Wall Street sets a positive tone for the long-term prospects of AI; the Fed's interest rate path for October remains uncertain! Cooling US Treasury yields open a window for tech sector rebound?

According to Zhitong Finance APP, during the midday session on October 16 Beijing time, Federal Reserve Chairman Kevin Walsh spoke with Kristalina Georgieva at the IMF and World Bank Annual Meeting in Bangkok, reiterating that the Fed will achieve price stability and described the rate hike in September as the removal of "some accommodation." Regarding the upcoming Fed FOMC monetary policy meeting set for October 27–28, he made it clear that he would not prejudge the path of the Fed’s monetary assets, nor would he pre-announce any actions that may be taken at the remaining meetings this year. He emphasized that every Fed decision combines output, employment, and inflation to form policy judgments, rather than following a pre-set policy path, and reiterated his personal preference that the Fed not provide any forward guidance.

Notably, Walsh's series of speeches at the IMF reinforced the judgment that cutting-edge AI technology is set to change the economy in the long run, but did not give any commitment to pause rate hikes in October. For the global semiconductor sector, which has recently suffered sharp sell-offs and entered a correction trajectory, the positive point was his lack of reinforcement of expectations for a near-term rate hike; however, whether U.S. Treasury yields will continue to cool, and whether battered AI compute-themed stocks, as well as the broader semiconductor and tech sectors, can stage a rebound will still depend on the simultaneous improvement of inflation, corporate financing conditions, and the realization of AI-driven profits.

Even more noteworthy than his "no prejudgment of monetary policy" stance is his statement that since the September rate-setting meeting, economic trends have generally been in line with expectations, with structural growth and a surge in investment currently present. This means that he has not equated the recent global tech stock correction, led by semiconductors, to a breakdown in the logic of economic growth.

At the same time, he emphasized that most parts of the world are synchronously tightening policies and that all nations need to maintain price stability; U.S. policy transmits overseas, and overseas policies also impact the U.S. in return. The policy framework remains to maintain inflation and institutional credibility, while retaining policy flexibility. Before his speech, MUFG's latest disclosure showed that the interest rate futures market priced in about a 27% probability of a rate hike in October, but fully accounted for a 25 basis point hike before year-end—thus, both before and after Walsh's speech, the interest rate futures market has been trading on the possibility that the timing of rate hikes may be pushed back, not that the tightening cycle has ended.

Recently, Wall Street giants have been concentrating their forecasts towards “continued rate hikes and an extended tightening cycle,” with the latest clear revisions from Citi and UBS. After the Fed's September hike of 25 basis points, the institutional focus is now on how much more may come. On October 14, Citi withdrew its forecast of “no further rate changes this year” and now expects a further 25 basis point hike in December followed by a pause, with rate cuts not expected until June 2027. UBS Global Wealth Management, between October 12 and 15, extended its projected path from one more hike in December to two more—one each in December this year and March next year, both by 25 basis points. Along with September’s move, the cumulative increase would be 75 basis points, with rates then maintained until around the end of 2027.

The key factor driving Wall Street’s reassessment of rate hike expectations is that the energy shock has started to spread to broader prices, while the economy is not yet weak enough to force the Fed to prioritize growth support. In September, core CPI rose by 0.3% month-on-month, exceeding the 0.2% market expectation for two consecutive months, and the year-on-year rate rose from 2.4% to 2.6%; the combined rise in housing, airfares, cars, clothing, and tech-related goods has made it more difficult for policymakers to view the inflation rebound as an isolated oil price disturbance. The following September PPI data showed a month-on-month rise of 1.3% and a year-on-year increase of 6.3%, further intensifying concerns about upstream cost transmission to end markets.

AI First Drives Demand, Then Unleashes Productivity: Why Does the AI Compute Boom Coexist with High Interest Rates?

In his speech, Walsh described AI as the most profound economic transformation of his life. The key to understanding this judgment is his emphasis that the impact of AI on the demand side will precede supply-side improvements. From the perspective of data center engineering and the AI compute supply chain, enterprises need to first purchase GPUs, TPUs, ASICs, HBM, network equipment, and invest in power, cooling, plants, and financing to build compute clusters; only after broad deployment of AI in R&D, production, and enterprise processes does the reduction in cost and increase in efficiency—which requires deployment, organizational adjustment, and business validation—materialize. This time lag allows AI to simultaneously boost semiconductor orders and increase capital and power demand, temporarily pushing up some prices and financing costs in the short term.

