By Dongyan Zheng in AI — 25 Jun 2026

ChinaBiz Briefing | Unitree's $4,150 Robot, Beijing's AI Trillion, MiniMax Meltdown

China's technology and capital markets are moving in two speeds simultaneously: the supply side is breaking cost barriers and benchmark records at a pace that outstrips Western expectations, while the demand side — and the companies caught between the two — is forcing a painful repricing of ambition. From Unitree's humanoid robot hitting consumer-electronics price points to Beijing mobilizing multi-trillion-yuan state capital for AI infrastructure, Wednesday's news flow reveals a system accelerating on multiple fronts. The stress fractures, however, are equally visible — in MiniMax's collapsing stock and Li Auto's cratering margins.

Unitree Puts a Humanoid Robot in the Shopping Cart for $4,150

Unitree has cut the R1 Air price to RMB 29,900 (US$4,150)—a $2,000 per unit reduction.

Why it matters: Unitree has effectively replicated the smartphone commoditization playbook in embodied AI hardware, years ahead of schedule. At this price point, the addressable buyer expands from elite research labs to universities and hardware startups. With rival Songyan Power already listing its Bumi humanoid below RMB 10,000, the industry's cost floor has structurally reset. The read-through for investors is bifurcated: vertically integrated players benefit; assemblers reliant on third-party actuators face a margin squeeze that is already visible in today's pricing data.

Beijing Maps a Multi-Trillion-Yuan AI Financing Machine

A Nomura research note published June 24 lays out the full architecture of China's state-led AI financing apparatus. The "Big Fund" semiconductor vehicle has deployed nearly RMB 700 billion (US$97 billion) across three phases, with Phase III co-establishing a RMB 60 billion National AI Industry Investment Fund and reportedly in talks to lead DeepSeek's initial financing round. Ultra-long central government bonds — RMB 800 billion allocated in 2026 alone — are being redirected toward AI, and a RMB 2 trillion data center buildout plan requiring 80%-plus domestic technology content is in active drafting. Policy bank lending tools are expected to catalyze over RMB 7 trillion in total project investment from an initial RMB 500 billion quota.

Why it matters: The scale and coordination distinguish this from prior industrial policy cycles. State capital is not substituting for private spending — it is running alongside it. ByteDance, Alibaba, and Tencent collectively account for roughly RMB 500 billion in annual AI capex, and all three are accelerating. For global investors, the implication is that China's AI infrastructure buildout is structurally insulated from the funding volatility that periodically disrupts Western venture markets. The constraint is no longer capital — it is execution and chip access.

Kling AI Repriced to $15B as Tencent and Alibaba Circle

Kuaishou’s AI video spinout Kling has cut its fundraising ask to a $15 billion pre-money valuation—down from $20 billion in early 2026—with Tencent, Alibaba, and Sequoia Capital China emerging as lead investors.

The fundraising deck projects ARR reaching $1 billion by December 2026, up from $500 million in March. Notably, 75% of Kling’s revenue originates outside China, and API calls account for 60% of total revenue—a B2B-heavy mix that supports the premium multiple.

Why it matters: The valuation compression is price discovery, not distress. At 30x current ARR, the multiple still demands hypergrowth — and ByteDance's Seedance 2.5, launching in July with 30-second clips and 50 simultaneous multimodal inputs, is closing the technical gap fast. Tencent and Alibaba's participation is strategically ambiguous: both run competing AI video programs, yet both need a hedge against video generation becoming foundational infrastructure. The $15 billion entry point will be tested by how convincingly Kling converts ARR momentum into defensible margins before its 2027 Hong Kong IPO.

Li Auto's Margins Collapse as the Embodied AI Bet Burns Cash

Li Auto reported Q1 2026 revenue of RMB 23 billion, down 11.4% year-on-year, and swung to a net loss of RMB 2.3 billion from a RMB 647 million profit a year earlier. Vehicle gross margin fell from 19.8% to 6.1%. The sole volume driver is the lower-priced i6 BEV; the flagship L8 and MEGA recorded fewer than 500 monthly deliveries each. Meanwhile, CEO Li Xiang is restructuring the entire R&D organization around embodied intelligence, with automobiles and humanoid robots classified under the same hardware body division.

Why it matters: Li Auto faces a classic innovator's dilemma with compressed timing. The extended-range vehicle segment it pioneered is now commoditized, with more than a dozen rivals spanning RMB 100,000 to RMB 500,000. The new L8 Livis and L9 Livis represent a genuine technology step forward — the L9 generated 10,000 firm orders in its first two weeks — but consumer trust in Li Auto's pricing stability has been eroded by aggressive discounting on prior flagships. The company needs car sales to fund its AI platform ambitions; through Q1 2026, it has not yet demonstrated it can deliver both simultaneously.

MiniMax Faces a Lock-Up Cliff After a Self-Inflicted Pricing Crisis

MiniMax’s Hong Kong-listed shares have shed HK$267.6 billion in market cap since their March peak, leaving the company valued at HK$149.8 billion—roughly one-sixth of rival Zhipu AI. The trigger was a June 1 unannounced API price hike that cut developer usage capacity to less than one-third of previous levels, prompting a 15.7% single-day stock decline. The structural problem runs deeper: consumer AI products generate 67% of revenue but carry only a 4.7% gross margin, versus 69.4% for the B2B API segment. In early July, lock-up expiry could increase available share supply nearly tenfold.

