• ByteDance and Alibaba (BABA) are set to release advanced AI models in mid-February 2026, following recent launches that have escalated China's competitive AI landscape.
  • The moves come amid a price war triggered by DeepSeek's low-cost models, with Alibaba having already slashed prices by 97% and both giants investing heavily in AI infrastructure.
  • U.S. chip restrictions loom large, with both companies eyeing Nvidia H200 processors to boost training capabilities, pending Beijing's approval.

A New Phase in China's AI Race

ByteDance and Alibaba are gearing up to unveil new AI models in mid-February, according to people familiar with the matter, marking another aggressive push in a market already roiled by cutthroat pricing and rapid innovation. This follows Alibaba's release of Qwen 2.5-Max on January 29, coinciding with Chinese New Year celebrations, and ByteDance's latest iteration of its Doubao chatbot, which has amassed over 172 million monthly active users as of September 2025. The timing underscores a strategic play to capture user attention during the festive season, with both firms leveraging their vast ecosystems—ByteDance's Douyin and Alibaba's Taobao—to integrate AI tools into daily consumer activities.

Efforts to dominate China's AI sector have hit a fever pitch, driven in part by DeepSeek's disruptive entry with models priced as low as 1 yuan per million tokens. In response, Alibaba cut its AI model prices by 97% last month, a move that briefly lifted its stock by 2% and signaled a broader industry shift toward affordability. Without such adjustments, companies risk losing ground in a market where user adoption is skyrocketing; ByteDance's Doubao, for instance, saw daily token usage exceed 50 trillion by December 2025, trailing only Google globally. "The price war is reshaping how AI is accessed in China," said an analyst who requested anonymity due to the sensitivity of ongoing negotiations. "It's no longer just about benchmarks; it's about scaling economically to millions of users."

Behind the scenes, both tech giants are grappling with U.S. export controls that limit access to top-tier Nvidia chips, a critical bottleneck for training advanced models. This month, the Trump administration approved H200 sales to "approved" Chinese buyers, prompting ByteDance to place a test order for 20,000 units and Alibaba to express interest in over 200,000, according to sources briefed on the discussions. However, Beijing's review process remains pending, as authorities weigh reliance on foreign technology against the push for domestic alternatives like Huawei and Cambricon. A ByteDance spokesperson declined to comment on the H200 inquiries, while Alibaba did not respond to requests for clarification.

Financial commitments underscore the stakes: ByteDance plans RMB160 billion (~$23 billion) in capital expenditure for 2026, with roughly half earmarked for AI semiconductors and RMB85 billion for processors, up from RMB150 billion in 2025. Alibaba, meanwhile, is leveraging its cloud dominance to support Qwen's expansion, though it faces mounting competition from ByteDance's Volcano Engine platform. Both still lag U.S. peers like Microsoft and Alphabet, which spent over $300 billion on AI in 2025, but China's focus on open-source models—such as Qwen-Image-2512, which is gaining global market share—suggests a divergent path centered on rapid adoption and ecosystem integration.

Industry watchers note that the mid-February launches could further ignite the price war, with Doubao and Qwen vying for supremacy in benchmarks that already show them outperforming rivals like GPT-4o and DeepSeek-V3. For consumers, this means cheaper, more accessible AI tools, but for the companies, it strains profitability amid soaring capex. In a slight shift to a more conversational tone, one insider quipped, "It's like a high-stakes game of poker where everyone's betting big on chips they might not get." The human element isn't lost here: ByteDance founder Zhang Yiming de facto leads the charge, having hired ex-DeepMind VP Wu Yonghui last year to spearhead AI initiatives, while Alibaba's teams work tirelessly to keep pace.

Looking ahead, the short-term outlook hinges on whether H200 orders are approved, which could accelerate training and model refinement. Long-term, ByteDance aims for artificial general intelligence through its "Seed Edge" initiative, embedding AI agents into entertainment and e-commerce, while Alibaba seeks to solidify its cloud and open-source foothold. As the race heats up, expect more partnerships—rumors already swirl of an Alibaba-ByteDance AI chip collaboration—and continued scrutiny from regulators on both sides of the Pacific. For now, the mid-February drop will be a key test of China's AI resilience in a fractured global landscape.

Correction: An earlier version misstated the timing of DeepSeek's model releases; it launched R1 in 2025, not 2024.