Goldman: The 6 AI Themes Defining China’s 2026
In a sprawling research note released on January 21, 2026, Goldman Sachs has outlined the definitive roadmap for China’s internet sector, declaring 2026 a "strategic pivot year." The era of blind hype appears to be fading, replaced by a ruthless focus on execution, monetization, and the desperate race for computing supremacy.
For investors navigating the volatile landscape of Chinese tech, the report offers a crucial signal: the "mega-caps" are stepping up capital expenditure to defend their turf, while the gap between US and Chinese AI capabilities is now entirely dependent on next-generation silicon access.
Here is the breakdown of the six key themes Goldman Sachs believes will dictate the market in 2026.
The "Super Entry-Point" War
The primary battleground has shifted from traditional search to AI-driven "super entry-points." Goldman analysts, led by Ronald Keung, argue that 2026 marks the proliferation of To-C (consumer-facing) assistants that will fundamentally alter user traffic patterns.
The bank notes:
"2026 could mark a proliferation of To-C assistants/AI super entry-points... Mega-caps AI chatbots/assistant functionalities to increase inference costs before potential for monetization (online marketing, commissions) to take place over the longer-term."
This transition threatens the traditional search market. The report highlights Alibaba and its Qwen app, as well as Tencent leveraging WeChat mini-programs, as potential winners in creating these dominant interfaces.
The Hardware Gap: Rubin vs. Blackwell
Perhaps the most geopolitical of the themes is the hardware deficit. Goldman posits that the performance gap between Chinese and US AI models over the next 12-18 months will be determined by access to Nvidia’s next-generation architecture.
While Chinese firms are optimizing software, the raw hardware advantage of the West is accelerating. Goldman cites Nvidia CEO Jensen Huang’s recent CES 2026 comments regarding the Rubin architecture:
"Rubin’s inference performance is five times higher than Blackwell’s, training performance is 3.5 times greater, and costs are reduced by tenfold."
Goldman warns that despite domestic policy support, Chinese players face "bottlenecks in production capacity and the software ecosystem," though they retain advantages in electricity and infrastructure.
The Token Explosion and Capex Surge
Usage is exploding, and with it, the cost of doing business. ByteDance recently announced that daily token usage for its Doubao Large Model exceeded 50 trillion in December 2025, ranking it first in China.
This hypergrowth in inference demand is forcing a massive uplift in capital expenditure. Goldman has pinned Alibaba’s FY26-28E capex forecast at a staggering RMB 454 billion yuan (US$63.1 billion), one of the highest estimates on the Street.
"We continue to see upside to capex for China hyperscalers... given strong AI demand and its continued commitment/focus on external customers."
Adtech: The Shift to "Answer Engines"
The monetization engine is evolving. Goldman forecasts a shift from traditional Search Engine Optimization (SEO) to "Generative Engine Optimization" (GEO). As AI chatbots become the primary way users find information, advertising budgets are shifting toward ROI-based ad products.
This benefits platforms with closed-loop ecosystems. The report points to Temu and its site-wide marketing tools as a prime example of this shift toward algorithmic, ROI-centric ad spend.
Global Ambitions: Open Source as a Trojan Horse
Chinese AI isn't staying behind the Great Firewall. Goldman observes a strategic bifurcation: companies are keeping their best "state-of-the-art" models closed while flooding the global market with efficient open-source variants to capture market share.
"We expect continued inroads of Chinese AI models into the global markets... We believe coding (Alibaba, DeepSeek and other AI independents) and multi-modal models (e.g. Kwai’s Kling) will be able to differentiate themselves on cost/speed."
Model Breakthroughs: The Physical World
Finally, 2026 will see AI move beyond text. Goldman predicts breakthroughs centering on "world models" and "physical AI"—essentially, AI that understands and interacts with the 3D physical environment. This is the next frontier for robotics and autonomous agents.
Investment Implications
Goldman Sachs remains bullish on the incumbents who can afford this expensive transition. They view Alibaba (for its full-stack AI) and Tencent (as the key application proxy) as the best-positioned mega-caps for the 1-3 year horizon.
However, they also highlight a Buy rating on Temu, citing its "valuation discount, strong value-for-money user mindshare and Temu growth," noting that Temu’s global monthly active users held steady at 534 million in December 2025 despite regulatory headwinds.
In summary, 2026 is the year the bill comes due for the AI pivot. For China’s tech giants, it is a year of massive spending, fierce competition for the "entry point," and a race against time to close the silicon gap.