The Zhitong Finance App learned that Wall Street financial giant Goldman Sachs recently released a research report saying that after experiencing the July AI deleveraging storm and massive sell-off in bullish positions, and after the global stock market was in a state of sideways trading and box shock for most of this summer, memory chip stocks and storage product lines related to the AI data center construction frenzy have begun to break through the investment target in a marked recent weak downward trend, moving towards a new round of upward bullish market trajectory.
The core opportunity that Goldman Sachs has captured is the expansion of demand for cutting-edge high-performance AI computing power brought about by the launch of OpenAI Astra and the RSI (recursive self-improvement) training paradigm beginning to dominate the expected rise in demand for memory chips in the context of AI training, and the low positions of traditional Wall Street asset management institutions and hedge funds.
According to a recent data compiled by Goldman Sachs, the net leverage of fundamental long and short funds is at the 4th percentile of the past year, which means that bullish exposure is close to the annual low; the implied volatility of the global semiconductor theme fell from 65 to 36, a drop of about 44.6%, indicating that previously extremely tight risk pricing has clearly eased. If the strong profit trajectory of memory chips driven by large-scale model upgrades and iterations and AI training sweeps the world by storm, continues to provide positive catalyst and institutions increase their risk exposure again, it is very likely that storage-related stocks will completely amplify their gains after breaking through consolidation.
The memory chip components of AI data center server clusters are still the clearest supply bottleneck at the AI computing power industry chain level. Market research agency TrendForce predicts that in 2026, server DRAM contract prices will increase by about 270%, and enterprise-grade SSD prices will increase by about 235%; HBM contract prices may still rise 70% to 140% in 2027. These data show the combined effects of AI computing power expansion and storage price increases. According to TrendForce's latest estimates, the combined share of DRAM and NAND in capital expenditure of major cloud service providers will rise from 47% in 2026 to 68% in 2027, behind which there is a simultaneous increase in procurement volume and price increases.
The Korean stock market has shown concrete signs of increasing capital participation. On September 7, Samsung Electronics rose 5.68%, and SK Hynix rose 8.26%. The Korean stock market benchmark index, the KOSPI Index, which has the title of “AI computing power weather vane,” rose sharply by 4.61% to 6,995.39 points; net purchases from foreign investors and institutions were about 2.55 trillion won and 2.64 trillion won respectively, and purchases have expanded beyond corporate buybacks. Based on 5,593.56 points on July 30, the index rebounded cumulatively by about 25.06%, which is in the technical bullish range. On September 8, the KOSPI index fell 0.58% to 6,954.52 points, still rising about 24.33% from this low, indicating that the storage market is strong, but the overall market is still driven by interest rates and energy risks.
Another Wall Street financial giant, Nomura Securities, recently stated that AI training and inference expansion are driving continued growth in demand, while supply expansion is limited, and the shortage is expected to continue until 2028. Furthermore, Nomura emphasized that long-term supply agreements (LTAs) will continue to improve the predictability of future profits of memory chip giants through volume locking, price protection, and advance payments.
Nomura expects that in the end, about 50% to 70% of sales will be covered by the Changxie Association, so the market continues to price traditional strong cycle stocks, about 3 times the expected price-earnings ratio of 2027, underestimating changes in the business model. This echoes Goldman Sachs's emphasis on low position filling opportunities: Nomura values the long-term profit base that supports the position replenishment market. Based on Samsung Electronics' closing price of about 269,500 won and SK Hynix's closing price of 1.793 million won on September 8, 2026, Wall Street financial giant Nomura's target price of 670,000 won for Samsung Electronics means a potential 150% increase over the next 12 months, and a potential increase of about 160% for SK Hynix's 4.7 million won.
Storage-themed stocks emerged from midsummer consolidation, and Goldman Sachs saw a breakthrough in positive signals
Lee Coppersmith, managing director of the fixed income, foreign exchange, commodities and stocks business division of Goldman Sachs Group, pointed out in the report that some of the most important investment targets in the storage industry chain — Micron Technology (MU.US), SanDisk (SNDK.US), iShares MSCI Korea Stock Market ETF (EWY.US), and Roundhill Memory Chip ETF (DRAM.US) showed similar bullish technology patterns. The latest trend chart compiled by Goldman Sachs shows that these targets are beginning to break out of the summer consolidation technical indicators, but related trends are still in the early to middle stages.
These potential breakthroughs come at a time when the broader AI computing power trading theme is ushering in a more favorable market environment: investors are bullish positions much lighter than during peak stock prices, implied volatility is declining, and a range of potential catalysts are on the horizon, including the Goldman Sachs Communacopia+ Technology Conference.
According to Goldman Sachs's main broker business data, the total leverage index of the US stock market's fundamental long and short hedge funds was only at the 27th percentile over the past year, while net leverage was only at the 4th percentile. It highlights that hedge fund institutional investors have begun to rebuild AI computing power positions, but the position level is still significantly lower than about two months ago.
The options market has also experienced a round of major adjustments. The implied volatility of semiconductors, measured by the Chicago Board Options Exchange Semiconductor ETF Volatility Index (VXSMH), has dropped from about 65 in July to 36, nearly halved, and fell back to the level close to the beginning of 2026.
