Samsung Electronics and Taiwan Semiconductor Manufacturing Company have both posted record profits, according to a Yahoo Finance summary, reinforcing the view that chip demand tied to AI has not yet rolled over.

The two companies sit at the top of the global contract chipmaking market. Their financial results offer one of the clearest reads on whether spending on AI infrastructure is still climbing or starting to fade. When both print all-time-high earnings in the same period, the inference is that orders for advanced processors and AI-relevant memory remain strong across the supply chain.

What do the record profits signal?

Chip designers such as Nvidia, AMD, and the in-house silicon teams at large cloud providers order wafers months before their own products reach end customers. That activity flows into foundry revenue and operating margin with a lag that mirrors production cycles rather than end-market cycles.

A record profit figure at a foundry therefore carries three signals at once. It points to high utilization of the most advanced process nodes, to pricing power where capacity is genuinely scarce, and to a product mix that tilts toward higher-value AI silicon.

Designer earnings can rise on margin expansion alone, because gross margins depend on average selling prices as well as volume. Foundry earnings, by contrast, rise mainly when wafer volume and the mix toward leading-edge nodes improve at the same time.

The takeaway: when both major foundries set profit records in the same quarter, the AI trade is still flowing through the production base.

How do Samsung and TSMC capture the AI trade?

Samsung and TSMC do not capture the AI boom through identical channels. Samsung combines a foundry business with one of the world's two large memory operations. Its memory division has been the larger beneficiary of the AI build-out, because high-bandwidth memory sits inside accelerator packages that have driven much of the recent demand.

TSMC is a pure-play foundry. Its revenue comes almost entirely from wafers it fabricates for external designers. That customer base includes the bulk of AI accelerators sold by Nvidia and AMD, plus custom AI silicon developed by hyperscalers.

When TSMC posts a record profit, the underlying read is that wafer demand from that customer set is at or above the previous peak. The simultaneous record prints suggest the AI trade remains broad-based. It is not confined to memory, and it is not driven by one designer alone.

The takeaway: the AI trade is reaching both memory vendors and logic foundries at full strength, not just one corner of the supply chain.

What could still break the pattern?

Visible risks include a sudden pullback in cloud capital expenditure, an inventory correction at a major customer, and yield problems at a leading-edge node. None of these yet show up in the reported numbers.

The absence is itself informative. Foundry results are usually the first place an underlying weakening of demand shows up, because foundries must commit capacity months in advance. If hyperscaler capex were about to soften, the order books would tighten before the data centers actually slowed down.

For now, the inference from Samsung's and TSMC's bottom lines is that the AI infrastructure spend cycle remains intact. The next test will arrive when each company issues guidance for the coming quarter, and that is where any softening would first appear in forward-looking language.