2026 Smartphone Shipments Forecast to Decline 13.9% Year-on-Year
When Doubao announced its official monetization starting late June and integration with Douyin's e-commerce, this sucker punch wasn't aimed at users but at the anxious nerves of the entire large model sector. ByteDance finally tore away the fig leaf of "fueling with love," acknowledging an industry-wide tacit truth: the window period for burning cash to acquire users has closed.
Analysis
When Doubao announced its official monetization starting late June and integration with Douyin's e-commerce, this sucker punch wasn't aimed at users but at the anxious nerves of the entire large model sector. ByteDance finally tore away the fig leaf of "fueling with love," acknowledging an industry-wide tacit truth: the window period for burning cash to acquire users has closed.
In the same week, news arrived from Silicon Valley—major tech giants had begun limiting employee token quotas. This signal was more glaring than any earnings report. When even OpenAI and Google are tightening belts internally, what confidence do those startups still surviving through a cycle of "funding, cash-burning, and re-funding" have to sell their stories?
The most ironic is Zhipu and MiniMax rushing to IPO on the A-share market. As overseas valuation logic weakens and US dollar funds become harder to secure, they've swiftly pivoted, hoping to find new "greater fools" in the A-share market. Behind this urgency lies the dilemma of homogenized model capabilities—when everyone can call open-source foundations and build "functional but unremarkable" products, the competition boils down to who can monetize faster and survive. But can retail investors in the A-share market truly absorb the valuation bubble of these "AI concept stocks"?
Looking at Unitree Robotics' IPO approval, its founder Wang Xingxing's net worth may exceed 14 billion. The robotics sector remains hot, but the capital market clearly favors the "hard tech" narrative. In contrast, pure software-based large model companies are stuck in an awkward middle ground: unable to reach the holy grail of AGI above, yet unwilling to become mediocre API service providers below.
Counterpoint's forecast of a 13.9% plunge in smartphone shipments by 2026 is, in a way, a mirror of AI's implementation struggles. Hardware manufacturers are betting on AI phones and AI PCs to spark an upgrade wave, but consumers aren't fooled—what can your large model do more for me? Is it worth switching to a new phone for this? The answer clearly isn't loud enough yet. When NVIDIA proclaims "reinventing the PC," figures like Lei Jun don't hear a bugle call but an alarm: if AI is merely about labeling products and adding feature portals, how is this different from the marketing gimmick of "Internet+" years ago?
Galaxy Securities' research report on craft beer, nestled among a pile of AI news, feels refreshingly clear. At least its logical chain is straightforward: rising costs → structural optimization → catalyzed by the World Cup in peak season. In contrast, the AI industry is flooded with research reports and PR materials brimming with correct but hollow words like "empowerment," "disruption," and "ecosystem." The real question should be: when token costs fall and model capabilities converge, what is your moat? Is it a data flywheel or channel monopoly?
Yin Xi, Harvard's youngest Chinese professor, joining OpenAI might seem unrelated to China at first glance, but it stings the Chinese AI academic community—top talent is still voting with their feet. We've invested heavily in computing power subsidies and policy support, but when it comes to the talent ecosystem for fundamental research, the gap cannot be bridged by money alone.
Piecing together this week's news reveals a panoramic view of the industry: commercialization anxiety, talent outflow, hardware stagnation, and capital retreat. AI isn't failing—it's the end of the illusion phase. What comes next isn't a contest of model parameters or funding amounts, but of who can truly identify viable payment scenarios and survive during contraction. Doubao's willingness to charge at least shows ByteDance sees the direction clearly. As for the other players—whether they'll keep drawing pies on PowerPoint or deliver tangible commercialization results—the market will have its say, and it won't be lenient.
Disclaimer: The above content is generated by AI and is for reference only.