Key Takeaways
- Demand for AI‑focused semiconductors remains robust, with TSMC and ASML reporting record‑high profits and strong order backlogs.
- TSMC’s Q3 net profit jumped 68.7% year‑on‑year to $763.1 billion, marking six consecutive quarters of all‑time highs.
- ASML posted €11.7 billion revenue and €4.96 billion net profit in Q3, and plans to lift capacity by 30% in 2027 and again in 2028, yet says this will still lag demand.
- CoreWeave, an Nvidia‑backed AI cloud provider, is carrying roughly $50 billion of debt to fund its data‑center build‑out, raising concerns about interest‑rate sensitivity.
- Investors are shifting focus from pure AI hype to tangible metrics: cash flow, debt levels, and profitability.
- Chinese memory maker Changsin Memory Technology (CXMT) is advancing new memory tech, adding uncertainty to the long‑term profitability of AI‑driven semiconductor demand.
- The market is now splitting AI‑related stocks into two camps: firms with strong cash flows (TSMC, ASML) versus those reliant on heavy borrowing (CoreWeave and similar “NeoCloud” players).
- While the AI boom continues, the prevailing question has become “how much you earn and how much you owe” rather than simply “Is it an AI company?”
Core Demand for AI Semiconductors Remains Strong
Despite recent market turbulence, the underlying appetite for AI‑centric chips shows no sign of waning. Analysts note that “Demand for AI semiconductors is still strong and some key companies are performing at record‑high performances,” a sentiment echoed by the steady stream of orders flowing to foundries and equipment makers. This persistence is driven by the relentless expansion of generative‑AI models, large‑language‑model training, and the proliferation of AI‑enabled data‑center workloads that require ever‑more advanced logic and memory solutions.
TSMC’s Record‑Breaking Profit
Taiwan Semiconductor Manufacturing Company (TSMC) delivered a striking performance in the third quarter, announcing a net profit of $763.1 billion, which represents a 68.7% increase year‑on‑year and marks the highest net profit ever for six consecutive quarters. TSMC’s management also forecast a 44% sales rise for the full year versus 2023 and reaffirmed its capital‑expenditure budget of $60 billion–$64 billion to sustain capacity for AI and high‑performance computing chips. The company’s commentary underscored that “Demand for advanced semiconductors for AI and high‑performance computing is still boosting performance,” reinforcing its role as the linchpin of the AI supply chain.
ASML’s Steady Orders and Capacity Plans
Dutch lithography leader ASML mirrored the strength seen at TSMC, reporting €11.7 billion in revenue and €4.96 billion in net profit for Q3. The firm highlighted that its order book remains “solid,” prompting it to target a 30% capacity increase in 2027 and another 30% in 2028. Yet ASML cautioned that even these aggressive expansions “will not be enough to keep up with demand,” signalling that the bottleneck in extreme‑ultraviolet (EUV) lithography could persist for years. The company’s confidence helped “somewhat allay concerns that investment in AI semiconductors is sharply declining,” as investors weighed the durability of equipment spending against short‑term market jitters.
CoreWeave’s Debt‑Driven Expansion Raises Alarm
While foundries and equipment makers enjoy robust balance sheets, the picture darkens at the cloud‑infrastructure layer. CoreWeave, an AI‑focused cloud provider backed by Nvidia, has amassed approximately $50 billion of debt to finance a rapid rollout of GPU‑laden data centers. Reuters highlighted investor worries that “Coreweave’s debt is worth about $50 billion and that the data center construction boom is largely dependent on credit.” The concern intensifies should U.S. interest rates stay elevated, as the interest burden on such leveraged builds could swell quickly, eroding profitability and testing the sustainability of the “NeoCloud” model that leases GPUs to AI developers.
Shifting Investor Priorities: Cash Flow Over Hype
The broader equity market is recalibrating its AI thesis. Rather than rewarding companies merely for bearing the AI label, investors now scrutinize real cash generation, debt service capacity, and operating margins. This shift is evident in the postponement or withdrawal of several AI‑related IPOs that had been rushed to market during the fervor of 2023‑2024. As one market observer put it, the focus has moved from “Is it an AI company?” to “how much you earn and how much you owe.” Firms that can demonstrate strong free cash flow while servicing sizable capital expenditures are being rewarded, whereas those that rely heavily on financing to chase growth are seeing their valuations compress.
China’s Semiconductor Push Adds Uncertainty
An additional layer of complexity emerges from Asia, where Chinese memory maker Changsin Memory Technology (CXMT) has reported progress in next‑generation memory technology. The article notes that this development makes it “difficult to guarantee whether memory companies’ high profitability will continue even if AI demand increases if Chinese companies’ pursuit accelerates.” Should CXMT and peers achieve cost‑effective, high‑density memory solutions, they could exert pricing pressure on incumbent players such as Samsung and SK Hynix, potentially altering the profitability outlook for the AI semiconductor ecosystem.
Market Fragmentation: Winners vs. Borrowers
The net effect is a bifurcation of AI‑linked stocks. On one side stand firms like TSMC and ASML, whose strong cash flows, modest leverage, and indispensable technological moats have kept investor confidence intact. On the other side lie companies such as CoreWeave and other “NeoCloud” operators, whose aggressive expansion is financed through substantial debt piles, rendering them more sensitive to rate swings and earnings volatility. This divide is reflected in recent price moves: Nvidia slipped ~8% and TSMC ~4% in the week of October 15, while GPU‑lending names experienced even steeper declines as market participants reassessed the risk‑return profile of heavily leveraged AI infrastructure plays.
Conclusion: AI Still Central but Metrics Matter
In sum, the AI boom has not evaporated; rather, it has entered a phase of selective enthusiasm. Foundries and equipment makers continue to ride a wave of robust demand, posting record profits and plotting capacity expansions that still lag behind market needs. Meanwhile, the data‑center and cloud segments are grappling with the consequences of aggressive debt‑funded growth, prompting investors to demand greater transparency around cash flow, leverage, and interest‑rate exposure. As the article concludes, the market’s lens has sharpened: “AI is still a key theme in the stock market, but ‘how much you earn and how much you owe’ is now more important than ‘Is it an AI company?’” Investors who can discern which companies possess the financial resilience to sustain their AI ambitions will be best positioned to navigate the next chapter of this transformative sector.
https://www.mk.co.kr/en/world/12178640