Samsung Electronics reported on 30 July that its Device Solutions division, the arm that makes memory and logic chips, produced 127.5 trillion won of revenue and 89.2 trillion won of operating profit, carried by demand for AI server components and high-bandwidth memory.1 SoftBank Corp., the Japanese telecoms operator, said on 4 August that Cloud and AI revenue rose 31% year on year, lifting enterprise revenue 11% to 260.4 billion yen, with a 1,000-person team deploying a new managed patching service.2
SoftBank's thousand-person team is a rare disclosure of the labour behind an AI revenue line. Most disclosures in this cycle describe AI revenue without saying how many people produce it, and a managed service staffed by a thousand engineers is a straightforward statement that this particular AI product needs labour to run. Patching, the routine work of applying security updates across a customer's estate, is exactly the sort of task automation was supposed to erase.
Samsung's numbers point the other way on employment. A memory fab is among the most capital-intensive facilities in manufacturing, and its output scales with wafer capacity rather than with staffing. Korea's chip industry can therefore book record divisional profit in the same quarter its central bank reports young workers disappearing from information services and publishing, because the profit and the jobs sit in different parts of the same economy.
Samsung and SoftBank both sit downstream of the money. Demand for AI infrastructure is being converted into earnings by the firms that supply components and operate systems, several rungs below the model developers who receive the coverage. Amazon cut jobs inside its own AGI research unit last month while its infrastructure spending climbed, which is the same pattern that puts the employment gains attached to AI in hardware, telecoms and construction rather than in software.
