SK Hynix posts record Q1 net profit on AI memory demand
Net profit rose nearly fivefold year-on-year, driven by surging prices for AI-grade memory chips
Briefing
- •SK Hynix reported record first-quarter net profit and revenue, with net profit rising nearly fivefold year-on-year.
- •Results came in line with analyst estimates, suggesting the AI memory demand cycle is tracking expectations rather than surprising to the upside.
- •The primary driver was continued price appreciation across its memory product range, fuelled by AI infrastructure buildout.
- •SK Hynix is a key supplier to Nvidia, making its results a proxy for high-bandwidth memory demand tied to GPU shipments.
Why it matters
- →SK Hynix results arriving in line rather than ahead of estimates matters more than the record headline profit. It confirms the AI memory demand curve is fully priced into consensus, removing the upside surprise mechanism that has driven multiple expansion in HBM-exposed names. Nvidia's upcoming print now carries the burden of proving GPU shipment volumes are accelerating faster than the memory supply chain already telegraphed.
- →SK Hynix's results, combined with TSMC's record Q1 beat, establish that both the leading foundry and the leading HBM supplier are running at cycle-high utilization rates driven by the same AI infrastructure buildout. This dual confirmation reduces the probability that any single hyperscaler CapEx pause would materially dent near-term semiconductor demand, given the supply chain is already committed through current order books.
Historical Context
- 2021-2022
The prior memory upcycle peaked when DRAM and NAND prices collapsed as consumer electronics demand cratered post-pandemic. SK Hynix's in-line rather than beat result echoes late-cycle dynamics where supply additions and demand plateaus compress upside surprise potential even as absolute profits remain elevated.
- 2016-2018
SK Hynix and Samsung posted multi-year record profits during the server DRAM supercycle driven by hyperscaler buildout. That cycle ended abruptly when cloud capex moderated and inventory overhang built. The current AI HBM cycle is structurally different in that supply is far more constrained, but the in-line result signals consensus has already captured the demand visibility.
Connected Events
TSMC Q1 profit surges 58% to record high, beating estimates on AI demand
TSMC's Q1 profit surged 58% to a record high on AI demand and raised its full-year revenue outlook, confirming durable AI foundry demand through at least H1 2026 and reducing the probability of downward guidance revisions from Nvidia and AMD.
Google Cloud launches 8th-generation TPUs targeting Nvidia in AI chip market
Google Cloud's eighth-generation TPU launch is positioned as a cheaper alternative to Nvidia GPUs for AI training and inference workloads. If adoption accelerates within GCP, it reduces the incremental HBM content per AI compute dollar, as TPUs use different memory architectures than H100 and B200 GPU clusters, creating a directional headwind to HBM volume growth at the margin.
