---
title: "Song Xuetao: Geopolitics, Macroeconomics, and Industry Intertwined—How to Value AI?"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/294033404.md"
description: "Song Xuetao points out that the correction in AI assets is jointly driven by geopolitical, macroeconomic, and industrial factors. Geopolitically, although US-Iran conflicts recur, they are constrained by oil prices and US Treasury yields, putting Trump under pressure to compromise; macroeconomically, oil price pressures keep the Federal Reserve hawkish; industrially, despite strong revenue from cloud providers, surging capital expenditures have turned free cash flow negative, shifting AI trading from \"rewarding input\" to \"verifying returns.\""
datetime: "2026-07-28T08:28:07.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/294033404.md)
  - [en](https://longbridge.com/en/news/294033404.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/294033404.md)
---

# Song Xuetao: Geopolitics, Macroeconomics, and Industry Intertwined—How to Value AI?

This round of adjustments in AI assets is the result of the combined effects of three threads: geopolitics, macroeconomics, and industry.

On the geopolitical front, US-Iran conflicts may recur, but oil prices, US Treasury yields, and the pullback in US stocks are compressing the space for Trump to continue exerting pressure, making TACO (Talk, Announce, Compromise, Operate) a more likely path. On the macroeconomic front, oil price pressures will prompt the Fed to retain the risk of rate hikes, but this is insufficient to prove that current rates are not restrictive enough; maintaining unchanged rates is itself a hawkish statement. On the technology front, while revenue and orders for leading cloud providers remain strong, surging capital expenditures, negative free cash flow, and rising external financing are pushing AI trading from the stage of "rewarding input" into the stage of "verifying returns."

**First, oil prices, US stocks, and US Treasury yields determine the boundaries of Trump's TACO strategy.**

Neither the United States nor Iran truly wants to escalate the war, but both sides want to prove on the negotiating table that they have not lost, so conflicts will recur and will not end easily. However, the ownership of the Strait of Hormuz holds vastly different significance for both sides. For Iran, it is an asset, a trophy, and its most important economic lever when facing the United States; for the United States, it is a concrete test of its long-term global strategic influence.

For Trump, compared to the "long-duration" interest of controlling the Strait of Hormuz, the financial and political pressures he faces are more immediate and urgent.

Both oil prices and Treasury yields have reached another critical point since the March conflict, but the market exhibits a strong learning effect: it did not panic immediately at the outset of the renewed conflict, but only began to significantly reflect war risks after oil prices broke through $100 and tech stocks faced additional adjustment pressures.

At current pressure levels, a new plateau is expected to be sought within a week. This plateau could involve pushing Pakistan to continue mediating, or the US first announcing a suspension of airstrikes and then discussing with Iran an extension of the previous 60-day ceasefire arrangement.

For Trump, abandoning the illusion of a quick victory, making rapid concessions to save face, and returning to the midterm election arena remains the optimal solution. Sharply falling oil prices and eased tightening expectations can at least keep him in the midterm election game. If he continues to struggle or even applies maximum pressure, the situation may drag on for another 1-2 weeks, but the outcome will not change.

The market has strong path dependence capabilities, trading tail risks when facing unknown shocks for the first time, and trading based on existing experience when facing the same shock for the second time. If US stocks had experienced a sharp decline before this, similar to March, it might have formed a buying opportunity similar to that in March. However, the overall impact of this round of geopolitical conflict on US stocks remains relatively mild.

**Second, keeping rates unchanged at the July FOMC meeting is a hawkish statement.**

For the July FOMC monetary policy meeting, the risk of a rate hike can be retained, but a actual hike is unnecessary; keeping rates unchanged is itself a hawkish statement. Currently, the market still prices in at least a 25bp rate hike within the year. Unless Fed Chair Walsh goes against the consensus, the possibility of an immediate rate hike in July is nearly zero.

If the Fed only needs to suppress financial conditions, then "not ruling out further tightening" is sufficient; but if the Fed truly raises rates, it needs to prove that current rates are still not restrictive enough, and this proof is becoming increasingly difficult.

Fluctuations in oil prices over the past two weeks will affect wording and may even keep Walsh in a state of further ambiguity, but it is difficult to reverse the policy direction alone. If the rise in oil prices mainly comes from geopolitical risk premiums rather than irreversible supply gaps, it is more like a phase-specific disturbance than a sustainable constraint for rate hikes. As long as core inflation does not accelerate continuously, and wages and inflation expectations do not rise synchronously, there is no necessity for a rate hike.

**Third, the game of capital expenditure intensifies, shifting from reward to punishment.**

Google's Q2 financial data remains relatively strong, with rapid expansion in revenue, operating profit, and cloud business, and unfilled orders indicating sufficient future demand. However, the market's current focus is not on revenue and orders, but on the lack of clarity in capital expenditure guidance for the next fiscal year, and free cash flow turning negative for the first time since its listing.

Google's single-quarter capital expenditure of $45 billion nearly doubled year-on-year, but the larger the capital expenditure, the more sensitive the company becomes to interest rates and external liquidity, and the harder it is to justify the investment. Google's negative free cash flow (-$5.85 billion) is not an isolated case; starting from the second half of last year, companies like Amazon and Oracle entered negative free cash flow states earlier due to heavy asset investments.

