Broadcom Hoktan “The biggest constraint is capital”... Cash flow pressure due to big tech facility investment
Stock prices plummet despite Samsung's record-breaking performance and Hynix's forecast of KRW 80 trillion
As artificial intelligence (AI) semiconductor supply shortages continue and U.S. big tech companies reap enormous profits, heads of global semiconductor companies have publicly mentioned the burden of financing due to expanded AI investment. As funds invested in data centers and cutting-edge semiconductors increase rapidly, there are concerns that investment resources and profitability may determine the growth rate of the AI industry as much as semiconductor supply capabilities.
At the ‘DBS Asia Future Forward Conference’ held in Singapore on the 12th, Broadcom CEO Hock Tan said, “The biggest constraint may be capital.” This means that the enormous investment costs required to expand AI infrastructure cannot be resolved simply by increasing semiconductor and power supply.
Intel CEO Lip-Bu Tan also cited capital and capital raising costs as immediate challenges. “In the long run, people will want to see the return on investment (ROI),” he said, emphasizing that the money invested in AI must lead to actual business results.
Cash flow pressure due to rapid increase in AI facility investment... External procurement demand of USD 1.5 trillion
Big American tech companies such as Microsoft, Alphabet, Amazon, and Meta are generating enormous amounts of cash from existing businesses such as advertising, cloud, and software. AI is also contributing to sales growth through expansion of cloud services and improvement of advertising efficiency.
The scale of facility investment, including AI infrastructure, by these companies is also rapidly increasing. According to Barclays Private Bank, the cash facility investment of these four companies in the second quarter of this year was $165 billion, a surge of 87% compared to the same period last year. During the same period, free cash flow excluding facility investment from cash earned from operating activities totaled only $6.7 billion. Although it generated a huge amount of cash, most of it was put back into facility investment.
Investment targets are not limited to semiconductors. It is necessary to have a data center site, power grid, cooling facilities, and network equipment. As competition for AI model performance becomes fiercer, more computing power and power are needed, and the initial investment cost to secure them also increases.
Morgan Stanley estimated that a total of $2.9 trillion will be needed for AI infrastructure investment by 2028, of which $1.5 trillion will depend on external funding. If major big tech companies have so far made investments based on their own cash flow, the role of the bond market and financial institutions may grow even greater in the future.
CEO Hock Thanh also raised questions about the profitability of companies that have adopted AI. “Most companies are not finding a return on investment,” he said, pointing out the reality of not achieving clear results even though they are pouring money into using AI.
He then asked, “Didn’t you have the experience of receiving $10 from your parents as a child and going to a game room, spending it all in 10 minutes and not learning anything?”
However, this does not mean that the two CEOs expected a sharp slowdown in AI demand. CEO Hock Thanh mentioned the possibility that AI infrastructure expansion will continue beyond 2029 and into 2030, and CEO Lipbu Thanh also predicted that strong demand will continue for the next two to three years.
The two CEOs' remarks focused on rapidly increasing investment costs and resulting profitability issues rather than AI demand itself.
Samsung 107 trillion, Hynix 80 trillion... Foreign selling despite record-breaking performance
So what about companies that supply semiconductors? Samsung Electronics and SK Hynix, which are benefiting the most from the AI investment competition, are also seeing mixed performance and stock prices.
On the 8th, Samsung Electronics announced preliminary results of 195 trillion won in sales and 107.4 trillion won in operating profit on a consolidated basis for the third quarter. Operating profit increased 782.5% compared to the same period last year, recording the highest quarterly performance ever.
SK Hynix is also expected to record its highest performance ever. Meritz Securities estimated SK Hynix's third quarter operating profit at 80.1 trillion won on the 12th. It is a 32% increase from the previous quarter, and Meritz Securities researcher Kim Seon-woo analyzed that increased shipments of DRAM and NAND and an increase in average selling price will lead to improved performance.
The background to the two companies' strong performance is a shortage of supply of high-bandwidth memory (HBM) and server DRAM, and rising memory prices. As investment in AI data centers surged, semiconductor supply could not keep up with demand, and the profitability of memory companies also improved significantly.
Stock prices moved in the opposite direction. On the 12th, Samsung Electronics closed at 250,500 won, down 4.39%, and SK Hynix closed at 1,605,000 won, down 4.52%. Samsung Electronics also fell 2.42% on the 8th, the day of its provisional performance announcement.
Foreign selling led the decline. In the stock market on the 12th, foreign investors net sold about 2 trillion won worth, and KOSPI closed at 6384.24, down 3.65%. Investment sentiment has weakened due to the rise in U.S. Treasury yields, instability in international oil prices, and profit taking due to the surge in semiconductor stocks.
US semiconductor stocks also showed weakness. On the 12th, the Philadelphia Semiconductor Index fell about 2.7%, and Broadcom and Intel stock prices fell 1.9% and 3.2%, respectively.
Performance was at an all-time high, but stock prices plummeted. How long the current semiconductor boom will continue, the market's attention is already turning to the next step.
Korean semiconductor capacity expansion accelerates... AI investment sustainability determines long-term demand
Samsung Electronics and SK Hynix are expanding investment in production facilities in response to demand for AI memory. Samsung Electronics is strengthening its cutting-edge memory, foundry, and packaging capabilities, and SK Hynix is seeking to expand production capacity centered on the Yongin semiconductor cluster.
SK Hynix is seeking to shorten the construction schedule for its Yongin production facility. A plan to build an additional production base in the Honam region is being discussed, but the investment scale and construction start schedule have not been confirmed.
Currently, the memory market is experiencing a supply shortage, especially for HBM and server DRAM. As major customers' investments in AI infrastructure are also expanding, there is limited evidence that demand will decline sharply in the short term.
However, the time to build a semiconductor factory and the time for AI companies to make investment decisions are different. Although it takes several years from construction to mass production for a memory factory, a customer's investment plan can be adjusted at any time depending on financing costs and business performance.
CEO Hock Tan predicted that wafer supply constraints could be eased around 2028-2029. Currently, it is difficult to receive all orders in the market due to a shortage of semiconductors, but as supply capacity increases, customer purchasing ability and additional investment plans may become more important.
The competition between Samsung Electronics and SK Hynix to expand capacity is an investment that goes beyond the immediate supply shortage and anticipates the growth of the AI market in the next few years. Currently, competitiveness depends on how quickly and how much semiconductors can be supplied, but the situation may change when the factory is completed. It is not only semiconductor production capacity that determines the growth rate of the AI industry. There also remains the question of whether customer companies can make enough money to continue purchasing the semiconductors and invest again.
