The US stock market’s “hidden bear market” harbors major opportunities? Goldman Sachs predicts a “three-way buying” is poised for a comeback, JPMorgan releases a bullish report on storage tokens

According to FinanceAPP, the well-known Wall Street financial research institution The Kobeissi Letter has just released a new report showing that while benchmark US stock market indices remain at high levels, valuations have already “gone into winter.” This institution is offering a contrarian bullish perspective just as the Nasdaq and the Philadelphia Semiconductor Index have seen significant declines. For the US equities bull market driven by the AI computing power theme in recent years, this is a heavy “shot of adrenaline.”

Kobeissi focuses on the fact that index resilience masks a significant correction in individual stocks: by its statistics, 84% of S&P 500 components are down at least 10% from their respective all-time highs, and 59% are down at least 20%. Meanwhile, some tech giants' profit growth continues to support fundamentals. Goldman Sachs’ fund flow team provides another bullish clue: the reopening of the buyback window, year-end retail inflows, and post-election foreign buying may bring incremental demand to stocks that have suffered valuation compression. Together, these point to high-quality technology growth assets that can continually deliver strong earnings, likely benefiting simultaneously from capital inflows and valuation recovery.

At the same time, for the South Korean stock market—which has recently experienced sharp sell-offs and plunges—Wall Street giant JPMorgan has adopted a contrarian, “buy the dip” approach, highlighting that macro-level concerns over the sustainability of AI investment are “clearly out of sync” with the bottom-up supply and demand data from the storage sector. Thus, the bank considers Q3 to be the “expectation reset season.” It is therefore optimistic about a major rebound in the South Korean market, where SK hynix and Samsung, the two memory chip leaders, account for 50% weight and provide a “barometer for memory chips.” This also largely means that, from JPMorgan analysts’ perspective, the memory chip sector, one of the main AI computing power beneficiaries in recent years, may return to the bull market trajectory after its recent sharp correction.

Latest price moves are already showing both severe corrections and some bottom-fishing. As of the close on October 15, the Nasdaq 100 dropped 0.05%, down 3.07% for the week and 1.56% for October. The Philadelphia Semiconductor Index fell 0.88%, marking seven consecutive down sessions and a 12.38% drop, pushing it into a technical correction; it’s down 7.88% for the week and 8.29% for October, but up about 63.5% year-to-date. However, the latest data shows that CoreWeave, known as the “canary of AI trading,” rebounded 5.94% after plunging 15.46% the previous day, though it is still down 15.10% for the week. These signs all indicate that sharp corrections and bottom fishing are already occurring, with the market now differentiating between industrial growth rates and financing capacity.

Kobeissi Letter proclaims: The current “stealth bear market” is “one of today’s greatest market opportunities”

The S&P 500 is currently only about 2% below its all-time high, but analysts at The Kobeissi Letter believe that the market is simultaneously experiencing a “stealth bear market,” representing “one of today’s greatest market opportunities.”

Kobeissi’s undervaluation call mainly comes from multiple compression rather than just the recent correction—the report notes that Nvidia’s stock price rose 21% over the past 12 months, yet its P/E ratio declined 46%. Its current forward P/E is about 24x, close to the S&P 500’s 22x. In the institute’s view, energy inflation, US Treasury yields, and rate hike expectations have depressed the market’s willingness to pay for valuations. If these pressures ease and earnings keep coming, tech stocks led by the Magnificent Seven may see strong valuation recovery.

This financial research institution states that currently 84% of S&P 500 constituents have fallen at least 10% from their historic highs, forming what it calls a historic “bull and bear coexisting” paradox in the market.

The macroeconomic backdrop has changed significantly since early 2026—at that time, the market expected the Fed to cut rates at least three times totaling 75 basis points.

Senior analysts at the institution wrote in their research report: “Today, the Fed has hiked rates once this year, and the rate futures market seems to fully price in three more hikes by June 2027; including the September hike, the cumulative increase will be 100 basis points.” They noted, this means a 175 basis point shift in rate expectations.

