AI Investment Sustainability Put to the Test…For Chip Stocks, Focus Shifts from Earnings to Shareholder Returns

The KOSPI advanced strongly enough to trigger a buy-sidecar during the session, trading around the 6,560 level, up about 3.5%. The KOSDAQ also gained 2.9% to 748 points. Buying spread across the market, led by semiconductor stocks.

Foreign investors were net buyers of about 450 billion won on the main board and about 18 billion won on the KOSDAQ, and purchased more than 400 billion won in the futures market. Program trading showed net buying in both arbitrage and non-arbitrage transactions. The won-dollar exchange rate moved within the 1,324–1,325 range with little change.

Anthropic’s release of its registration statement and gains in the U.S. Nasdaq and semiconductor indexes the previous day were cited as immediate catalysts for the market’s rise. In Big Tech earnings, investors are expected to focus on cloud revenue growth, the pace of AI capital expenditure growth and free cash flow. With investment still needing to expand, a key measure will be how closely AI revenue and cash generation keep pace.

SK hynix was due to report earnings and hold a conference call the following day, while Samsung Electronics is scheduled to release its final results on the 29th. The market is watching not only the earnings figures themselves but also the companies’ shareholder-return plans and explanations of the profitability of AI investment.

Anthropic’s registration statement revealed a sharp rise in revenue alongside the scale of its losses. The disclosed materials put revenue for the third quarter of 2026 at $16.3 billion, while explaining that a substantial portion of the loss reflected accounting valuation losses resulting from an increase in the fair value of convertible bonds. This prompted the view that total losses alone do not provide a sufficient basis for assessing cash flow or the profitability of the business.

An analysis of the financial data in the filing found that the gross margin improved from negative 94% in 2024 to about 40% in 2025, and rose sharply from 12% in the first half of 2025 to 85% in the first half of 2026. Anthropic also held substantial cash and investments; one disclosed comparison found that its holdings exceeded Amazon’s. The simultaneous evidence of revenue growth, improving profitability and cash reserves eased some concerns about AI companies’ ability to raise capital.

The cost of securing computing infrastructure remains a risk. Anthropic typically contracts for infrastructure 1–1.8 years in advance, but if demand exceeds contracted capacity, it may have to use more expensive short-term infrastructure, putting pressure on margins. The company was reported to have agreed to pay about $3.1 billion for a contract to use SpaceX’s AI infrastructure in earnest from December 2026.

The filing also described concerns about timely HBM supplies from Samsung Electronics, SK hynix and Micron, as well as past chip procurement delays caused by capacity shortages at TSMC. It also mentioned the possibility of placing advance orders at above-market prices for power equipment with long lead times, such as transformers and turbine generators. This illustrates how AI demand could extend beyond semiconductors to power infrastructure and data-center equipment.

After SK hynix halted its review of a listing for Solidigm, some market observers linked the decision to potential measures to enhance shareholder value ahead of the earnings announcement. Kim Sun-woo, a researcher at Meritz Securities, raised the possibility that the company was preparing more specific and substantive shareholder-return measures. Bank of America also anticipated expanded shareholder returns from Samsung Electronics and SK hynix, and said the companies could address the issue in their earnings announcements.

Kim Dong-won, head of research at KB Securities, forecast additional shareholder returns of 80 trillion won for Samsung Electronics and 40 trillion won for SK hynix. Some market participants also suggested that SK hynix’s actual return could exceed this estimate, though nothing has been confirmed. For Samsung Electronics, investors are focused on total returns, including its existing dividend, and on how any additional dividends, share buybacks and cancellations would be allocated.

An analysis citing Goldman Sachs data put the forward P/E ratios of Samsung Electronics and SK hynix at about 3 and 2.8, respectively. These are low compared with the KOSPI’s average P/E of 10, and another analysis suggested that, based on a fair P/B ratio, current share prices are around half that level. Discussion continued over whether the low valuations indicate genuine undervaluation or reflect the market’s view that earnings growth is temporary—a question that depends on what the companies say and do.

Shareholder returns can raise dividend yields while also signaling that a company expects to remain profitable. The market is looking beyond the earnings figures to the scale and duration of return measures and to long-term plans combining dividends with share cancellations. The extent to which the companies can explain the durability of growth and downside risks in their earnings announcements and conference calls was identified as key to a re-rating of their shares.

Meta was reported to be pursuing a share issuance of up to $95 billion to expand its AI investment. The plan would initially raise $35 billion through common and preferred shares, with an at-the-market equity offering program of up to an additional $60 billion, depending on market conditions. An at-the-market offering issues shares over time in line with the share price and demand conditions.

