China EV Supply Chain Faces Battery and AI Shift

China’s EV industry delivered a revealing snapshot of its next phase this week: GAC Aion moved to contain battery-quality concerns with an expanded warranty program for affected vehicles, AI chipmaker Axera spun off a dedicated inference-computing unit to chase edge and embodied-AI growth, and semiconductor giant STMicroelectronics posted a strong Q2 2026 recovery while signaling that autos remain important—but no longer the only growth engine. Taken together, these developments show how the Chinese EV market is being shaped not just by vehicle launches and price wars, but by after-sales accountability, battery durability, automotive semiconductors, and the race for AI compute.

GAC Aion Extends Warranty on 177Ah Battery Vehicles

GAC Aion has announced a warranty-service upgrade for some vehicles equipped with CALB 177Ah lithium iron phosphate (LFP) battery packs after battery failures were reported in certain AION S cars used for commercial operations.

According to the official notice and after-sales explanation, the affected scope includes:

• AION S
• AION V
• AION Y
• Vehicles equipped with CALB 177Ah power batteries

The core issue appears to be concentrated in high-intensity ride-hailing use cases rather than private ownership. These vehicles can reportedly cover more than 300 km per day, with frequent fast charging and deep discharge cycles—conditions that place much greater stress on LFP battery systems.

What Aion Changed

For private ride-hailing vehicles, GAC Aion upgraded the battery warranty from:

The company also said it will strengthen big-data monitoring and provide:

• Free inspections for abnormal vehicles
• Free repairs where necessary
• Free battery replacement in qualifying cases

That is an important signal in China’s highly competitive EV market, where brand trust increasingly depends on after-sales execution as much as headline specs.

Why No Formal Recall Yet?

One point of controversy online has been why GAC Aion did not immediately announce a recall. The answer is regulatory: in China, an automaker cannot simply declare a recall on its own timetable. Recalls must follow a formal process involving technical validation, reporting, and filing with the State Administration for Market Regulation.

In practice, companies often issue preliminary customer notices before a formal recall process is completed. That appears to be the case here. The report suggests GAC Aion and its supplier may still be working through follow-up product and service measures with relevant authorities.

Compensation for Earlier Repairs

GAC Aion also said that users of 177Ah battery vehicles who previously experienced failures under normal use and paid for repairs themselves may apply for reimbursement or compensation. To do so, owners must provide supporting documents such as:

• Repair work orders
• Invoices
• Relevant service records

For Chinese EV buyers—and especially ride-hailing operators—this is a meaningful step. Battery durability under intensive duty cycles has become a major differentiator in the fleet and mobility segment, where total cost of ownership matters more than marketing promises.

Axera Spins Off AI Inference Business as Compute Race Intensifies

In a separate but highly relevant development for the EV technology stack, Chinese AI chip company Axera announced the creation of a wholly owned subsidiary called Axera Computing. The new unit will take over AI inference-related business, separating it from the parent company as competition in the compute market heats up.

Axera, which only surpassed RMB 1 billion in annual revenue last year, is effectively restructuring to respond faster to a market defined by rapid iteration, heavy R&D spending, and increasingly brutal pricing pressure.

What Axera Computing Will Do

The new company inherits the parent’s “AixTongyuan” NPU technology route and is focused on:

• Edge AI inference
• End-side AI compute
• Compute cards
• Inference servers
• Physical AI applications

Target scenarios include:

• Embodied intelligence
• Smart office systems
• Retail
• Industrial applications

At WAIC 2026, Axera showcased its Yuanxi series inference cards, with claimed compute performance of more than 1,000 TOPS. These products are expected to form the initial lineup of the new subsidiary.

Why It Matters for Automotive

Although this is not an automaker story on the surface, it matters for EVs because Axera’s customer base reportedly includes a substantial number of:

• Automotive electronics companies
• Smart hardware manufacturers

Those customers are highly sensitive to:

• Compute cost-performance
• Supply-chain security
• Product continuity
• Fast deployment cycles

That is increasingly relevant as software-defined vehicles add more in-cabin AI, driver assistance functions, sensor fusion, and edge inference workloads. Chinese EV makers are under pressure to secure not just batteries and power semiconductors, but also domestic AI compute options that can support cost control and geopolitical resilience.

