Tao Dong: Has the Bear Market for AI Stocks Arrived?

Money doesn't give you the right to do whatever you want! The market delivered a loud slap in the face, letting the hottest tech companies in the stock market know who the real boss is.

The selloff in AI stocks has recently spread from NeoClouds to Hyperscalers, as investors have lost patience with tech giants burning through free cash flow. Alphabet, Amazon, and Meta plunged, while Intel's stock suffered a cliff-edge collapse, falling 29% in a single day. Anthropic is reportedly unable to obtain an investment-grade bond rating, and OpenAI is reportedly being forced to slash the amount it aims to raise in its IPO to $20 billion. Next, if demand fails to hold up, the massive orders for semiconductor chips will also need to be reassessed; the Philadelphia Semiconductor Index fell 7.2%, and Nvidia dropped 4.2%.

By contrast, the bond market's reaction to interest rate policy was relatively mild. The rate decisions of the three major central banks—the US, Europe, and Japan—the elections in Brazil and Israel, funding concerns at new cloud companies, and the oil/diesel supply shortage all failed to stir up much of a splash in the bond market. The US 2-year Treasury yield fell 2 basis points, while the 10-year was flat. Hit by street unrest in France, the French 10-year Treasury yield briefly broke above 5%, and the spread between French and German government bonds widened. As US-Iran friction intensified, crude oil prices continued to climb, and the average US diesel price hit a record high. In this turbulent autumn, the dollar's effective exchange rate continued to rise, gold plunged, and silver suffered a steep fall. Following news that hackers stole Ether on a trading platform, cryptocurrency prices slumped.

The AI bull market has run into its largest wave of selling in the past four years, and investors are re-examining the investment logic behind massive infrastructure spending. Investing in AI infrastructure is no longer a reason to chase stocks; what matters more is earnings certainty, and companies must also protect their free cash flow. A consensus is gradually forming in the market that the growth momentum of AI infrastructure construction may be peaking. If so, it is necessary to assess how much of the orders currently accumulated by AI hardware companies may be canceled.

The AI revolution is undoubtedly the greatest revolution seen in the past two hundred years, and in the long run it will have a huge impact on human production, life, and even society. However, the prospects of the AI revolution and the financial condition of AI companies are two different concepts. The biggest contradiction in the stock market's AI theme is that the companies spending heavily on AI infrastructure and the hardware companies making big money are separate. AI investment is awkwardly running into cash flow difficulties: not only do the weaker new cloud vendors lack funds, but even the deep-pocketed tech giants have burned through their free cash flow.

The second contradiction is the obvious mismatch between short-term capital input and long-term capital returns. When the market no longer sets limits on its tolerance for future dreams and uncertainty, the story becomes hard to tell. Third, hyperscale cloud vendors are pouring huge sums into hoarding high-end computing power, creating excess capacity. They generally book it under an accounting standard of seven-year depreciation, but computing power iterates quickly, and two years later its value is very different. Fourth, the closed-source paid model is facing low-price competition from open-source models, forcing high-end large models to cut prices sharply, which hits long-term valuation levels. Finally, AI companies are disrupting themselves: the emergence of agents has sharply reduced search traffic, and other moats are also being eroded.

R&D and investment in AI were originally corporate actions, and their financial condition should be a credit risk. But the circular orders and cross-shareholdings that have become popular over the past two years have tangled the risks together, turning credit risk into financial risk and market risk. Moreover, most AI assets are illiquid, low in transparency, and difficult to mark to market. Once one company erupts in crisis, it may "trigger a fire that spreads through the whole camp"—exactly the lesson left to us by the subprime crisis.

The author does not believe this stock market correction is putting an end to the AI fever in the stock market, but the AI industry does need a reshuffle, and the market needs to refocus from "poetry and distant places" back to applications, profitability, cash flow, and debt repayment capacity. The author is bullish on the AI stock market, but expects the AI stock market to be stormy in the coming months.

The Fed has no reason to pivot abruptly for now

The FOMC passed a resolution by a 10-2 vote to keep the federal funds rate unchanged after the September hike, as more time is needed for observation.

Although there was no rate hike, the change in Fed Chair Warsh's assessment of the economic situation is visible to the naked eye. Two months ago at the Fed's annual meeting, he insisted there would be no wavering in fighting inflation; this time, however, he hinted that price pressures had improved slightly, and whether to raise rates in December would depend on the situation at that time. He again refused to provide forward guidance for December, reiterating that inflation had not yet returned to target, yet not telling the market what the next step would be. Traders believe Warsh's anti-inflation tone has softened somewhat, and the 2-year Treasury yield fell after the meeting, but the market's expectation of a December rate hike remains high, above 80%.

