Newsletter commentary Nov 2025

AI Investment Enters the Second Stage, Competition Emerges

The market faced challenges in November. As the year-end approaches, the market is reevaluating opportunities for the coming year. Coinciding with this, doubts have arisen about the AI investment chain, and Google’s new model has prompted a reassessment of the existing investment landscape. Additionally, the distribution of some domestic phased expansionary policies has been completed, with new policies under formulation for the future. The market is exploring new investment directions. Bank stocks have risen, while the communications sector has performed strongly driven by news of additional orders for optical modules. Other industries have posted lackluster performance.

Google’s new model brings new insights in both training methods and capabilities, posing significant challenges to the previously dominant OAI-NV-centric investment framework. The former leads in model innovation, while the latter offers the best cost-effectiveness in hardware and ecosystem. For a long time, this framework faced no substantial competition. However, the situation has now changed. OAI has acknowledged that Google’s new model has surpassed its own and may require some time to catch up, as OAI might have misallocated resources in model training, focusing more on skills rather than pre-training intelligence. Google trained its new model on its own TPUs, challenging the uniqueness of NVIDIA’s GPUs. Previously, there was no alternative, as NVIDIA’s chips led in both flexibility and performance. Now, only flexibility may remain, marking a potential turning point for downstream manufacturers to reconsider their choices. Meta and Anthropic have already begun taking action.

AI investment has largely passed the initial stage, where a few companies pioneered the market and achieved dominance. However, with the entry of new players offering diverse technological alternatives, the landscape is shifting. Initially, dominant players held a significant lead, excelling in almost every aspect, such as single-chip performance, scale-up, scale-out, scale-across, and business ecosystems, as they defined the market. In this phase, no component was cheap, but this was not an issue, as investment driven by necessity was price-insensitive. However, unlocking real B2B and B2C demand is an economic issue, and reducing costs to align with the current economic pricing system will become imperative. Over time, more companies are offering alternative solutions. While these may not be the best in every aspect, they enable more globally optimal choices for themselves. As the focus shifts from model training to application inference, pricing will become an increasingly important consideration. Models will gradually converge, leading to a transition from universal solutions to specialized, high-speed solutions. Google’s full-stack R&D advantage is a product of this trend.

For investors, this represents a significant shift. Similar to the breaking of a monopoly, greater choices will lead to a larger market and lower prices.

A potential variable is that if models begin to converge and AGI remains elusive in the near term, the winners will be those with richer, more cost-effective application ecosystems. This could reshape the logic of AI investment. While Google has disrupted the OAI-NV-centric narrative, it also presents an opportunity for China’s AI narrative to gain more recognition—another Google, or many more Googles.

The global economy remains in a challenging position. U.S. stocks faced significant pressure after hawkish remarks from Federal Reserve officials, until dovish comments on the preceding Friday raised expectations of interest rate cuts, leading to a market recovery. The Fed’s dilemma is not merely about defending its formal independence but reflects the genuine difficulties of the current situation. The K-shaped development, rising living costs, and affordability issues contrast sharply with record-high stock indices. The former has already caused the Republican Party to lose many local elections, while the latter’s new highs are closely tied to AI investment. What if AI, instead of reaching for the stars, enters a period of stagnation?

China’s economy weakened quarter-on-quarter in Q4, as expected. Domestic policies have tightened marginally on the basis of easing: strengthening constraints on local government investment and debt, reducing tax incentives, and accelerating the cleanup of financing platforms. Under government planning, the driving force of investment and exports will diminish, increasing reliance on consumption. How policies can help consumption recover is a major challenge. Of course, investing in Chinese companies is gradually shifting away from being solely about betting on China’s macroeconomy, as more Chinese companies are achieving global operations.

In the past, investors favored inflation because EPS is a nominal value. We believe the flip side of inflation-affordable living costs-was overlooked amid the flood of EPS. However, as this issue has now become prominent, it can no longer be ignored. Supply capacity may become a tangible investment similar to gold, which is not impossible. China, with its strongest supply capacity, was once associated with deflation but may now be recognized as the country with the strongest supply capability.