BYD Quietly Rips Out Momenta's Brain, Betting Its 10M-Unit Base on Unproven In-House ADAS
BYD has begun replacing Momenta-supplied autonomous driving algorithms across its entire lineup with proprietary software—a cost-driven gamble that benchmark data shows has already cut real-world urban navigation performance by as much as 41%, while threatening to strip Momenta of roughly 62% of its urban NOA installation base.
The switch, which began rolling out silently in late September 2026, was confirmed to First Electric by BYD insiders on Oct. 19 and was corroborated by independent road testing on Nov. 1. BYD issued no public changelog, no OTA notification, and no product announcement—a conspicuous silence from a company that in February 2026 declared its "DiPilot Tianshen Eye" (天神之眼) system "full-stack proprietary." Owner communities across WeChat and Autohome flagged the change first, describing the overnight shift as a "stealth modification" after vehicle settings reset to factory defaults and driving behavior changed materially.
Market implications crystallized quickly: BYD stands to recapture at least RMB 2,000 (US$278) per vehicle in bill-of-materials savings; Momenta faces a potential revenue cliff given BYD's RMB 522 million (US$72.5 million) contribution in fiscal 2025, equivalent to 21.6% of Momenta's total revenue.
Benchmark Data Exposes a 41-Point Performance Gap
The numbers are stark. First Electric's evaluation team drove a BYD Tang EV over Beijing's standardized "Smart Driving Leaderboard" test routes on Nov. 1, 2026, with the vehicle displaying firmware version V1.1.0—the same version string it carried when equipped with Momenta's R6 model.
On Route A (low-to-medium difficulty), the Tang scored 78.14 points with 8 interventions, including one safety-critical takeover. The identical vehicle platform running Momenta R6 had scored between 101 and 118 points across five tests from May through September 2026, most recently 101.19 on Sept. 18—a 23-point regression in under six weeks.
Route B (medium-to-high difficulty) was more revealing. Over 70-plus minutes, the Tang stopped dead in front of speed bumps, twice halted mid-bridge at Ciyunsi requiring manual override, missed navigation exits on the East Fourth Ring Road auxiliary lane, and failed to avoid a pothole on Huihe South Road. Fourteen interventions. Final score: 56.3 points—against Momenta R6's 96.6 on the same route Sept. 18, a 40.3-point collapse.
Critically, the failure taxonomy shifted. Momenta R6's deductions had concentrated in edge-case scenarios: extremely narrow intersections and U-turns. The Nov. 1 Tang's deductions concentrated in everyday commuting situations: speed bumps, standard intersections, and highway on-ramp/off-ramp navigation. Eight of 14 interventions occurred because the vehicle hesitated or stopped when it should have moved; four were missed exits. First Electric's evaluation team, having benchmarked multiple Momenta software generations, concluded unambiguously: "This is not Momenta's algorithm."
Qualitative owner reports corroborate the quantitative verdict. Drivers describe following distances doubling without user adjustment capability, acceleration lagging well below posted speed limits, and—paradoxically—increased lateral aggression, with the system initiating lane changes when a lead vehicle travels even 1 km/h below the speed limit. One owner noted the system attempted to merge into an occupied lane, was manually blocked, then re-initiated the maneuver seconds later.
Cost Arithmetic Forces BYD's Hand
The commercial logic behind the switch is traceable to a single price point: RMB 12,000 (US$1,667)—the current opt-in price for DiPilot Tianshen Eye B-tier (the Momenta-powered urban NOA system) on models as affordable as the Seagull (海鸥).
That figure is structurally incompatible with the RMB 100,000–200,000 (US$13,900–US$27,800) segment BYD dominates. Per First Electric's China Urban NOA Installation Ranking data, urban NOA penetration in the RMB 100,000–200,000 passenger vehicle band surged from 4.7% in January 2026 to 17.4% by September—crossing the 5%-to-15% "liftoff acceleration band" in just seven months.
