Samsung Electronics said on Thursday that its third-quarter operating profit rose nearly ninefold to a record high, driven by robust demand for chips used in artificial intelligence systems. The world’s largest memory chipmaker reported operating profit of 107.4 trillion won, about 80.04 billion dollars, up from 12.2 trillion won a year earlier and in line with its earlier estimate, even as investors questioned how long the artificial intelligence spending boom can last.
The headline numbers
A near-ninefold rise in operating profit is not a normal seasonal improvement, and Samsung itself presented the result as a record. The company also said it would pay 30 trillion won in dividends for the third quarter, which ended in September, at 4,601 won per share. That payout, reported alongside the results, underlines how much cash the memory cycle is currently generating for the company.
Investors remained wary about the durability of artificial intelligence spending, a caution that has shadowed chip stocks even as their earnings have surged. The worry is familiar in a cyclical industry: today’s shortage can become tomorrow’s glut, and customers who are ordering heavily now may already have more memory on order than they can use.
A shortage that deepens
Samsung said the chip supply shortage would deepen next year, echoing the outlook given by its cross-town rival SK Hynix a day earlier. When supply cannot keep pace with orders for the memory used in artificial intelligence servers, prices and profits rise, but buyers face longer waits and higher costs for the hardware they need.
SK Hynix reported strong earnings on Wednesday, with an upbeat outlook that helped ease market concern that artificial intelligence-driven chip earnings could lose momentum. The two results landed in sequence, and together they told a consistent story: demand for high-bandwidth memory and other server memory remains strong, and both of South Korea’s memory giants are benefiting.
Why memory is at the centre of AI
Training and running large models requires vast amounts of fast memory close to the processor, and the servers that host those models are packed with it. When data-centre operators expand capacity, they buy processors and memory together, so a boom in artificial intelligence infrastructure becomes, almost directly, a boom in memory sales.
Memory prices are set by a small number of large suppliers, and production cannot be expanded quickly. New fabrication capacity takes years to build and bring online, so a sudden surge in orders translates into tight supply and rising prices rather than a rapid increase in output. The result is record profit for suppliers, and a higher bill for everyone building artificial intelligence capacity.
Why It Matters
It shows that the money is not only flowing to the best-known model makers and processor designers, but also to the suppliers further down the chain whose components make those systems possible. A record quarter for the world’s largest memory maker is, in that sense, a measure of how much physical hardware the current wave of artificial intelligence investment actually requires.
The caution in the report matters just as much. If the shortage deepens next year, as Samsung expects, the cost of building artificial intelligence capacity will stay high, and smaller buyers may be squeezed first.
That tension, between record supplier profits and wary investors, is one reason chip executives keep warning about a glut even while reporting record quarters. Nvidia chief executive Jensen Huang has made a similar point about the longer term, as covered in our report on his memory outlook.
Background
Samsung is the world’s largest memory chipmaker, and its quarterly results are watched as a barometer for the wider semiconductor cycle. Its rival SK Hynix reported its own strong third-quarter results the day before Samsung.
Source: Reuters report
0 Comments