Global fund flow data compiled by LSEG Lipper, along with stock market trading activity, reveals a bifurcated equities landscape over the past two weeks—one defined by broad risk reduction alongside targeted opportunity-seeking.

Based on closing prices from August 28 to September 10, the KOSPI index—often dubbed the "AI computing power barometer"—rose approximately 3.61%, with heavyweights Samsung Electronics and SK Hynix climbing roughly 4.67% and 12.10%, respectively. The Philadelphia Semiconductor Index gained about 1.26% over the same stretch, though it first advanced 3.37% on September 4 before sliding 2.66% on September 10, signaling that the global AI computing theme and its supply chain remain under pressure from interest rate shocks even as capital continues to pour into Asian equity funds and technology sector funds.

Correspondingly, the latest global equity and bond fund subscription and redemption data from LSEG Lipper indicates that investors have significantly trimmed exposure to US large-cap equity funds while persisting with purchases of European and Asian stock funds as well as global technology sector funds. The frontier high-performance AI computing demand expansion brought by OpenAI's Astra, alongside the emergence of the RSI training paradigm dominating AI training workflows, keeps profit opportunities from surging AI compute demand in focus—yet market tolerance for valuations and financing costs is clearly diminishing.

Morgan Stanley, a Wall Street financial heavyweight, has weighed in on OpenAI's Astra model—which NVIDIA CEO Jensen Huang has described as ushering in the "AGI era"—highlighting that enhanced AI model capabilities are making more workloads economically viable, thereby tightening supply constraints across AI compute, data center power chains, substrate, and memory manufacturing. In its scenario projections, power capacity for hyperscaler compute deployments expands from approximately 35 gigawatts in 2025 to roughly 145 gigawatts by 2028, a multiple of about 4.1 times. GPT-6 Astra's launch on September 3, 2026 has undeniably intensified market debates about the arrival of the AGI era, driving stronger compute demand trajectories—particularly through the new "pay-per-outcome" growth model that promises even more robust total compute demand. More immediate evidence of compute demand comes from the AI R&D process itself: the "recursive self-improvement" trajectory where AI begins to "build AI."

As Astra's advanced frontier models generate increasingly potent AI compute demand, Morgan Stanley projects that capital expenditures for data centers at the four major North American hyperscale cloud and AI application vendors will rise from $917 billion in 2026 to $1.47 trillion in 2027 and $1.64 trillion in 2028, with deployment capacity expanding from 35 GW in 2025 to 145 GW in 2028. Astra's pivotal industry signal is that more complex work now holds commercial value when delegated to AI. On September 10, reports emerged that OpenAI launched a ChatGPT product tailored for the financial services sector, combining GPT-6 Astra with proprietary data sources to support research, financial modeling, and client material production. Extrapolating from this, AI demand growth variables will further expand to include concurrent agent counts, task execution durations, tool invocation frequencies, and context sizes: when the cost of completing a task drops and success rates climb, enterprises gain justification to deploy more workflows. This opens vast new space for cloud-based AI inference computing and high-performance storage demand related to AI, serving as the latest basis for the market to reassess the sustainability of AI infrastructure growth.

US Equity Funds Bleed Heavily While Tech Funds Attract Inflows

With global capital shrinking exposure and selectively focusing on specific tracks, the latest fund subscription and redemption data reveals that investors have dramatically reduced holdings in US large-cap equity-linked stock funds while maintaining purchases of European and Asian equity funds and global technology sector funds. Value stocks, high free cash flow stocks, and bonds have not formed a broadly rising safe haven either. During the same period, using adjusted closing prices of corresponding index ETFs as the observation gauge, the IWD tracking the Russell 1000 Value index fell roughly 1.94%, the COWZ focused on high free cash flow yield stocks dropped about 3.79%, the AGG tracking the Bloomberg US Aggregate Bond index declined approximately 1.13%, and the SPY tracking the S&P 500 index fell roughly 1.50% over the same period.