Competitors moved aggressively into BYD's core price territory: Xpeng MONA L03 ramped rapidly; Geely Galaxy TT and MG07 each exceeded 5,000 monthly units in their launch months; Leapmotor pushed city navigation capability down to RMB 80,000 vehicles. The competitive message reaching BYD's customer base became blunt: no city NOA at RMB 100,000, no sale.
According to a BOC International research note cited by Aastocks Financial News in October 2026, BYD demanded a substantial reduction in Momenta's per-vehicle licensing fee and was refused. First Electric calculates Momenta's average per-unit royalty at approximately RMB 1,890 (US$263), derived from Momenta's H1 2026 licensing revenue of RMB 607 million (US$84.3 million) against approximately 321,000 new installations—consistent with the RMB 1,898 average implied by Momenta's 2025 annual report (RMB 968 million licensing revenue, ~510,000 vehicles).
A consumer survey leaked in late October 2026 suggests BYD intends to price its proprietary Tianshen Eye C "quasi-city navigation" at a one-time fee of RMB 1,000 (US$139)—below Momenta's per-unit royalty. Combined with the elimination of one Nvidia Orin-X chip (approximately US$380, or RMB 2,736) through substitution with BYD's proprietary Xuanji A3 4nm chip announced in May 2026, the total BOM reduction reaches approximately RMB 2,000 (US$278) per unit. At BYD's sales volumes—the company delivered over 1.76 million new energy vehicles in September 2026 alone—that arithmetic compounds rapidly into margin and competitive firepower.
Momenta Loses Its Largest Client, Pivots Toward SAIC
The revenue exposure is concentrated and immediate. First Electric's October 2026 urban NOA installation data shows BYD-branded vehicles accounting for 62% of Momenta's total urban NOA installations. The loss of that business, even partially phased, represents an existential recalibration.
BYD's RMB 522 million (US$72.5 million) in fiscal 2025 payments constituted Momenta's single largest customer relationship at 21.6% of total revenue. The pace of the transition—already underway across the full model range, per BYD insiders—suggests Momenta's forward revenue curve faces material downward pressure beginning Q4 2026.
Momenta's strategic response arrived four days after First Electric's Nov. 1 test: on Nov. 5, Momenta and SAIC Motor announced a joint venture in which Momenta holds a 51% controlling stake—a structural deepening of the SAIC relationship that signals Momenta's intent to anchor its next growth phase to China's third-largest automaker by volume. The timing is unlikely to be coincidental.
BYD's Strategic Bet: Closing the AI Loop Before the Robotics Era
BYD's silence is explicable, if not defensible, on two grounds. First, DiPilot Tianshen Eye has been marketed as "full-stack proprietary" since the February 2026 launch event; acknowledging that the city NOA capability was Momenta-sourced would retroactively complicate that narrative. Second, the proprietary algorithm is, by the benchmark data, not yet competitive—and broadcasting a mid-cycle capability downgrade to a user base of tens of millions would amplify churn risk.
The strategic upside, however, extends beyond unit economics. BYD's May 2026 announcement of the Xuanji A3 chip and a "Physical AI large model" incorporating LiDAR data was a declaration of intent: BYD wants a closed loop across operating system, compute, algorithm, and training data. In the context of China's accelerating embodied intelligence investment cycle, that loop—once closed—becomes a defensible moat. Xpeng, Li Auto, NIO, and Huawei-backed brands have already established their own vertically integrated stacks. BYD's window to build a comparable architecture is narrowing.
The risk is symmetric. BYD's installed base—the overwhelming majority concentrated in the RMB 100,000–200,000 band—is precisely the segment where urban NOA competition is most intense in 2026. If the proprietary algorithm fails to converge toward industry-average performance within two to three OTA cycles, the same scale that makes the transition strategically logical becomes a reputational liability. At BYD's volumes, a cohort of dissatisfied urban NOA users is not a niche complaint; it is a trending topic.
What is now clear is that the three largest players in China's RMB 100,000-and-below passenger vehicle market—BYD, Geely, and Changan—are all deploying high-order autonomous driving stacks. For Leapmotor, whose city navigation proposition has until now been uncontested in the sub-RMB 80,000 tier, the competitive clock just started ticking faster.
Related Coverage: