Momenta Stays Below Its Hong Kong IPO Price as Horizon Robotics Regains Valuation Lead

Momenta's Hong Kong Listing: A 9% Discount to the IPO Price

Momenta's Hong Kong debut has turned into a sustained discount. The autonomous-driving software company priced its IPO at HK$295.60 on 8 July 2026, implying a market capitalization of about HK$69.6 billion. The stock touched HK$314.80 on day one, but the initial pop faded and the shares barely held the offer price with help from the greenshoe. Once that stabilization support ran out during the 30-day window, the stock slid as low as HK$243.60.

By 14 August, Momenta was trading at HK$268, down roughly 9.3% from the issue price and worth about HK$63.1 billion. The slide is not explained by weak financials: revenue grew from RMB743 million in 2023 to RMB2.413 billion in 2025, a compound annual growth rate above 80%, while gross margin improved from 17.5% to 71.6% and adjusted net loss narrowed from about RMB1.09 billion to RMB303 million.

The more important explanation is the valuation set against Horizon Robotics. During bookbuilding, Horizon was near a recent low of about HK$52.9 billion; by listing day it had recovered to HK$64.6 billion. Momenta, with 2025 revenue equal to only about 60% of Horizon's, listed at HK$69.6 billion. On a static 2025 price-to-sales basis including the greenshoe, Momenta's IPO implied roughly 28.8 times sales, against about 14.9 times for Horizon.

How the Market Re-rated Momenta Against Horizon Robotics

The core vulnerability was the IPO's pricing anchor. Momenta was valued at about HK$69.6 billion when Horizon Robotics had recovered to about HK$64.6 billion, even though Momenta's 2025 revenue was only about 60% of Horizon's. That gave the market a simple repricing trade: as Horizon recovered, Momenta's software premium had to be defended on revenue and moat quality, not just on narrative.

The 'Global Physical AI First Stock' Pitch Runs Ahead of Revenue

Shortly before listing, Momenta shifted its positioning from passenger-car intelligent driving supplier to 'the first global physical AI stock.' The difficulty is that 100% of reported revenue still comes from passenger-vehicle autonomous driving solutions, with no commercial robotics or industrial AI revenue. The prospectus mentions the R7 world model once and does not use 'physical AI' as a term. The R7 model's ability to predict object motion and judge complex driving decisions is technically forward-looking, but the distance from concept to monetization remains long.

How the Market Rotated From Momenta to Horizon Robotics

By 14 August, Horizon Robotics was worth about HK$79.9 billion and Momenta about HK$63.1 billion, a gap of roughly HK$16.7 billion. Momenta's 2025 price-to-sales multiple had fallen to 22.7 times, still above Horizon's 18.4 times. That residual premium is not a contradiction: Horizon's revenue base is roughly 1.5 times Momenta's, so a lower multiple on much larger revenue produces a higher market capitalization.

Why Momenta Keeps a Valuation Premium Despite the Selloff

Momenta's asset-light software model gives it a 71.6% gross margin, against 64.5% for Horizon's chip business, and a much smaller adjusted net loss of RMB303 million versus RMB2.812 billion. Pure software has near-zero marginal cost; chip production carries wafer, materials and inventory costs. Momenta also has a broad customer base — Mercedes, BMW, Audi, Volkswagen, Toyota and General Motors among 24 OEMs, including nine of the ten largest global automakers — and a 65% share of third-party urban NOA. Software can be reused across those customers cheaply, whereas Horizon's chip exports require repeated country-level vehicle certification and supply-chain adaptation.

The counterpoint is that hardware creates stickier customers. Once Horizon's chips are embedded in a production platform, replacing the computing architecture involves hardware adaptation, safety certification and extensive road testing. Momenta's software switch is shorter at about six months or more, but BYD, Li Auto and Nio are advancing full-stack self-development, which may compress the long-term market for third-party software vendors.

Horizon's Numbers Also Need a Second Look

Horizon's apparent first-half 2026 net profit of RMB3.5 billion to RMB4 billion is largely non-cash: it reflects fair-value gains on the CARIAD convertible and a disposal gain from ending the D-Robotics merger, not operating profit. Excluding those items, its adjusted operating loss widened. Its 2025 R&D spend of RMB5.154 billion rose 63.3% and equalled 137.1% of revenue, so profit realization remains long-dated. That helps explain why Horizon's higher market value has not produced an outsized multiple.

Three Specific Re-Rating Triggers for Momenta

For investors comparing the two listings

• Do not treat Momenta's 22.7 times 2025 price-to-sales ratio as evidence of market-cap leadership. Horizon's 18.4 times multiple applies to a revenue base about 1.5 times larger, which is why Horizon is worth about HK$79.9 billion against Momenta's HK$63.1 billion.
• Strip non-cash items from Horizon's first-half 2026 profit of RMB3.5 billion to RMB4 billion: the CARIAD convertible fair-value gain and the D-Robotics disposal gain do not reflect recurring operations, while its adjusted loss is widening.

For Momenta's management and shareholders

• The 'global physical AI first stock' story currently has no robotics, industrial AI or robotaxi revenue behind it; all reported revenue remains passenger-car autonomous driving. The next update needs to demonstrate a second revenue source, not a new label.
• Three re-rating triggers are already visible: mass production of Momenta's planned in-house chip, first commercial physical AI revenue, or large-scale overseas design wins. None of these is yet visible in reported revenue.

For automakers assessing suppliers

• Momenta's 65% share of third-party urban NOA and coverage of nine of the ten largest global automakers are current strengths. The industrial risk is that BYD, Li Auto and Nio are developing their own full-stack systems, so the key question is whether a software-only vendor can stay indispensable as more OEMs insource.