According to the institution, what drove this change was a historic surge in oil prices: diesel prices rose more than 80% from the lows in January 2026, and Brent crude is now trading at $106 per barrel.

Market concentration has reached extreme levels. Since the March 30 market bottom, only 10 stocks have contributed 70% of the S&P 500’s gains. Nvidia contributed the most with 13% of the total index gain, followed by Micron, Apple, and Microsoft, each contributing 9%.

The research organization said: “Meanwhile, the other 490 stocks only contributed 30% of the S&P 500’s gain since March 30.” Real estate, materials, utilities, and consumer discretionary sectors saw the greatest declines.

Despite the index being close to record highs, The Kobeissi Letter believes that some tech companies whose fundamentals far outpace market benchmarks are currently trading at significant and unreasonable discounts.

Discussing Nvidia, analysts stated: “Even though earnings growth during this period has been incredibly strong and upwardly revised by analysts, pushing the stock up 21% and solidifying market cap above the $5 trillion milestone, the P/E ratio has still dropped a sharp 46%.” Nvidia’s current forward P/E is approximately 24x, nearly equal to the S&P 500’s 22x.

The institution states, the forward P/E of the S&P 500 Information Technology sector and the S&P 500 index as a whole are now nearly the same—this is very rare, as tech stocks historically have “almost always enjoyed a notable valuation premium vs. the general market.”

The Kobeissi Letter believes that this valuation compression has formed a “stealth bear market,” weighing down overall valuations of the world’s largest tech giants.

The senior analyst team at the institution wrote: “If this backdrop is ultimately no longer priced in by the market, it could catalyze another powerful rally for the S&P 500, where tech giants comprise roughly 40% of the weight.” They add: “Even a modest broadening of the rally would rapidly increase buying demand.”

Buying, Valuation, and Earnings Begin to Converge: What Support Does the AI Computing Power Rebound Need?

Compared to The Kobeissi Letter, the latest research reports from Goldman Sachs and JPMorgan not only support the bullish call that “AI computing and semiconductors have strong conditions for a rebound,” but also provide crucial evidence on the funding and earnings side.

Goldman’s so-called “three reinforcements” (buybacks, retail investors, and foreign investors) are expected to improve the capital environment for AI assets to be re-rated. Goldman’s fund flow team pointed out in the latest “Witching Hour” report that by Witching Hour’s estimate, about 48% of S&P 500 components will be back in buyback open windows by the end of October. Historically, November accounts for about 10.1% of all annual buybacks. On top of that, mutual funds’ year-end tax selling pressure fades, year-end retail inflows and overseas allocations following midterm elections—together, these could gradually improve stock supply-demand dynamics. Among this, around $190 billion of foreign buying is projected by historical extrapolation. For semiconductors that have seen concentrated de-risking, buying resumption means even just proving orders and earnings haven’t deteriorated much could trigger position replenishment. However, the tech sector’s buyback execution is down 40% year-on-year, indicating that capex is competing with shareholder returns for capital. To attract sustained allocation to AI stocks, they must prove that investment in GPUs, network infrastructure, data center optical interconnect, CPUs, and other key AI infrastructure resources can produce significantly higher future cash flow.

Goldman’s bullish rationale first comes from a marginal improvement in supply-demand. The fund flow team estimates that by the end of October, about 48% of the S&P’s components will be back in buyback windows. Historically, November’s buyback execution is about 10.1% of the annual total, second only to August. Meanwhile, tax-driven selling by some funds around their October fiscal year end is expected to subside, and retail inflows in November–December, plus foreign allocation post-midterm elections, may combine to strengthen market absorption capacity.

For AI semiconductor assets that previously saw concentrated selling, Goldman’s analyst team stresses that the new buying doesn’t need to be huge—it may be enough to propel a significant recovery in prices. However, this support first applies to the broader market. Where the capital flows on a large scale ultimately depends on which companies can show new profit growth evidence in this reporting season, and AI computing-related semiconductor stocks and cloud computing leaders may have an edge.