Long-end Treasury yields, however, rose rather than fell, indicating that what the Treasury market fears most is not a Fed rate hike but the debt burden. With US fiscal deficits remaining high, investors need a higher term premium to compensate for fiscal risk over the next decade. Short-term bonds are trading Warsh; long-term bonds are trading Washington. The two do not necessarily move in the same direction. The author maintains the previous expectation that the Fed will raise rates again in December and possibly once more in the first quarter of next year. Current economic growth remains resilient, inflation is above the policy target, and the Fed has no reason to pivot abruptly for now.

Next year's story may be divided into two acts. High interest rates, rising energy prices and tightening financing conditions will eventually slow the US economy; the tax-cut effect from early this year has largely ended, base effects from oil prices and rents will emerge, and inflation should fall step by step. At the same time, AI investment growth will peak in the first half of 2027, while AI applications have already created deflationary forces in the job market. The need for the Fed to keep rates high may then quickly dissipate, while funding liquidity pressure grows ever greater.

The author expects the Fed may begin a rate-cutting cycle in the second half of next year, and that cuts will then come faster than the market imagines. Over the next year the US rate path may resemble an inverted V: first hikes, then cuts; the policy rate at year-end may actually be lower than at the start of the year. Of course, major uncertainties exist in energy prices, AI investment and financial markets, so the policy rate path remains highly uncertain.

Washington and Wall Street are both preparing for a reshuffle

As the US midterm elections enter their final stage, Washington and Wall Street are both preparing for a reshuffle. The Democrats need a net gain of only four seats in the House to regain control, and likewise only four in the Senate. PolyMarket (international version) shows the probability of the Democrats taking the House has risen to 94%, and of taking both chambers to 68%. Although the Democrats raised only a quarter as much money as the Republicans, they succeeded in focusing the election debate on prices and corruption.

A Democratic sweep would be major political news, but the Democrats would also lack the ability to substantially reverse US policy. President Trump still sits in the White House, holds veto power and controls the executive branch, so it would be hard for the Democrats to advance major policy corrections on tax reform, health-care reform or climate legislation. The change would occur in the power structure. For the past two years the Republicans have controlled the White House and both chambers of Congress, and Trump has taken or left policies as he pleased, but once both chambers turn blue his role will shift from agenda-setter to defender.

The majority party in Congress also controls committee chairmanships, subpoena power, hearing power, budget power and other agenda tools; once the Democrats retake the House, many investigations currently suppressed are likely to be launched quickly. When the Democrats retook the House in 2018, the Trump White House faced numerous and escalating investigations, subpoenas and hearings that ultimately evolved into impeachment. If the Democrats take both chambers this time, they will inevitably restart checks on executive power. Impeachment of the president is initiated by the House, but final conviction requires support from more than two-thirds of the Senate. The Democrats currently cannot clear the threshold for successful impeachment, but an impeachment inquiry itself is enough to occupy the political agenda, consume White House resources and magnify every government misstep.

Once the space for domestic legislation is blocked, presidents often rely more on executive tools that do not require congressional approval, and trade policy is one of them. The author believes Trump may stake more political capital on tariffs and trade issues, with his focus more concentrated on less constrained overseas policy. After the Democrats control Congress, they may use appropriations, hearings and the legislative agenda to exert greater pressure on the Iran war, aid to Ukraine, Israel policy and military spending.

A blue wave could push the United States into a more intense period of institutional collision: the White House would rely on executive orders and presidential authority to advance policy, while Congress would rely on budget, investigation and legislative powers to check it. Government shutdowns, the debt ceiling and appropriations negotiations could again become routine in Washington. Whichever party wins, a fundamental improvement in the US fiscal deficit will be very difficult. Political gridlock is sometimes what financial markets like, because the government being "unable to do anything" is itself a form of policy certainty. Historically, with a Republican president and a Democratic Congress, the S&P 500's average performance in the 12 months after midterms has not been bad.

If a blue wave really occurs, what is most worth watching is not the rise or fall of Wall Street after the election, but the subsequent changes in power relations in Washington. Trump remains president, but no longer has a Congress that can smoothly carry out his will; the Democrats regain Congress, but still cannot dominate policy. US politics shifts from unilateral advance to comprehensive checks and balances, and the struggle over investigations, budgets and executive power again becomes the norm. Democrats and their supporters have never lacked anti-Trump sentiment, but they currently lack a leader who can turn that sentiment into a national political coalition.

[The author is President and Chief Economist of Springs Capital (Hong Kong). This article is purely personal opinion, does not represent the official position or forecasts of the institution, and is not investment advice or solicitation.]