CATL's AI Infrastructure Push Exposes Pressure From Carmakers

CATL's reported interest in investing in DeepSeek should not be read only as another sign that artificial intelligence has become attractive to China's largest industrial companies. It also points to an uncomfortable shift in CATL's core business: the battery leader is looking for growth outside a vehicle market where its customers are becoming less willing to live under one supplier's pricing power.

The company has not historically been known for direct bets on general-purpose foundation-model companies. Its recent deal activity instead shows a clearer industrial logic around AI data centers. In April, CATL agreed to invest 4.1 billion yuan for a 49 percent stake in a shareholder of Zhongheng Electric, a supplier of high-voltage direct-current power systems used in intelligent computing centers. In May, it moved again, paying 6.4 billion yuan for a 38.1 percent stake in VNET, becoming the data-center operator's largest shareholder.

Those two moves sit close to CATL's existing competence. AI data centers are not only server rooms. They are electricity-intensive industrial assets that require stable grid access, backup systems, power conversion, thermal management and increasingly large storage capacity. A single high-density rack can consume many times the power of conventional equipment, and a large campus can resemble a small city in electricity load. CATL understands batteries, storage and power systems better than most new entrants.

The DeepSeek angle, if completed, would add the demand side of the same chain. A model company needs training and inference capacity. That capacity needs data centers. Those data centers need power architecture and storage. From that perspective, CATL would be trying to place itself near the electricity spine of China's AI economy, not merely buying a fashionable AI name.

The darker reading is that CATL is being pushed there by automakers. Chinese car companies have spent several years trying to dilute their reliance on CATL. BYD has its own battery arm and sells to outside customers. Geely has built internal battery-pack capability and shifted more work across suppliers. Great Wall has Svolt. GAC Aion moved closer to CALB. Nio, Li Auto, Xpeng and Leapmotor have all pursued more diversified procurement. The message is blunt: carmakers do not want one battery supplier to control cost, delivery rhythm and technology access.

The economics explain the hostility. Batteries can account for a large share of electric-vehicle cost, while vehicle manufacturers are trapped in a price war and shrinking margins. CATL's 2025 net profit and margin profile looked far healthier than many of its customers' vehicle businesses. That creates resentment even if CATL's position was earned through scale, engineering and execution. The more brutal China's EV market becomes, the less tolerance automakers have for a supplier capturing the most reliable profits in the chain.

The result is a strategic squeeze. CATL remains powerful, but its peak share in China's power-battery market has already fallen from earlier highs. Automakers are not abandoning it wholesale; they are making sure it is no longer unavoidable. That is a major difference. Once customers believe credible alternatives exist, they can negotiate harder, spread orders and demand more price concessions.

AI infrastructure offers CATL a second battlefield. Demand for computing power is large, long-duration and electricity hungry. Model companies and cloud platforms need stable capacity, and the cost of downtime is severe. Storage, backup power, high-voltage direct-current systems and energy management are not peripheral expenses; they can determine whether a data center operates efficiently. CATL can argue that its industrial background gives it a role beyond passive financial investment.

But this is not a painless escape. Data centers are capital-heavy assets exposed to land, grid approvals, utilization cycles and large-customer bargaining power. If the market overbuilds capacity, the same problem CATL is trying to avoid in batteries could reappear in computing infrastructure: too much supply, falling returns and pressure from the biggest buyers. AI demand is strong, but it is also volatile, and today's favored model company may not control tomorrow's traffic.

There is also client-concentration risk. VNET's large capacity arrangements with internet platforms may make the industrial story easier to tell, but dependence on a handful of technology customers can become a weak point. If those customers change suppliers, delay projects, renegotiate pricing or bring more infrastructure in-house, CATL's AI-adjacent investments may deliver less strategic control than hoped.

The deeper question for CATL is whether it can turn energy storage and power systems into a durable AI-infrastructure franchise before vehicle-battery margins erode further. The answer is not obvious. CATL has industrial advantages, yet it is moving into a market where cloud companies, data-center operators, power-equipment suppliers and local governments all have competing claims.

For investors, the reported DeepSeek interest should therefore be treated with caution. It may expand CATL's access to a high-growth sector, but it also reveals that the company is less comfortable relying on the old EV-battery formula. A leader that has to buy its way into the next demand pool is still a leader. It is also a leader that can see pressure coming from its own customers.