JPMorgan, on the other hand, interprets why the memory correction may be creating opportunities from the strength of industry-side profits: Market concerns over continuous AI capex spending may not have equally translated into deteriorating storage orders and earnings. AI inference requires constantly reading model weights, saving context and KV cache, and relies on server memory and SSDs for different data needs. HBM upgrades also occupy wafer, packaging, and yield ramp-up resources, making effective supply lag behind demand growth.

Meta’s launch of Muse Spark Open can be seen as a new technical driver for wider application and inference demand growth. With Meta Muse and OpenAI’s GPT-6 Astra fully igniting the AI agent wave—and the latest developments featuring Muse Spark Open, “the strongest open-weight AI foundation model in the US to date,”—“AI FOMO trading” (where investors rush to buy or rebuy previously reduced AI assets, fearing to miss out on gains) is likely to return in the near term.

From an AI infrastructure architecture perspective, Muse, Astra, and Meta’s latest open-sourced Muse Spark Open are expected to jointly further expand AI agent usage. Their massive AI inference workload expansion will simultaneously increase model inference, tool running, and state management needs, deepening competitive pressures on some software companies. GPUs, TPUs, and other accelerators handle model computation; high-performance server CPUs run browsers, virtual machines, product search, database queries, and trade orchestration; HBM and server DRAM handle model data, context, and concurrent work environments; enterprise SSDs store product indexes, task logs, and persistent states; and high-speed networks and optical interconnects support distributed data exchange.

Therefore, JPMorgan analysts state, trading ahead of a cyclical cooling in memory chip prices can occur in parallel with a prolonged profitability supercycle in the global AI computing value chain, including memory chip components. Led by Jay Kwon, JPMorgan’s Asia-Pacific analyst team is optimistic on Samsung Electronics, SK hynix, Kioxia, Samsung Electro-Mechanics, ISU, and ISC over the next six months, betting on a longer memory upcycle and sustained demand for substrates, PCBs, and testing.

JPMorgan’s semiconductor research report on South Korea shows that macro concern over AI investment sustainability is clearly out of sync with the bottom-up supply and demand data from the storage sector. Thus, Q3 is seen as the “expectation reset season,” and the next six months will overweight Samsung Electronics, SK hynix, Kioxia—and extend to Samsung Electro-Mechanics, as well as ISU and ISC, the industry leaders in AI PCBs and test sockets, respectively. This “duration trade” bets on a longer upcycle and extended high profitability: even if DRAM/NAND price increases slow, as long as HBM, server DRAM, and enterprise SSD demand continues to absorb new capacity and achieve significant growth, the market’s previous expectations of profits falling rapidly may need to be revised. This also explains why memory stocks can fall ahead of profit peaks, but rebound early once earnings sustainability is verified. The trading opportunity is decided by the ongoing gap between real earnings trends and the profit expectations priced into the stock.

According to Kobeissi’s contrarian bullish logic, Nvidia’s stock rose 21% over the past 12 months and its P/E fell 46%; its current forward P/E ratio is about 24x, close to the S&P 500’s 22x, reflecting a significant shrinkage in the tech premium. Latest AI computing supply chain data show that lithography machine giant ASML (ASML.US) still expects that according to the current supply plan, it cannot fully meet client demand for 2027–2028, supplying more industry evidence for the sustainability of AI hardware investment. That’s why, in Kobeissi’s view, the attractiveness of contrarian positioning in AI computing technology and the US stock market is concentrating on those companies with deliverable orders, realizable earnings, and cash flow strong enough to support expansion. If financial reports maintain profit expectations and energy inflation and financing pressure ease, returning capital could prompt valuation recovery on previously compressed stocks. The truly promising rebound will be when earnings resilience, improved financing conditions, and expanding buying power join forces.