul; oilfield as near-term commercial pivot
• EasyMile, Oxa, Embotech: fixed-environment deployments producing commercial contracts
• Stack AV: 2.5 years since launch, no announced customers or business model
• Udelv: 1,000+ reservations from 2021-2022; no visible commercial deployments 2025-2026
• Sector research: freight buyers qualify on safety history; AV companies have none

Moderate-confidence conclusions (reasoned inference):

• Fixed-route/fixed-environment commercial model closes faster because qualification evidence is achievable without FMCSA carrier history
• Kodiak oilfield pivot accumulates commercial operational history that is portable to subsequent freight buyer conversations
• Cost-per-mile translation gap is the primary sales cycle stall driver , inferred from sector research finding and named company patterns

Known evidence gaps:

• Gatik per-mile economics not publicly disclosed
• Waabi driverless commercial timeline , Volvo validation dependency
• Kodiak long-haul timeline from 78% ARM to commercial readiness
• Stack AV SoftBank runway and commercial activation timeline
• Aurora and Torc Robotics , not in primary research dataset; commercial status not covered