Semiconductor Sell-Off Spurs Rotation to Bitcoin, Gold as ETF Flows Shift

Money rarely disappears from markets—it rotates. After months in which AI-driven semiconductor stocks vacuumed up capital on the back of data-center spending and high-bandwidth memory (HBM) constraints, that trade is starting to unwind, and investors are already scouting the next destinations: gold, Bitcoin (BTC), and a widening slice of payment- and fintech-led financial stocks.

The shift matters because the semiconductor rally had become the defining expression of the AI boom. As that pillar wobbles, the question for global markets is less about whether AI demand is real—and more about how quickly crowded positioning and high valuations can push capital into new narratives.

Bloomberg data cited in the report points to a striking symmetry in flows. Year-to-date, roughly $17 billion has been withdrawn on a net basis from U.S.-listed gold and Bitcoin ETFs, while a near-matching amount has moved into U.S. semiconductor ETFs. The alignment does not prove a one-for-one transfer—investors, mandates, and time horizons vary—but it does underline how decisively ‘risk appetite’ pivoted toward AI hardware earlier in the year, and how that preference could now reverse as semiconductor prices correct.

Semiconductors are no longer facing a routine pullback confined to a handful of mega-caps. The Philadelphia Semiconductor Index has fallen about 20% from its recent peak, a decline that typically places an index in bear-market territory. The weakness has spread broadly across the supply chain—from Nvidia ($NVDA) and Micron ($MU) in the U.S. to Samsung Electronics (005930.KS), SK Hynix (000660.KS), Taiwan Semiconductor Manufacturing Company (TSM) and European chip-equipment makers.

At the same time, several non-AI sectors—including parts of financials, healthcare and industrials—have held up comparatively well, reinforcing the idea that this is less a market-wide breakdown and more a rotation away from a single, crowded theme. The irony is that fundamentals across much of the chip complex have not collapsed overnight. Orders and revenue commentary remain constructive for many companies. But markets price expectations, not just текущие earnings, and when ‘growth optimism’ becomes fully embedded in valuations, even good results can fail to support prices.

That dynamic helps explain why cross-asset flows are now under scrutiny. Earlier in the year, gold and Bitcoin looked stagnant next to the relentless march of AI-linked equities. Investors who prioritised momentum had little patience for assets that were not making new highs. Now the script has flipped: chip stocks have stumbled, and investors nursing losses are scanning for assets that feel less expensive—or at least less crowded.

Gold and Bitcoin can rise for similar macro reasons, but they are not interchangeable. Both are often framed as alternatives to traditional fiat-linked assets, and both are now easily accessed via U.S.-listed ETFs. Yet gold remains the classic shelter in periods of geopolitical stress and economic anxiety. Bitcoin, despite its long-run ‘digital gold’ branding, still behaves like a high-volatility risk asset over shorter horizons and has often traded in sympathy with tech equities. Some research has suggested that Bitcoin’s correlation with equity markets strengthened after the launch of spot ETFs, complicating the idea that BTC automatically benefits when investors flee stock risk.

In practical terms, a deeper semiconductor drawdown does not guarantee that both gold and Bitcoin rise in tandem. If fear dominates—driven by geopolitics, recession risks, or falling real rates that primarily support defensive positioning—gold may attract flows first. If liquidity expectations improve and risk-taking returns, Bitcoin may respond more aggressively. The more accurate framing is that Bitcoin can become an alternative ‘liquidity asset’ for capital previously concentrated in speculative growth trades, while gold remains a more traditional endpoint for capital seeking protection.

A third destination is emerging in U.S. sector flows: financials, particularly the parts of the sector tied to payments, data, and market infrastructure. The Bloomberg figures referenced in the report show financial-sector ETFs absorbing the largest net inflows among major U.S. sectors over the past month, while technology funds saw notable outflows. Healthcare, utilities, and industrials followed financials in relative strength.

There is also a structural logic behind that move. AI hardware is ‘capital intensive’—chip fabs and data centers require enormous up-front investment in land, power, cooling, servers, and network equipment. Growth often depends on continuously expanding capex. By contrast, many payments and software-driven fintech models can scale transaction volume and customer activity with a lighter balance-sheet footprint, monetizing ‘network’ effects and data rather than physical plant.