The Strategic Logic Behind the Spin-Off

Axera founder Qiu Xiaoshen described the move as a strategic decision, but the market sees a deeper commercial rationale. The AI inference market is crowded by:

• Nvidia
• AMD
• Cambricon
• Horizon Robotics
• Other domestic inference-chip players

Axera’s 2025 financial report showed R&D spending exceeded 40% of revenue, highlighting just how capital-intensive the sector has become. A standalone entity could improve:

• Decision-making speed
• Cost discipline
• Product-market focus
• Customer-specific execution

Still, the challenge is clear: AI inference hardware is increasingly vulnerable to commoditization. High TOPS figures alone are not enough to build a long-term moat. The real test will be mass-production delivery, software ecosystem support, and customer retention.

STMicroelectronics Recovers in Q2—but Guidance Disappoints

Further upstream in the EV supply chain, STMicroelectronics reported a much-improved Q2 2026 result on July 23, underscoring the ongoing recovery in automotive and industrial semiconductors.

The company posted:

Despite the headline beat, investors focused on weaker profitability and soft guidance. The midpoint of ST’s Q3 revenue guidance was just $3.7 billion, well below analysts’ average estimate of $3.9 billion. After the earnings release, the stock reportedly fell as much as 17% intraday in European trading.

What Drove the Mixed Reaction?

The weakness was not about top-line recovery alone. Analysts pointed to several pressure points:

• Operating profit missed expectations
• EBITDA and free cash flow also came in below consensus
• Earnings quality was affected by asset impairment, restructuring, and accounting impacts linked to the acquisition of NXP’s MEMS business

At the same time, there were some constructive signals. Inventory days fell from 140 days to 126 days, moving below the company’s normal target level. That suggests the demand cycle may be stabilizing after the long semiconductor correction.

Segment Performance

ST’s business mix was uneven:

For the EV industry, the underperformance in power and discrete is particularly notable. Power semiconductors remain central to electric drivetrains, onboard chargers, DC-DC converters, and energy management systems. Weakness here suggests that while parts of the auto semiconductor market are recovering, not every category is rebounding at the same speed.

Autos Are Recovering—But AI Is the Bigger Growth Story

ST said its automotive business rose 14% quarter-on-quarter and 16% year-on-year, while industrial revenue increased 20% quarter-on-quarter and 34% year-on-year. Communications equipment and computer peripherals jumped 50% year-on-year.

The biggest highlight, however, was AI infrastructure. ST raised its AI data-center target again and now expects:

• More than $1 billion in AI data-center revenue in 2026
• Well above $2 billion in 2027

CEO Jean-Marc Chery said Q4 revenue could exceed $4 billion, supported by AI data-center demand and low-Earth-orbit satellite communications projects.

This is a crucial signal for anyone tracking Chinese EVs: automotive semiconductors remain important, but global chip suppliers are increasingly chasing AI infrastructure as the faster-growth, higher-excitement market. That could affect future capital allocation, product roadmaps, and capacity priorities across the broader electronics supply chain.

Comparison: Three Stories, One Supply-Chain Reality

These three developments sit at different layers of the stack, but they are closely connected.

Why This Matters

China’s EV market is no longer just a story about who sells the most cars. It is increasingly about who can manage the entire technology chain—from battery life in demanding real-world use, to secure and affordable AI compute, to resilient semiconductor sourcing.

A few broader takeaways stand out:

• Battery reliability is becoming a brand issue. For ride-hailing and fleet users, durability and service response can outweigh acceleration, range claims, or cabin tech.
• AI is moving closer to the vehicle edge. As more processing shifts into cars, domestic inference-chip players have an opening—but only if they can deliver at scale and at the right cost.
• Semiconductor competition is changing priorities. Global chipmakers still care about autos, but AI data centers are attracting a growing share of attention and investment.
• Supply-chain resilience matters more than ever. Chinese EV makers need dependable partners across batteries, power devices, sensors, and AI processors to protect margins and product cadence.

Global Implications

For overseas observers, these stories offer a useful correction to the common view that Chinese EV competition is purely about sticker prices. In reality, the market is maturing into a full industrial contest involving:

• After-sales responsibility
• Battery life-cycle management
• Automotive-grade semiconductors
• AI computing independence
• Vertical integration and supply-chain control

That matters beyond China because Chinese EV brands are expanding globally. Their competitiveness abroad will depend not only on low-cost manufacturing, but on whether they can sustain quality, software capability, and component security over millions of vehicles and years of use.

What to Watch Next

Several follow-up questions now matter:

• Will GAC Aion and CALB move from service measures to a formal recall filing?
• Can Axera Computing turn its 1,000+ TOPS promise into high-volume commercial deployments, especially in automotive electronics?
• Will ST’s improving automotive demand translate into stronger profitability in power semiconductors, or will AI continue to overshadow the auto market?

The bigger picture is clear. China’s EV race is entering a more demanding phase—one where reliability, compute, and supply-chain execution may prove just as decisive as sales volume.