IBM fell 25 percent on Tuesday, and ASML raised its annual forecast overnight. The results look like separate company stories, but together they suggest that the AI boom is changing the order in which businesses spend their technology budgets.
Companies aren’t pulling back from AI. They’re directing more money toward servers, storage, memory and chipmaking capacity that could become harder or more expensive to secure. Because those budgets aren’t unlimited, software deployments, consulting engagements and other large projects can be delayed even while overall AI spending continues to rise.
IBM offered the clearest view from the buyer’s side. The company warned that second-quarter revenue and profit would fall below expectations, and management acknowledged that poor execution contributed to the shortfall. Several large deals didn’t close on schedule, while its mainframe business performed worse than expected.
But CEO Arvind Krishna also said customers had shifted quarterly capital spending toward servers, storage and memory ahead of expected price increases. In practical terms, companies spent first on physical computing equipment they feared might cost more or become difficult to obtain later.
That urgency changes the purchasing order. A company worried about shortages can’t assume the infrastructure it needs will still be available at the same price six months from now, while a software implementation or consulting project is usually easier to postpone.
IBM’s internal results support that distinction without turning the quarter into a clean hardware-versus-software story. Its distributed-infrastructure business grew 37 percent and ended the quarter with a backlog of about $500 million, while Red Hat revenue increased 11 percent.
Technology demand didn’t disappear across IBM. Customers became more selective about which commitments they made first.
Poor execution still played a meaningful role, and some of the delayed deals may close in the third quarter rather than disappear. But the strength of IBM’s infrastructure business suggests that the company’s problems weren’t simply the result of customers abandoning technology spending.
ASML showed the same shift from the supply side. The Dutch company makes the specialized lithography machines used to produce advanced semiconductors, placing it close to one of the hardest physical constraints in the AI buildout.
ASML reported €9.3 billion in second-quarter sales and €2.9 billion in net income, both above its previous guidance. It also raised its 2026 revenue forecast from a range of €36 billion to €40 billion to a new range of €43 billion to €45 billion.
Management said AI-related investment is causing chipmakers to accelerate their expansion plans. ASML now expects to increase production capacity for its most advanced lithography systems by roughly 30 percent in 2027, with another increase under consideration for 2028.
Service revenue from machines already installed at customer facilities also contributed to the quarter, so the upside wasn’t entirely the result of new AI equipment orders. Even so, companies don’t expand manufacturing capacity on that scale unless they expect demand to persist.
The reaction spread across South Korea’s semiconductor supply chain. SK Hynix, a major producer of the high-bandwidth memory used in AI systems, rose nearly 13 percent. Samsung Electronics, which is also one of the world’s largest memory-chip manufacturers, gained almost 8 percent, while Hanmi Semiconductor, a maker of advanced chip-packaging equipment, rose about 25 percent.
Those gains followed softer U.S. inflation data as well as stronger demand expectations, so the price moves alone don’t prove that corporate budgets are being reallocated. But the companies occupy different parts of the same physical buildout, and all are benefiting from the difficulty of adding memory, manufacturing and packaging capacity quickly.
That constraint is beginning to shape which technology purchases happen first. When buyers expect servers, memory or chip capacity to cost more later, those commitments move forward, while projects with more flexible timing move back.
The result can be a narrower technology boom than broad indexes suggest. Semiconductor suppliers may continue reporting strong orders while software and consulting companies face longer sales cycles, not because businesses have lost interest in technology, but because infrastructure is absorbing more of the money available now.
The distinction should become clearer in upcoming results from Microsoft, Oracle, ServiceNow, Salesforce, SAP and Workday. Stable bookings and normal sales cycles would suggest IBM’s problems were mostly company-specific. Repeated references to delayed approvals or customers prioritizing infrastructure would point to a broader change in purchasing behavior.
Memory prices, ASML’s orders and cloud-provider capital spending will show whether the urgency continues on the supply side. If shortages ease and delayed software projects return quickly, the pressure may prove temporary. If infrastructure commitments keep rising while other projects continue slipping, the AI boom isn’t expanding the technology budget evenly. It’s deciding which purchases get funded first.