Why it matters: MiniMax's crisis is a case study in the pricing power deficit facing China's second-tier AI model companies. The company pivoted late to AI coding — now the primary enterprise monetization battleground — and is perceived as a follower behind Zhipu and Moonshot. The next 60 days are decisive: M3 coding benchmark performance will determine both stock price defense and eligibility for a planned STAR Market dual listing. If enterprise revenue can demonstrably match consumer revenue on a run-rate basis, as management claims, the margin profile improves materially — but that claim still requires verification in reported financials.

Sugon Tops IO500 Double Crown, Rewriting China's Storage Narrative

Sugon's ParaStor F9000 all-flash storage system claimed first place on both the full-node and 10-node IO500 production-tier rankings at ISC 2026, displacing long-time incumbents DDN and Intel. The IO500 production tier requires continuous real-world deployment — Sugon's system had been running in clusters exceeding tens of thousands of GPU cards for over a year before submission, supporting more than 100 production AI and HPC applications. The company claims a 50% improvement in cluster training efficiency and an 80% reduction in inference latency versus prior configurations.

Why it matters: IO500 and TOP500 collectively define the global benchmark hierarchy for supercomputing. Long-term Western dominance of the production tier meant that de facto performance standards for AI and scientific computing infrastructure were set outside China. Sugon's double crown introduces a Chinese reference point into that standard-setting process for the first time — and does so through production-first validation rather than benchmark-optimized hardware. In an environment where GPU supply remains constrained and expensive, the ability to extract materially more output from existing hardware is not an engineering footnote; it is a capital efficiency argument that procurement teams at Chinese hyperscalers are running in real time.

Nomura: ESS and Robot Batteries Redefine the Battery Investment Thesis

Nomura’s 50-page battery sector report, published June 24, projects global EV battery demand reaching 1.8 TWh by 2030, with China holding 78% of EV battery market share and 80% of ESS share. The more actionable call is energy storage: Nomura has raised its ESS demand forecast by 25%, now projecting 926 GWh by 2030 at 17% annual growth, driven by AI data center backup power demand and grid modernization. Robot batteries—commanding cell prices of US$200–400/kWh versus roughly US$100/kWh for EVs—are projected to reach a US$2–4 billion market by 2030. CATL is Nomura’s top global pick, with 49.6% upside implied to its CNY 612 target price.

Why it matters: The battery sector is fracturing into distinct competitive arenas with different winners. China dominates LFP-based volume; South Korea is carving out a tariff-protected ESS niche in the US market. The robot battery segment is small in volume but structurally attractive on economics — high C-rate requirements favor high-nickel NCM chemistry in the near term, but improving LFP performance could allow Chinese manufacturers to gain share over time. The replacement cycle dynamic — robot batteries last two to five years versus eight to ten for EVs — adds a recurring revenue dimension that EV-centric battery models have not yet priced in.

J.P. Morgan's European Automotive Conference summary, published June 24, projects Chinese OEMs — including BYD, Geely, and Chery — will capture approximately 20% of European auto market share, up from roughly 12% currently. China's vehicle exports are expected to approach 10 million units in 2026, a near-40% increase year-on-year. Spain is emerging as a key production localization beachhead. On credit, S&P carries a negative outlook on 50% of covered European OEMs versus just 30% for parts suppliers — a reversal of prior positioning driven by supplier agility in restructuring and contract rationalization.

Why it matters: The "in China, for global" strategic shift is no longer a forecast — it is a supply chain reality. Chinese battery suppliers are traveling with their OEM customers into overseas markets, compounding demand from multiple directions simultaneously. For European legacy automakers, the competitive threat has moved from theoretical to measurable within a single conference cycle. The credit market's verdict — suppliers outrunning OEMs — suggests the adjustment burden is falling disproportionately on the brands rather than the component ecosystem.

What to Watch Next

The convergence of state capital mobilization, cost-floor resets in hardware, and valuation stress-testing in software creates a layered risk-reward picture for the second half of 2026. Key inflection points: Kling's ARR trajectory into Q3 as Seedance 2.5 launches in July; MiniMax's lock-up expiry and M3 coding benchmark data; Li Auto's L8 order volumes as a test of whether Livis-generation technology can rebuild consumer trust; and the first batch of STAR Market listings under expanded hard-technology eligibility criteria. On the infrastructure side, watch for the formal announcement of Beijing's RMB 2 trillion data center plan and the timeline for CATL's sodium-ion commercial deployment in Q4.

Related Coverage:

Unitree Breaks the Humanoid Robot Cost Barrier With RMB 29,900 R1 Kling AI Valuation Falls to $15B as Tencent, Alibaba Bet on Video AI Infrastructure MiniMax Faces Lock-Up Cliff as Pricing Misstep Exposes Structural Fault Lines in China AI Race Li Auto's Revenue Craters as Li Xiang Doubles Down on Embodied AI Pivot J.P. Morgan’s Europe Auto Warning: China’s OEMs are coming for 20% Market Share China's AI Funding Machine: How Beijing Is Mobilizing Trillions to Win the Tech Race Nomura: ESS and Robot Batteries to Drive the Next Battery Boom as China Dominates Sugon All-Flash Storage Tops IO500, Highlighting China’s Shift to Industry Standard-Setter