Within financials, the leaders have not been limited to traditional banks. Payment companies such as PayPal ($PYPL) and Global Payments ($GPN), financial-information and credit-rating firms such as S&P Global ($SPGI) and Moody’s ($MCO), and market-infrastructure names like Nasdaq ($NDAQ) have shown notable relative strength. PayPal’s sharp move, in particular, was influenced by reports that Stripe and private equity investors have explored an acquisition valued around $53 billion—an idiosyncratic catalyst that should not be attributed solely to sector rotation. Still, the willingness of strategic and financial buyers to discuss deals at that scale underscores renewed attention on the strategic value of global payment rails and consumer transaction data.

Even here, investors are parsing nuance. ‘Financials’ is not a monolith: banks and brokerages are far more exposed to interest rates, credit losses, and capital-markets activity than payments and data businesses. A sharp equity sell-off can cool IPOs and M&A, pressure investment-banking fees, and tighten credit conditions—risks that may not apply in the same way to transaction- and data-fee models. The rotation, in that sense, is less “sell tech, buy banks” and more a shift from heavy-capex growth stories to companies that collect tolls on financial activity with comparatively lower incremental capital requirements.

Looking ahead, one variable looms particularly large for the next leg in semiconductors: rates. The report notes that the U.S. 10-year Treasury yield has recently traded above 4.5%, while the 30-year fixed mortgage rate is around 6.55%, near its highest level in roughly a year. A cooling housing market adds to the caution—pending home sales in June fell 5.4% from the prior month, according to the figures cited.

Higher rates matter for AI hardware more than the narrative often admits. Data centers are built in the real economy, financed through borrowing and capital markets. When funding costs rise and credit conditions tighten, investors tend to discount distant cash flows more aggressively, and companies face steeper hurdles for incremental investment. However compelling AI demand may look, it sits alongside constraints imposed by financing, electricity, and property markets—areas where the bond market effectively sets the price of ambition.

The implications could be especially direct for South Korea, where the equity index is heavily influenced by Samsung Electronics (005930.KS) and SK Hynix (000660.KS). When semiconductors lead, they can lift the entire market; when they fall, strength elsewhere may not be enough to defend headline indices. Concentration risks also extend beyond one country. The report highlights claims that TSMC (TSM), Samsung Electronics (005930.KS), and SK Hynix (000660.KS) together represent roughly 29% of the MSCI Emerging Markets Index—an unusually high dependence on a narrow slice of the AI supply chain.

For Korean retail investors, the rotation story intersects with another reality: crypto’s accessibility. Bitcoin can be traded around the clock and purchased directly in KRW on domestic exchanges, creating a fast channel for risk capital exiting equities to reappear in digital assets. But that mechanism works in both directions. In a global risk-off shock, investors may sell Bitcoin to cover equity losses rather than buy it as a haven. Both semiconductors and Bitcoin are highly sensitive to leverage and liquidity, which can amplify moves during stressed conditions.

Three indicators stand out for monitoring whether the rotation is sustainable. First, ETF flows: whether semiconductor ETF inflows stall and turn into net outflows, and whether gold and Bitcoin ETFs begin to stabilise or flip back to net subscriptions. Second, rates: if the 10-year yield continues to climb, pressure on high-valuation growth segments is likely to persist; if yields stabilise and liquidity expectations improve, Bitcoin and growth equities can rebound together. Third, internal differentiation within financials: sustained leadership from payments, fintech, and financial-data firms—alongside weaker performance from banks and capital-markets-exposed institutions—would suggest the market is choosing ‘capital-light financial infrastructure’ over traditional cyclicals.

Ultimately, the semiconductor correction does not signal the end of AI. Data centers are still being built, and demand for advanced chips and HBM has not evaporated. What may be ending is the period in which semiconductors monopolized market attention and absorbed an outsized share of incremental capital. As positioning unwinds, money will seek new stories—some defensive, some speculative, some focused on monetizing AI through services rather than hardware.

Markets often move before narratives catch up. The more consequential question now is not only why semiconductors are falling, but where the capital leaving the trade is going—and whether those inflows become the next durable leadership signals across gold, Bitcoin, and the evolving fintech and payments complex.