Meanwhile, even after a massive rebound of nearly 7% last Friday, the stock market technology long and short momentum trading portfolio compiled by Goldman Sachs still retreated about 50% from its high in late June. Coppersmith said that this round of adjustments narrowed the potential performance range reflected in stock pricing related to AI computing power, but there was no clear market bearish trend.
The Korean market may also provide further momentum for the rise in the hashrate theme sentiment level. Goldman Sachs said that in the past four weeks, capital inflows to the Korean stock market were entirely supported by corporate repurchases. The rest of the investors' capital flows were in a state of net sales, but the sales rate narrowed significantly. This has left room for more investors to participate deeply in the Korean stock market, especially in the new bullish frenzy of SK Hynix and Samsung Electronics, the two largest memory chip stocks in the world.
Astra detonates the AGI fanatic+AI to participate in AI development, and memory chip revaluation ushered in a double demand expansion curve
The Astra model just launched by OpenAI continues to actively expand the professional tasks that AI can undertake. On Sunday, Nvidia CEO Huang Renxun made a big statement on social media that the launch of GPT-6 Astra means “AGI has arrived,” and Nvidia's confirmed strong revenue range and subsequent strong shipping guidelines, combined with the strong AI model development is entering a new stage of “recursive self-improvement (RSI)”, which opens up another stage of the AI computing power demand surge curve — that is, Astra is expected to expand AI computing power demand for commercial applications. Beginning the R&D trajectory of “building AI” may increase investment in cutting-edge operator experiments, evaluations, and long-term continuous training, and jointly extend the computing power investment cycle.
The investment significance of Astra and recursive self-improvement (RSI) is that cutting-edge high-performance AI models, and AI research and development itself are becoming new scenarios that continue to consume computing power. OpenAI revealed on September 6 that it has achieved the goal of being an “automated research intern” and can complete some tasks that would have required skilled researchers for several days under human guidance; as of mid-August, each human working day corresponds to about 3.1 smart device operation days. This measures run time; it is not a 3.1 times increase in scientific research output. It can be deduced from this that research automation will simultaneously increase the reasoning required for code generation, experimental evaluation, and candidate model training requirements. However, the complete RSI has not yet become an established dominant paradigm, and research direction and resource allocation are still determined by humans.
Astra represents the most advanced performance demand expansion mechanism: the ability of large models has been increased, making tasks that were previously difficult to complete reliably into the scope of commercialization. Furthermore, Astra may shift the overall demand curve — that is, when the AI model is smarter, companies can try jobs that could not be done reliably before, and competitors also need to continue to invest in R&D and training, which provides new strong support for the AI spending cycle.
The GPT-6 Astra model launched by OpenAI and the RSI technology path focused on by AI leaders are expected to become the two core driving forces driving the exponential expansion of AI computing power demand, namely the AI big model with better performance, the use of a wider range of AI application tools, and a next-generation AI training path with stronger computing power requirements, which are an important basis for the continued growth in AI computing power infrastructure demand.
OpenAI revealed that Astra achieved 98% in the FrontierMath Level 4 test and 99.9% in ARC-AGI-3. Based on this, Hwang In-hoon expressed the judgment that “AGI has arrived” and stated that the model used more than 100,000 Nvidia GPUs for training, and 400,000 GPUs will soon be launched. It is worth noting that “AGI has arrived” is still a controversial judgment, and the announcement of the deployment of the larger Nvidia AI GPU cluster directly reinforces the strong AI computing power demand expectations that the large front-end AI model will continue to expand investment in training resources.
From a technical perspective, storage requirements depend on parameter size, context length, number of concurrency, and experimental density. HBM is responsible for model weight, training intermediate state, and active key value caching on the GPU side; server DRAM undertakes data processing, operating environment and cache offloading; NAND enterprise-grade SSD stores data sets, training checkpoints, and reusable caches. An example given by Nvidia shows that the Llama 3 70B requires about 140GB of memory to load weights with FP16 accuracy, and the KV cache corresponding to the 128,000 token context for a single user still needs about 40GB. As stronger models handle longer tasks, more agents run simultaneously, and the RSI research process adds parallel experiments and checkpoint storage, the demand for capacity, bandwidth, and read/write throughput will expand dramatically at the same time.
KB Securities from South Korea predicts that storage will increase from 14% in 2025 to 40% in 2026, and then 57% in 2027. KB Securities' core bullish logic for SK Hynix and Samsung focuses on the gap between extremely thin inventory buffers and low forward profit estimates.
KB Securities said that Samsung Electronics and SK Hynix have been in storage inventory for less than 10 days, and it is estimated that AI infrastructure investment by hyperscale cloud vendors will reach 1.3 trillion US dollars in 2027, an increase of about 60% over the previous year. According to the stock price and profit forecast used in the report, the two companies retreated by about 38% from their previous high, corresponding to a price-earnings ratio of only about 3 times in 2027. Therefore, KB is betting on expanding demand and rising prices to push up profit expectations and drive valuation repairs. For KOSPI, the benchmark stock index for the Korean stock market, Goldman Sachs even set a target of 12,000 points. As of September 8, the KOSPI index fell 0.58% to 6,954.52 points; Goldman Sachs pointed out that the market has systematically underestimated the duration of the demand cycle for AI-driven memory chips and drastically raised the capital expenditure forecast of major US technology companies to 1.2 trillion US dollars next year, believing that the “storage shortage” caused by data center expansion will further intensify in 2027.