ChinaBiz Briefing | AI Apps Go Paid, EV Makers Build Chips, US Taps China Battery Tech

China's technology and mobility sectors are converging on a single underlying pressure in June 2026: the cost of ambition. AI platforms are burning compute budgets faster than they can build revenue models. EV startups are rewriting their corporate identities around silicon rather than steel. And Detroit's legacy automakers are licensing Chinese battery IP to compete in a market their own engineers cannot yet replicate. Across every sector, the question is no longer who can scale — it is who can make scale pay.

China's 710M-User AI Market Hits a Monetization Wall

China's top four AI native apps — Doubao, Qianwen, DeepSeek, and Yuanbao — collectively reached 710 million monthly active users in May 2026, according to QuestMobile data. But the industry's defining event of the month was not a user milestone: it was ByteDance quietly listing Doubao subscription tiers at RMB 68–500 per month, triggering an estimated 6.1 million churns and a survey finding that 43% of users would quit upon any paywall introduction. ByteDance's 2025 net profit reportedly declined more than 70% year-on-year, with AI compute costs cited as a primary driver.

Why it matters: The churn data establishes a sector-wide benchmark with uncomfortable implications. A free-access culture built over three years cannot be unwound without friction — and every major AI platform faces the same compute cost pressure that forced ByteDance's hand.

The divergent responses are instructive: Alibaba's Qianwen is routing monetization through Taobao transaction infrastructure; Tencent is hedging via a RMB 10 billion stake in DeepSeek and embedding AI into WeChat's 1.4 billion-user Mini Program ecosystem; Kimi has abandoned MAU entirely, crossing US$200 million ARR by April 2026 after a deliberate pivot to enterprise subscriptions.

DeepSeek, meanwhile, closed a RMB 50 billion-plus funding round at a post-money valuation exceeding US$50 billion — with founder Liang Wenfeng retaining effective absolute control through a structure that grants investors no voting rights. The monetization reckoning is not a ByteDance problem. It is a sector-wide event in progress.

Alipay's AI Redesign Confronts a Decade of Institutional Failure

Ant Group has launched an invitation-only beta of an AI-native Alipay, bifurcating the interface into a financial dashboard and an AI assistant named "Ābo" that executes voice and text commands across the app's ecosystem. The redesign, internally codenamed "Ābo Plan," has been in development for over a year. Notably, the AI version defaults to opt-in — users must actively switch from the legacy interface.

Why it matters: The opt-in default is a rare act of institutional self-awareness from a platform that has historically forced interface changes on its 1.04 billion monthly active users. But the deeper challenge is organizational, not technical. Alipay's core payment business operates near breakeven; its profit engine has always been financial products stacked atop payment traffic. That structural dynamic — where nudging a credit limit upward outperforms grinding through merchant onboarding — eroded the company's capacity for "slow work" across four failed strategic pivots since 2014. Ant's 2024 R&D spend reached approximately RMB 23.45 billion, but full-year 2025 profit fell roughly 60% year-on-year. The AI pivot's outcome will be determined not by model performance but by whether the organization can rebuild institutional patience that a decade of financial product superprofits systematically destroyed.

NIO, Xpeng, Li Auto Complete China's EV-to-AI-Platform Transition

Li Auto's June 15 unveiling of the Mach M100 — a 5nm automotive-grade AI chip delivering 1,280 TOPS — completed a proprietary silicon trifecta among China's three leading EV startups. The chip achieves 82% AI compute utilization running Li Auto's VLA model, versus 30–40% for general-purpose automotive chips including NVIDIA's Orin. Li Auto has reallocated 50% of its annual R&D budget to AI. NIO's chip subsidiary Shenji has shipped over 550,000 units and raised RMB 2.257 billion at a valuation approaching RMB 10 billion. Xpeng's Turing chip is now powering Volkswagen production vehicles in China.