Therefore, Walsh’s AI optimism does not naturally align with lower rates. Productivity improvement helps expand supply and curb unit costs, but higher expected investment returns can also attract more firms to compete for capital, pushing up real interest rates. For the trajectory of 10- and 30-year U.S. Treasury yields, the market must also price in future policy rates, fiscal financing, and term premium. Strong compute demand and persistently high long-term capital costs can both stem from the same round of AI investment boom; concurrent semiconductor profit growth and valuation pressure can occur simultaneously.

His latest remarks also reveal how policy research is catching up with this transformation: there are plans to hold a live AI hearing in the Bay Area on November 16, with the Fed’s working group forming new research perspectives to aid FOMC judgment; AI may also improve the central bank’s data comprehension, though short-term employment effects remain uncertain. Regarding private tech investment, the Fed will not judge who is right or wrong but will focus on tail risks in finance that can be addressed via policy. These signs all suggest that the Fed and other global central banks will attach more importance to AI’s contribution to growth and will be examining financing structures and risk transmission but have not committed to protecting AI asset prices.

U.S. Treasuries Have Cooled, but Semiconductor Rebound Still Needs Cooling in Credit Costs and Profit Resonance

Long-term yields have indeed begun to fall. According to the U.S. Treasury, on October 15, the 10-year and 30-year Treasury yields were 5.10% and 5.43%, respectively, each down 9 basis points on the day; for the week up to the 15th, they were down 14 and 17 basis points, respectively, and compared to the end of September, down 19 and 21 basis points. However, both are still 35 and 18 basis points above their respective levels at the end of August, indicating that recent cooling has only partially reversed previous funding cost shocks.

Notably, a key driver of this pullback has been safe-haven Treasury buying driven by volatility in AI stocks and the credit market. TD Securities observes that investors are lowering growth expectations, and Treasuries have resumed their role as a hedge against equities. This creates an apparently contradictory market mix: compute stocks under pressure push Treasury yields downward, but the yield drop is not yet sufficient to dispel funding concerns for compute projects. Corporate funding costs are determined by both benchmark rates and credit spreads; if the former declines but the latter widens, data center projects reliant on external financing may still face high actual borrowing costs.

Global equities were also in a post-plunge recovery window as Walsh spoke. As of October 15, the Philadelphia Semiconductor Index fell 0.88%, recording a cumulative drop of about 12.4% over seven trading days, a 7.9% drop for the week, and about an 8.3% drop in October, though still up about 63.5% year-to-date, with short-term momentum clearly damaged. During Asian midday trading on October 16, Korea’s benchmark KOSPI index—seen as an “AI memory chip” bellwether—jumped 3.17% in a single day, with Samsung and SK hynix both rebounding. This rebound came before Walsh's speech but the rally continued thereafter.

Walsh's remarks have, to some extent, provided support for the recovery of sentiment in the battered global semiconductor sector, but what can truly drive a sustained semiconductor rally is the combination of “long-term yield declines + stable credit spreads + continued realization of compute orders and cash flows.” If financing pressures ease while AI demand remains strong, the market will once again be able to pay higher valuations for growth; within this, leading chip and equipment firms with ample cash flow, stable customer funding sources, and strong order conversion ability are poised to benefit first. The AI-driven long-term transformation emphasized by Walsh provides a macro explanation for this growth theme, while whether the short-term rebound becomes a sustained trend still depends on whether cost of capital relief can truly flow through to industrial investment.

Wall Street financial giant Goldman Sachs recently described three “main sources of support” (buybacks, retail, and foreign investors), which may improve the funding environment for AI assets to be repriced. The Goldman flow team noted in their latest report "Witching Hour" that, according to their estimates, about 48% of S&P 500 constituents will re-enter buyback windows around the end of October; historically, November has contributed about 10.1% of annual buyback execution. In addition, mutual funds' fiscal-year-end tax-related selling will subside, year-end retail inflows will arrive, and post-midterm demand for foreign allocation will increase—collectively, stock supply-demand may gradually improve. Of this, about $190 billion in foreign buying is a potential scale extrapolated from historical patterns. For semiconductors that have seen concentrated selling, renewed buying could mean that as long as companies prove orders and profits have not worsened significantly, it could spark position rebuilding. However, tech sector buyback execution has declined 40% year-on-year, reflecting the competition for capital between capex and shareholder returns; for AI stocks to continue attracting incremental allocation, it still needs to be demonstrated that investments in GPUs, networking infrastructure, optical interconnects for data centers, CPUs, and other key AI infrastructure resources can deliver higher future cash flows.