Meta was reported to plan to use non-voting Class C shares. Issuing shares can reduce interest costs compared with borrowing, but it also dilutes existing shareholders’ ownership stakes. Reports also said Nvidia, AMD and Broadcom had expressed interest in investing in Meta.

Meta’s AI service Muse was described as limiting its launch to North America because of insufficient data-center capacity. A launch in Japan has been discussed, but agentic AI requires more computing power than existing services, making further infrastructure investment necessary. Whether Meta can turn the funds it raises into service expansion and monetization will determine whether it can offset the burden on shareholders.

SpaceX signed a $40 billion AI chip-related deal with Apollo Global. Under the arrangement, Apollo was reported to buy GPUs and related equipment and then lease them to SpaceX; the final interest rate and closing conditions have not been determined. Elon Musk argued that SpaceX’s funding base differs from that of typical neocloud companies because it generates independent revenue from Starlink and launch services.

Musk said the neocloud industry could see large-scale bankruptcies and consolidation, with few companies surviving. He also said SpaceX did not necessarily need to pursue acquisitions directly. His remarks prompted speculation that the company could be positioning itself to acquire data centers and GPU assets at lower prices after industry restructuring.

Neocloud companies use debt to acquire GPUs and lease data-center computing capacity to customers. Even if an operator runs into financial trouble, tangible assets such as GPUs and data centers remain, making the situation different from the bankruptcy of a typical company centered on intangible assets: an acquirer may be able to obtain those assets at a low price. There was speculation that financially strong companies, including SpaceX, could absorb assets as the industry is reshaped.

The AI industry is at a point where it is difficult to stop investing, but continued investment raises the question of where the funding will come from. Frontier-model companies such as Anthropic and OpenAI are growing both investment and revenue, but there may be a lag before revenue catches up with the funding required to secure data centers and chips. Investors are assessing companies’ cash reserves, access to external funding and free cash flow together.

Meta’s share issuance is a way to raise capital-market funding for AI investment, while Anthropic’s registration statement showed that revenue and cash holdings could support some of the investment burden. The market has entered a phase of testing not only whether AI leaders can continue investing, but also whether they can generate revenue and profits that exceed their investment costs. Differences in companies’ financing capabilities and the pace at which they monetize their businesses could widen gaps across the industry.

Whether concerns about neocloud bankruptcies signal a broader contraction in AI demand needs to be assessed separately. Calculations presented on air put CoreWeave’s share of the overall cloud market at about 3%, while CoreWeave’s disclosed 1.8 GW represented about 3.5% of roughly 40 GW of U.S. data-center capacity. Even if an individual operator encounters difficulties, hyperscalers and other companies could acquire its assets, potentially limiting the impact on the industry as a whole.

One forecast cited financing pressure at CoreWeave around 2027 and at Nebius around 2030. However, there are also a range of operators beyond neocloud companies, including Sorin, while some overseas operators have government shareholdings and therefore different funding structures. Discussion emphasized the importance of not equating the potential bankruptcy of an individual company with a slowdown in demand for AI infrastructure itself.

South Korea’s advance estimate of gross domestic product (GDP) showed growth of 1.0% quarter on quarter and 3.4% year on year in the third quarter. Semiconductor exports made a major contribution to growth. A broadcast forecast suggested that annual growth could reach the high-3% range, with 4% growth also possible if a recovery in consumption provides support.

Whether improving exports and large-company earnings translate into private consumption and household income remains a separate challenge. Private consumption declined in the third quarter, and although the Bank of Korea expects it to recover in the next quarter, the scale and durability of that recovery are uncertain. Whether strong semiconductor exports can spread to suppliers, workers and households more broadly will be a key factor in determining the quality of growth.

The current-account surplus was put at around $110 billion, described as the second-largest after China. Comparisons were also made with the combined surpluses of Germany and Japan, which the figure was said to exceed, while the current-account growth rate was also assessed as high. One view was that strong exports and the current-account surplus reduce the grounds for uniformly undervaluing the Korean economy.

Noh Geun-chang, CEO of Semicon Research Lab, viewed positively the improvement in revenue and profitability evident in Anthropic’s public disclosures. Revenue in the first half rose 14-fold year on year. The company remained loss-making under generally accepted accounting principles, but its adjusted operating margin was put at about 10%. The gross margin was about 85%, and Noh said profitability was improving as computing costs also declined.

Noh highlighted Anthropic’s position at the point where end consumers and corporate customers use frontier AI models. Semiconductor companies are further removed from end users, so investors may question the durability of their results even when earnings are strong. But he said evidence of revenue and profits at frontier-model companies helps validate AI demand. He added that the public filing made it possible to examine a debate that had until then relied on rumors and narratives against actual data.