In the past, internet companies were cash cows because profits and cash flows ultimately returned to shareholders, but in the AI era, investors need to accept a new reality—they may have to share the costs for the future "stars and seas" first; there are neither long-term buybacks boosting earnings per share nor stable dividends.

Overall, the current market state is somewhat similar to the period after last October: value and non-AI assets are recovering, differentiation is occurring within tech stocks, and investors are beginning to reward performance, cash flow, and lower capital expenditures while punishing high inputs and weak returns. Rising external financing and credit risks have increased the denominator fragility of tech stocks.

AI investments by large cloud providers are shifting from internal cash flow to external financing. AI-related bond financing in 2026 will exceed at least $500 billion, with capital expenditures by major cloud providers exceeding $700 billion and total US AI capital expenditures approaching $1 trillion. After leveraging, tech stocks become more sensitive to interest rates. Recently, the credit risk of highly leveraged AI infrastructure companies has been significantly repriced, but the overall spread in the tech industry remains low, and risks have not systematically spread.

Free cash flow and credit spreads determine the valuation "denominator," but they do not determine the AI trend. In the second half of last year, the same denominator concerns did not end the rally because model capabilities and revenue growth re-accelerated, with numerator expansion covering denominator fragility. How far the AI rally can ultimately go still depends on revenue and business models on the numerator side.

**AI narrative enters the investment return verification period: from model ARR to US-China technological competition.**

Discussions on AI business models and monetization capabilities will become the main market theme for a considerable period; ultimately, Agents and AI applications must create revenue at the forefront for subsequent cloud computing capital expenditures, GPU purchases, storage price increases, and equipment expansion to be justified. Therefore, the watershed of the market lies in whether the foremost business models can work. If AI cannot make money, the closer to the back end of the industry chain, the more severe the distortion of investment returns, resulting in greater volatility.

On one hand, rapidly growing ARR is currently the most important support for the AI narrative. The combined ARR of leading US model companies as of mid-July was approximately $115 billion; the rapid increase in Codex usage has also driven OpenAI's ARR to re-accelerate. To achieve the ARR growth rate expected by the market, model capabilities also need to continue to leap forward; as long as Agents continue to create token consumption and revenue maintains high growth, the current trillion-dollar capital expenditure expectation still has support.

On the other hand, potential revenue slowdowns may come from various factors: slowing technological progress, low cost-performance ratio of models, slower-than-expected transformation of enterprise workflows, and another pressure appearing in the increased depreciation corresponding to the continuous accumulation of AI net capital. Low-cost Chinese models are another variable: from DeepSeek to Kimi, Chinese solutions are compressing the pricing space for US models and pushing the US to extend competition from technology to distillation restrictions, enterprise usage rules, and pricing power.

All these factors may lead the market to continue its trading scale of rewarding profits and cash flows while punishing high capital expenditures and weak returns.

**Look at narratives (industry trends) during declines, and look at valuations (technical analysis) during rises.**

The market often only looks at trends during rises, becoming more optimistic as prices rise; during declines, it only looks at short-term technicals and risks, becoming more pessimistic as prices fall. A more reasonable approach is exactly the opposite: during rises, focus on technical indicators, valuations, crowdedness, and leverage; after declines, re-evaluate whether industry trends and business models have truly been falsified.

In the short term, attention should still be paid to the blockade of the Strait of Hormuz, oil prices, the global monetary policy cycle, and deleveraging in South Korea, as these determine liquidity and the valuation denominator. But in the medium to long term, the numerator is more important: whether model ARR can continue to grow, whether Agents can enter enterprise workflows, whether AI revenue can cover depreciation and operating costs, and whether hardware supply shortages will be quickly alleviated by expansion in East Asia.

If model revenue and enterprise adoption continue to accelerate, the current adjustment may just be a repricing of capital expenditures; if the foremost business models cannot work, capital expenditures, hardware profits, and valuations all need to be revised downward again.

Risk Warning and Disclaimer

The market involves risks, and investment requires caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial status, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this content is at your own risk.

### Related Stocks

- [GOOGL.US](https://longbridge.com/en/quote/GOOGL.US.md)
- [GOOG.US](https://longbridge.com/en/quote/GOOG.US.md)
- [AMZN.US](https://longbridge.com/en/quote/AMZN.US.md)
- [ORCL.US](https://longbridge.com/en/quote/ORCL.US.md)
- [OpenAI.NA](https://longbridge.com/en/quote/OpenAI.NA.md)
- [ORCL-D.US](https://longbridge.com/en/quote/ORCL-D.US.md)

## Related News & Research

- [Sugar Prices Trade on a Weak Note as Oil Prices Fall](https://longbridge.com/en/news/293781967.md)
- [AI's new North Star: Intelligence per dollar](https://longbridge.com/en/news/293823098.md)
- [Gulf states back plan to let Iran collect voluntary fees to use Hormuz](https://longbridge.com/en/news/294051854.md)
- [The AI Hangover Has Arrived](https://longbridge.com/en/news/294191198.md)
- [AI is quietly becoming an unofficial and potentially unwanted 'third' in relationships](https://longbridge.com/en/news/293839765.md)