Dongwu Securities: Can NASDAQ go bottom after the midterm elections?

The Zhitong Finance App learned that Dongwu Securities released a research report saying that it may be cautious about bottoming out the NASDAQ index. In the next 12 months, the growth rate of AI infrastructure investment in the US will face downward pressure, and the profit expectations of upstream hardware companies are still at risk of adjustment. An index rebound may occur, but Chen Li of Dongwu Securities is more willing to place new investment opportunities in downstream applications, especially to C AI agents that can use cost-effective models to reduce service costs.

If investment in AI infrastructure actually slows down, it will gradually become clear how much the US economy will be affected. The data provided by the chief analysts of major overseas banks is very different. Morgan believes that 67% of the increase in the US economy is related to AI, while Goldman Sachs believes that only up to 10% is contributed by AI. The year ahead will be a good time to test their views. Whether the slowdown in AI growth will be enough for the US to quickly enter recession is still uncertain. The key is whether consumption, services, and other corporate investments can withstand these changes in demand, and whether financing pressure extends to more sectors.

The main views of Dongwu Securities are as follows:

Today's consistent expectations and variables after the US midterm elections

On November 4, 2026, the first trading day after the election results were announced, the Dow rose about 3%, the NASDAQ rose about 1.4%, and WTI crude oil futures fell more than 7%. Markets expect the Republican Party's defeat to bring about policy adjustments, and the political costs of the Iran war have also received more attention. The next day, the Nasdaq fell 5.5%, and the semiconductor index fell about 12%. Investors focused on discussing how much resistance the Democratic Party's victory in congressional and state elections would put on data center construction.

Expectations that the war will cool down are well-founded. Voters are dissatisfied with the cost of living, and the government continues to bear the pressure of war expenses and rising energy prices, making it more difficult politically.

But it remains to be seen how the president responds to the failure and how Congress uses funding and investigative powers. Iran also has its own demands for military action and negotiations. Even if the US increases its will to end the war, order in energy supply and transportation will not necessarily be restored immediately. The current situation in the Strait of Hormuz fully shows that this is no longer something that the US can decide unilaterally. Even if the US unilaterally withdraws from the battlefield, Iran, Saudi Arabia, Israel, Palestine and other forces may continue to destabilize the situation in this region.

On the AI side, market concerns have been implemented into projects and financing. Some investment banks estimate that the potential suspension of construction could affect about 30% of the expected additional data center capacity in the US in 2028. Delays in approval may result in project relocation, and may also prompt companies to increase energy, water, and community compensation expenses. Facilities that have been licensed and connected to electricity may even become more scarce as a result.

But for companies that rely on external financing, delays are indeed dangerous. The computer room has not yet been put into use, yet interest on the loan still has to be paid. On November 5, CoreWeave's five-year credit default swap spread rose to about 1,289 basis points, and the stock price fell about 10% on the same day. This set of transaction data shows that investors are already demanding higher compensation for financing and debt repayment risks. The widening of credit spreads may also further increase the financing costs of enterprises.

Entrepreneurs who have become global “heroes” in the AI industry may want to take a close look: whether their business practices actually provide benefits to ordinary people, rather than simply increasing their water, electricity, oil and gas prices. Of course, the Democratic Party is not against AI for the sake of opposing AI. If AI companies pay real money to promote local employment, no one can get along with the money. But this necessarily means that businesses need larger amounts of capital.

Interest rate markets, on the other hand, remain concerned about inflation. The employment data released on November 6 was stronger than expected, and the unemployment rate fell to 4.1%. The market's pricing probability for the December rate hike increased from about 70% to 76%. However, if AI projects slow down significantly, equipment procurement, construction activities, and capital requirements will all be affected; if the war eases, the energy shock may also abate. At the same time, wages and service prices are sticky, and falling oil prices may not be enough for the Federal Reserve to turn around quickly.

Judgments about the US economy are therefore divided. Consumption and employment are resilient, supporting the view that the economy can withstand some AI investment adjustments. However, shrinking stock wealth, tightening financing, and delays in corporate recruitment may also affect household expenses in a few months. AI investment, which the market is currently worried about, can coexist in the short term with expectations of continuing interest rate hikes; if investment contraction continues to expand, the outlook for the US economy in 2027 will be overshadowed.

An opportunity to bottom out the Nasdaq? Will it be an AI app

Over the next 12 months, the growth rate of new US AI infrastructure investment will face greater slowing pressure. Higher interest rates, energy costs, and changes in the political environment will raise the actual investment threshold for projects. This will affect the profit expectations of upstream hardware companies.

I prefer to focus on companies that can generate revenue from downstream applications, especially AI Agents for individual users.

There is also an important reason to be optimistic about this direction: what the public needs is a product that is affordable and can reliably complete tasks, and does not necessarily need to use the strongest and most expensive model for every service.

We expect that as the product matures, To-C applications will pay more attention to the cost performance ratio of the model, and the lightweight and flexible small model will be more widely used. For application companies, how much it costs to complete each task is often more important than how many parameters the model behind it has. Users pay the same, and the cost of model call and task execution is reduced, so enterprises have the opportunity to leave more profits, and they can also expand the user base by lowering prices.

This is why I think investment opportunities may shift downstream. In the past, the market focused on continuous breakthroughs in model capabilities, as well as how many chips and how much electricity are required for breakthroughs. Next, part of the commercial value may come from another ability: turning existing model capabilities into services that the public is willing to buy at a lower cost.

AI skits, video production, and other ancillary applications are similar business opportunities. However, different tasks require different model capabilities, so it cannot be assumed that all applications can be completed by a small model. The common direction is to reduce the cost of service delivery through model selection and process optimization.

As a result, we may be wary of depreciating the NASDAQ index. In the next 12 months, the growth rate of AI infrastructure investment in the US will face downward pressure, and there is still a risk that the profit expectations of upstream hardware companies will be adjusted. An exponential rebound may occur, but I prefer to place new investment opportunities in downstream applications, especially to C AI agents that can use cost-effective models to reduce service costs.

If investment in AI infrastructure actually slows down, it will gradually become clear how much the US economy will be affected. The data provided by the chief analysts of major overseas banks is very different. Morgan believes that 67% of the increase in the US economy is related to AI, while Goldman Sachs believes that only up to 10% is contributed by AI. The year ahead will be a good time to test their views. Whether the slowdown in AI growth will be enough for the US to quickly enter recession is still uncertain. The key is whether consumption, services, and other corporate investments can withstand these changes in demand, and whether financing pressure extends to more sectors.

At the end of 2006, strong employment and consumption convinced investors that the real estate problem could be solved. As a result, the real estate market declined, and it had a severe impact on the financial system, which eventually led to a deep recession in the US economy.

These issues will add great uncertainty to the 2027 stock market.