Noh said growth continued after revenue slowed in August, supported by OpenAI’s Astra and other developments, and that Anthropic’s revenue was rising again in October. He also said combined monthly revenue at OpenAI and Anthropic continued to grow steadily. However, he said further examples were needed of companies that had adopted AI actually outpacing competitors and generating profits before investors’ doubts about monetization would be dispelled.

He assessed that business-to-business (B2B) demand was more likely than end consumers paying AI subscription fees to drive demand initially. Companies are still at the stage of experimenting with AI as a matter of course, and need time to change their workflows and organizations. Noh said that as more companies such as Palantir demonstrate actual gains in industry-specific productivity and competitiveness through AI, resistance to the costs of frontier models could ease and valuations across related companies, including semiconductor firms, could recover together.

Noh Geun-chang said supply constraints remain, making it premature to conclude that demand for AI semiconductors has weakened. Earnings announcements from Lam Research and ASML also indicated that equipment supply could be less smooth than expected, and bottlenecks in equipment are making it difficult to rapidly increase memory production capacity. Noh estimated that net annual additions of about 300,000 DRAM wafer starts would be needed to meet demand, and said this could exceed the combined expansion plans worldwide.

Noh said Samsung Electronics and SK hynix could differentiate themselves in premium products such as HBM and high-speed DRAM. Rising memory prices can increase revenue simply by raising the value of output, but expanding supply could lead to lower margins, making pricing and product mix important. He said HBM and high-speed DRAM command higher premiums than standard products and could lift overall profits if demand grows.

He said that because the market questions the durability of semiconductor companies’ earnings, companies need to explain their growth paths, supply constraints and responses to a slowdown in price increases in concrete terms during investor relations. They need to explain how they will maintain profitability if production does not rise as much as expected and what product strategies they will pursue even if memory prices fall over the longer term. Simply waiting for earnings to be reported before assessing the results and investing is not enough; he emphasized that IR should involve communicating forecasts and scenarios in advance.

Noh said shareholder returns can signal a company’s confidence and the durability of its profits. He explained that presenting specific return plans based on free cash flow and combining them with measures such as share cancellations could affect return on equity and a re-rating of the P/B ratio. He said Samsung Electronics and SK hynix were in the rare position of recording both low P/E and low P/B ratios, and that they should candidly explain to investors both the basis for their growth and unfavorable scenarios.

Noh Geun-chang said financing difficulties at individual neocloud companies such as CoreWeave do not automatically mean that the AI industry as a whole is in crisis. CoreWeave’s shares of revenue and U.S. data-center capacity were each put at around 3%. Hyperscalers are larger, and even if an operator runs into difficulty, another party could emerge to acquire its data-center and GPU assets.

Neocloud companies use borrowed money to buy GPUs and lease data-center computing capacity, leaving tangible assets in place even when financial pressures mount. Noh said CoreWeave’s and Nebius’s financing pressure could become an issue around 2027 and 2030, respectively, but cautioned against interpreting financial difficulties at a few companies as a collapse in overall demand. He added that the industry’s financing structures are not uniform, taking into account overseas operators with government shareholdings and other operators such as Naver.

Noh also suggested that the semiconductor cycle could prove more resilient than the market expects. Concerns about memory prices and supply volumes remain, but demand for HBM and high-speed products, as well as bottlenecks in equipment supply, could allow Korean companies to outperform. He urged shareholders of Samsung Electronics and SK hynix to be patient, saying their low valuations are difficult to match in past examples.

Even if semiconductor exports rise, the improvement may not translate into better economic conditions for most people if the gains remain concentrated among a small number of large companies. If companies return earnings to shareholders through dividends and share cancellations, investor income can rise and create a path toward stronger consumption. The argument is that a structure is needed to spread the benefits of growth to households and suppliers.

A comparison was made with the United States, where a cycle has operated in which corporate earnings flow into shareholder returns, investor income and increased consumption. Expanded shareholder returns by Samsung Electronics and SK hynix could also serve as a link between income and consumption, rather than merely supporting their share prices. To earn market confidence, companies need to present a long-term vision and execution plan alongside the scale of their returns.

When considering AI companies’ financing, the dilution of existing shareholders’ stakes from share issuance needs to be weighed against the need to secure capital to sustain investment. Meta’s planned capital increase burdens shareholder value, but some observers said it could also present an investment opportunity if the funds enable the company to expand AI services and generate profits. It was also argued that Korean semiconductor companies should look for opportunities to invest in overseas AI companies and infrastructure and expand their role in the global ecosystem.