WuBlockchain’s WhiteLine Daily said the market’s main thread for the day comes down to one question: what does a high valuation need in order to be justified? Its latest note linked that question to Changxin Technology’s listing terms and to the commercialization progress of embodied AI companies shown at WAIC 2026.
Changxin’s listing valuation leaves investors debating which earnings year matters
Changxin Technology is priced at 8.66 yuan per share, implying a listing valuation of about 579.2 billion yuan. Base fundraising is about 57.9 billion yuan. Online subscriptions reached 243.93 times, the final allotment rate was 0.4714%, and the company is expected to officially list on July 27.
On the surface, WhiteLine Daily described the deal as a concentrated bid for a scarce domestic DRAM asset. The actual tension, though, lies in which year of earnings the market should use to value Changxin.
Using 2025 profit, the implied offering P/E is above 300x and price-to-book is about 5x. Yet the company’s first-quarter 2026 net profit already reached 33 billion yuan, and first-half net profit is projected at 66 billion yuan to 75 billion yuan.
That creates two very different valuation readings. On historical profit, the stock looks expensive. On current cycle earnings, it looks much cheaper. WhiteLine Daily said the market is effectively making a judgment on whether the high profit seen in 2026 will become a new earnings base or remain only the top of the current memory cycle.
Changxin’s global DRAM market share was about 7.7% in 2025, according to the report, making it the fourth scaled producer after Samsung Electronics, SK Hynix, and Micron. At the same time, faster capacity expansion would also carry greater implications for the future structure of global DRAM supply.
WhiteLine Daily said the most important thing to watch on the first day of trading is not how much the stock rises, but how it rises.
- If only Changxin moves higher while the broader STAR Market chip segment continues to lose funds, that would suggest the trade is centered mainly on scarcity, with Changxin pulling capital away from other chip names.
- If DRAM, semiconductor equipment, and materials companies rise together, that would point to a broader repricing of the domestic memory supply chain.
At WAIC 2026, robots were judged less by how human they looked and more by whether they could work
In the industry section, WhiteLine Daily shifted to WAIC 2026. The event placed embodied AI alongside intelligent computing as its two core tracks. More than 1,100 companies attended, with more than 3,000 exhibits and more than 300 global debut products. More than 200 of the participating companies were in embodied intelligence.
The report said industry expansion is also starting to move from funding rounds and prototypes into real procurement. Data from China’s State Taxation Administration showed that in the first five months of 2026, sales revenue at domestic embodied intelligence companies rose 22.4% year over year, while industrial enterprises’ purchases of embodied intelligent robots increased 2.3 times from a year earlier.
WhiteLine Daily argued that the real change at this year’s WAIC was a change in the standard used to evaluate robots. The focus used to be on degrees of freedom, parkour, and backflips. Now the question is whether a robot can complete a full set of tasks. The next step is whether it can run stably over long periods and be replicated at scale.
By that logic, embodied intelligence is moving from capability demonstrations to productivity validation: first prove it can be manufactured, then prove it can do useful work, and then prove it is more cost-effective than human labor.
AgiBot: first prove robots can be produced at scale
The first case in the report was AgiBot. Ahead of the opening of WAIC, AgiBot’s 15,000th general-purpose embodied robot rolled off the line, less than three months after the 10,000th unit. The company said its flexible manufacturing and delivery capacity now exceeds 100,000 units per year.
WhiteLine Daily said mass production matters for more than reducing hardware costs. Only after larger numbers of robots enter real-world settings can the sector build a loop of deployment, data return, model improvement, and wider deployment.
AgiBot’s previously disclosed AgiBot World data platform includes more than 1 million trajectories, 217 tasks, and five application scenarios. Models trained on that data have achieved a success rate above 60% on complex long-horizon tasks.
For the report, this case addresses whether the robotics industry can move from trials measured in the hundreds of units to production measured in the tens of thousands.
MagicLab: moving from general-purpose demonstrations to specific jobs
The second company highlighted was MagicLab. WhiteLine Daily said its focus is not to build one all-purpose humanoid that handles every task, but to match product form factors to different job types.
- MagicBot X1: a full-size humanoid robot designed to explore complex general-purpose tasks;
- MagicBot D1: a wheeled humanoid robot used for high-frequency procedures such as handling and loading or unloading materials;
- MagicDog T1: a quadruped robot for inspection and work in narrow spaces.
At the event, Magic-VLA K02 demonstrated long-horizon tasks including box stacking, gluing, and clothing organization. WhiteLine Daily said that after box movement, changes in tape state, or task interruption, the system could re-recognize the environment and resume execution rather than continue replaying preset motions.
The company said embodied models reduced the demonstration training data needed for a new job by about 60%. The report treated that as more important than completing a single action, because it directly affects adaptation time and deployment cost for new roles.
That case addresses whether a robot can move beyond being a hardware body and become a labor unit capable of delivering a specific result.
WhiteLine Daily’s takeaway: high valuations still have to come back to profit and delivery
WhiteLine Daily said WAIC did not prove that humanoid robots have already reached full commercialization, but it did show that the industry’s rules of competition are changing. AgiBot is beginning to answer whether robots can be manufactured at scale. MagicLab is beginning to answer whether robots can complete specific jobs. The factors that will decide real orders next are utilization rate, manual takeover rate, per-position cost, and payback period.
The report closed by linking both topics back to the same market test. Changxin needs to show that its 2026 profit can hold through the memory cycle. Robotics companies need to show that production numbers and live demos can turn into long-duration operation, real orders, and repeatable job deployment.
The market may be willing to pay a premium for scarcity, but the final test still comes back to whether profits can last, whether products can be delivered reliably, and whether expansion can generate real returns.