Why it matters: These are not supply-chain hedges. When a company engineers a chip that benchmarks against data-center hardware, restructures its engineering hierarchy around embodied intelligence, and treats the vehicle as an AI training incubator rather than a product, the identity transition is already complete internally. The chip launch is the external announcement. China's EV sector entered 2026 as the world's most competitive automotive market. It is exiting 2026 as the proving ground for the next generation of physical-world AI computing platforms.

Ford Licenses CATL IP; GM Bets on Sodium-Ion — Both Chasing AI Data Center Power Demand

Ford Motor formally established Ford Energy in May 2026, committing US$2 billion to retrofit a Kentucky plant for 20 GWh of annual storage capacity — built entirely on licensed CATL fifth-generation LFP technology, with CATL holding no equity and collecting an estimated US$250–280 million in annual royalties.

General Motors is taking the opposite route: near-term LFP production through its LG Energy Solution joint venture, paired with a reported US$900 million bet on Peak Energy Technologies to co-develop sodium-ion cells for grid-scale storage, targeting 2028 trial production with a fully domestic supply chain.

Why it matters: The demand signal driving both pivots is AI infrastructure. ChatGPT alone consumes an estimated 500,000 kWh daily; analysts project AIDC-driven U.S. stationary storage requirements could reach 122–245 GWh by 2030. Tesla's Megapack business generated US$12.77 billion in energy storage revenue in 2025 — the template both automakers are chasing. Ford's CATL licensing arrangement accelerates market entry at measurable political cost; GM's approach avoids that exposure but accepts longer timelines and unproven chemistry. When Ford announced its storage pivot in late 2025, its stock rose 25% in a single session — a signal that markets are re-rating legacy automakers capable of demonstrating recurring energy services revenue over one-time vehicle transactions.

Leapmotor Posts Record 81,569 May Deliveries — Then Faces the Hard Questions

Leapmotor recorded 81,569 vehicle deliveries in May 2026 — the highest single-month figure ever posted by a Chinese EV startup, surpassing Li Auto's prior peak of 58,000. Cumulative January–May deliveries reached 263,100 units, with 28.5% exported, primarily through a technology-licensing and local-production agreement with Stellantis that provides access to approximately 850 European sales and service points. The D19 sedan posted 7,000 first-month deliveries at a higher price point. Gross margin has remained positive since Q3 2023, underpinned by a 65% in-house component development ratio.

Why it matters: The record delivery figure is credible because it is margin-positive — a threshold that eluded now-defunct peers Neta and WM Motor at comparable volume peaks. Leapmotor's vertical integration strategy, targeting above 80% in-house component ratios, creates a cost wedge that cannot be replicated quickly by capital-deploying competitors. The Stellantis partnership is a capital-efficiency model that has delivered the highest overseas delivery share among Chinese EV startups. The structural risk ahead is twofold: quality consistency at scale — owner reports cite recurring software issues on the A10 and D19 — and the brand ceiling that will determine whether "China's Toyota" aspiration can extend above the RMB 200,000 price band.

What to Watch Next

The monetization experiments now running across China's AI platforms — Doubao's subscription tiers, Qianwen's transaction-layer integration, Kimi's enterprise ARR pivot — will produce the industry's first real data on willingness-to-pay by Q3 2026. DeepSeek's governance structure and Huawei Ascend training compatibility deserve close monitoring as the AI infrastructure arms race accelerates. In EVs, Huawei's AITO is introducing Gotion and CALB as secondary battery suppliers to cut costs roughly 10% per unit versus CATL pricing — a supply chain restructuring that signals CATL's pricing power is under structural pressure across the industry. Regulatory approval of those supplier transitions remains pending.

Related Coverage:

Leapmotor Breaks EV Delivery Record With 81,569 Units, But the Hard Part BeginsHuawei’s AITO Diversifies Battery Supply Chain, Challenging CATL’s DominanceAlipay's AI Pivot Confronts a Deeper Organizational ReckoningDetroit Giants Pivot to Energy Storage, With CATL's Technology Blueprint in HandNIO, Xpeng, Li Auto Abandon Auto Identity to Claim Next Computing Platform710 Million Users, One Hard Question: Can China’s AI Apps Turn Scale Into Profit?