Momenta is reporting losses while its gross margin surges—given its heavy reliance on major customers and automakers’ in-house R&D efforts, what is its IPO outlook and growth ceiling?

It is transforming from a 'per-head project-based company' into a 'software platform charging per vehicle equipped'—a model akin to Microsoft selling Windows licenses: development costs are incurred once, and nearly all revenue from each additional vehicle shipped flows straight to profit. The ceiling for the former model is constrained by team size; the ceiling for the latter is defined by total industry vehicle deployment volume. Once this structural shift is complete, Momenta’s path forward becomes increasingly effortless. Many fixate on its ongoing losses—but look at its adjusted net loss: ¥1.093 billion in 2023, ¥0.959 billion in 2024, and ¥0.303 billion in 2025—a cumulative reduction of nearly ¥0.8 billion over three years. This loss stems primarily from R&D investment: cumulative R&D spending over the past three years reached ¥4.66 billion, with R&D expenses accounting for 77.5% of revenue in 2025 and R&D personnel comprising over 80% of total headcount. Losses driven by high R&D investment to secure future advantage are fundamentally different from burning cash to chase scale. A 'data flywheel' generated by one million vehicles on the road An industry truism goes: code can be written by hired engineers; models can be replicated from academic papers—but real-world road data collected daily from hundreds of thousands of mass-produced vehicles? That’s irreplicable. According to the latest update, as of July 7, more than one million production vehicles equipped with Momenta’s system are on the road. That number stood at just 680,000 by end-2025—meaning an additional 320,000 vehicles were added in only 188 days. Faster vehicle deployment accelerates data accumulation; richer data enables faster algorithm iteration; stronger algorithms win more customers. Once this virtuous cycle gains momentum, the gap between Momenta and latecomers will widen continuously. Many claim that, as the entire industry races toward world models and end-to-end architectures, Momenta lacks technical moats. That view is only half-right: while algorithmic frameworks are widely accessible, the 'ingredients' required to train them—authentic, real-road driving data—are not equally available to all. Data gathered from dozens of modified Robotaxi test vehicles covers only limited scenarios; one million mass-produced vehicles traversing roads across China and globally encounter every conceivable condition—and countless edge cases—every single day. Models trained on such authentic data possess generalization capabilities orders of magnitude superior. The competitive dimension of intelligent driving has already shifted—from 'Can it drive?' to 'Does it drive well—and safely?' At this point, it’s worth highlighting Momenta’s 'one flywheel, two legs' strategy. Simply put: dual-track execution—L2 mass-market ADAS and L4 autonomous driving. Crucially, Momenta uses a unified underlying tech stack for both tracks: data from mass-produced L2 vehicles directly feeds L4 algorithm iteration, while cutting-edge L4 technologies can be downgraded and integrated into量产 solutions. However, sustaining the flywheel demands continuous investment. Currently, the Robotaxi business contributes no material revenue; R&D expenses are likely to rise further, and overseas expansion also requires capital. The direction is right, the growth trajectory is sound—but the inflection point hasn’t yet been fully realized. Controversies and Challenges No investment candidate is flawless—and Momenta faces objective criticisms as well. First, the customer concentration issue. Early on, Momenta was indeed heavily reliant on SAIC Motor Group. Yet with OEMs including Mercedes-Benz and Nissan ramping up production of their designated models, its customer base is diversifying—a natural phase in corporate evolution.

U.S. earnings report — Marvell (MRVL), fiscal year ending March 6, 2026

Management explicitly guided for >30% revenue growth in FY2027, implying full-year revenue approaching or exceeding $11 billion. Custom silicon revenue is expected to grow >20% in FY2027 and double again in FY2028—suggesting custom silicon could reach a $6–7 billion scale by FY2028. III. Peer Comparison Analysis 3.1 Semiconductor Peer Overview Company Revenue (TTM, $B) YoY Growth Gross Margin Forward P/E EV/Sales Market Cap ($B) Marvell (MRVL) 8.2 +42% 59.0% 24.7x 9.8x 80 Broadcom (AVGO) 56.4 +44% 66.9% 30.9x 17.2x 1,050 AMD 25.7 +14% 52.1% 22.5x 7.5x 190 NVIDIA (NVDA) 130.5 +114% 73.0% 30.5x 25.0x 2,800 Lattice (LSCC) 0.5 -15% 68.5% 45.0x 12.0x 7 3.2 Valuation Comparison Marvell’s forward P/E of 24.7x is the lowest among AI semiconductor peers—below Broadcom (30.9x) and NVIDIA (30.5x). Given MRVL’s >30% FY2027 growth guidance, its PEG ratio stands at ~0.8x—highly attractive within the high-growth semiconductor space. Its EV/Sales multiple of 9.8x is also meaningfully below NVIDIA’s 25.0x and Broadcom’s 17.2x, though above AMD’s 7.5x—reflecting market recognition of MRVL’s differentiated positioning in custom silicon + optical interconnects, while expressing some concern over its declining gross margin trend. 3.3 Growth–Profitability Positioning Within the growth–profitability quadrant, MRVL sits in the 'high growth / mid-tier profitability' segment: its growth rate (+42%) trails only NVIDIA (+114%), but its gross margin (59%) lags behind both NVIDIA (73%) and Broadcom (67%). The key distinction lies in MRVL’s growth drivers—custom silicon (lower-margin but high-barrier) combined with optical interconnects (higher-margin)—versus NVIDIA’s reliance on standardized GPUs and Broadcom’s blend of custom silicon and software (VMware). 3.4 Marvell vs. Broadcom — The Custom Silicon Duopoly Dimension Marvell Broadcom Custom Silicon Customers Amazon, Google, Microsoft (3+) Google, Meta, ByteDance (6) Custom Silicon Revenue Doubled in FY2026; doubles again in FY2028 AI revenue of $44B in FY2027 (+40%) Differentiation Optical interconnects + Ethernet switching + CXL Network ASICs + software (VMware) Gross Margin 59% (dragged down by custom silicon) 67% (lifted by software)

Mativ Holdings Inc (MATV) Earnings Forecast: Future EPS & Revenue Growth Estimates

Jun 18, 2026

Mativ Holdings Inc projects next quarter revenue of $508.50M and EPS of $0.28. Price targets range from $15.50 to $21.00, benchmarked against a Chemicals industry average score of 7.31.

TradingKey

2026-05-07

Mativ Holdings Q1 2026 Earnings Beat EPS Forecast, Stock Rises

2026-05-08

Mativ Holdings Swings Back To Quarterly Loss And Renews Earnings Volatility Debate

2026-05-10

Mativ Announces First Quarter 2026 Results

2026-05-11

Mativ Holdings: A Speculative Buy On Improving Fundamentals (NYSE:MATV)

2026-05-11

Mativ Holdings Reports Q1 2026 Earnings Results