These data points underscore that high cash flow can improve a company's ability to cope with financing pressures but cannot eliminate risks stemming from excessively strong industry cyclicality, positioning structures, and valuation downside. Bond funds receiving net subscriptions also does not imply that bond prices are simultaneously rising. The clearer allocation shift at present is that capital is becoming more selective and tending to control the overall interest rate sensitivity of stock-bond portfolios. LSEG Lipper data shows that for the week ending September 9, global equity funds saw net outflows of $15.52 billion, the largest weekly outflow since March 18. US equity funds experienced net outflows of $32.27 billion, unexpectedly marking the highest level since the week of December 17, 2025, when net outflows hit $52.45 billion. European and Asian equity funds, however, attracted net inflows of $11.16 billion and $3.03 billion, respectively. The scale of US market withdrawals exceeds global equity fund net outflows, indicating that net buying in other regions provided partial offset. This reflects regional allocation divergence and cannot be directly inferred as the same group of investors redirecting all US asset proceeds to Europe or Asia; fund subscription and redemption data also does not equate to cross-border capital flows across the entire stock market.

Within the US market, withdrawals are concentrated primarily in large-cap funds: a single-week net outflow of $40.44 billion, a record, while mid-cap funds saw net outflows of $682 million, and multi-cap and small-cap funds recorded net inflows of $3.52 billion and $274 million, respectively. Sector selection runs concurrent with overall withdrawals: global sector funds saw net inflows of $2.92 billion, with technology and financials attracting $1.89 billion and $1.25 billion, respectively; US sector funds recorded net inflows of $1.46 billion, with technology seeing $1.71 billion in net inflows and financials $720 million. The combined net buying in technology and financials exceeds the overall net inflow of sector funds, largely implying offsetting large-scale one-way withdrawals in other sectors. Capital is thus exhibiting a pattern of trimming broad equity exposure while preserving specific sector opportunities; notably, regional, sector, and market-cap classifications belong to different LSEG Lipper statistical dimensions and cannot simply be repeatedly added or subtracted.

Inflationary Pressures Reshape Capital Destinations: Short-Duration Bonds Build Defenses, Energy Gains Allocations

The macro pressure driving this adjustment remains the energy supply shock and its interest rate consequences: Brent crude broke above $100 per barrel on September 9 and touched $109.97 intraday on September 11; WTI surpassed the July high of $93.50 during the week and reached $104.46 on September 11, with both hitting four-month highs. US PPI and CPI data released on September 10, combined with earlier price rises in Europe and Japan driven by energy inflation, further reinforced inflation stickiness and rate hike concerns. The most distinctive shift in bond allocation is investors' greater willingness to take on shorter-duration risk.

Global bond funds saw net inflows of $8.95 billion that week, the lowest weekly inflow since July 29, with short-term bond funds attracting $6.65 billion, marking the second-largest weekly inflow in three months. Loan participation funds and government bond funds recorded net inflows of $1.01 billion and $743 million, respectively, while corporate bond funds saw net outflows of $2.37 billion. US bond funds logged their 21st consecutive week of net buying, with $6.56 billion in inflows that week; short-to-intermediate investment-grade bond funds attracted $3.75 billion, the highest in nine weeks, and short-to-intermediate government bond and US Treasury funds drew $2.78 billion. Overall, investors still require bond coupon income but place greater emphasis on duration control and credit quality; global corporate bond fund outflows and US short-to-intermediate investment-grade fund inflows also reflect differences at the regional and tenure levels.

Money market funds further reveal that "risk aversion" is not a globally synchronized one-way trade. Global money market funds recorded net inflows of $10.72 billion, marking a second consecutive week of net buying; yet US money market funds saw net outflows of $10.41 billion, following net inflows of approximately $48.76 billion the prior week. Thus, large redemptions from US equity funds did not mechanically convert into net subscriptions for US money market funds in the same week. Redemption pacing, capital usage, and regional distribution can all influence outcomes; existing data is insufficient to track the final destination of every redeemed dollar, but it can confirm that rising global liquidity preference and phased outflows from US cash funds can occur simultaneously. Commodity and emerging market capital distribution also demonstrates selectivity. Global gold and precious metals funds ended eight consecutive weeks of net inflows, recording net outflows of $537 million that week, while energy funds attracted $211 million. Energy receiving additional allocations is consistent with investors focusing on supply shocks and their potential earnings beneficiaries, but this cannot be taken to mean that precious metals capital was directly redirected into energy. Data covering 28,984 funds also shows that emerging market equity funds ended eight consecutive weeks of net buying, turning to net outflows of $1.56 billion, while emerging market bond funds continued their sixth consecutive week of net inflows, attracting $537 million that week.