Which three chip stocks are down 15% or more?

Broadcom (down 23% from its spring highs), AMD (about 18% off its highs), and SK Hynix (around 20% following its IPO this year).

What is driving demand for these chip makers?

Cloud computing companies are seeing strong returns on AI infrastructure investments with locked-in contracts. Broadcom benefits from Alphabet's aggressive capex on custom TPUs and from other hyperscalers building their own AI chips. AMD is capturing inference work (where AI produces answers after training) through partnerships with OpenAI and Meta. SK Hynix supplies specialized high-bandwidth memory (HBM) that GPUs need to optimize performance.

How long is the AI spending cycle expected to last?

SK Hynix sees the HBM market remaining imbalanced until at least 2030, though the company notes there is a good chance the supercycle lasts much longer.

AI Stocks & MarketsYahoo Finance AI1h agoAug 21, 2026, 19:03 JST2 min read

3 AI chip stocks down 15%+ as investors seek value

Send on LINE

Key takeaway

Three major AI chip makers have tumbled 15–23% from recent highs, creating potential buying opportunities. Broadcom leads in custom AI chips and networking for cloud firms.
AMD is positioning itself in inference and agentic AI, where CPUs are becoming critical. SK Hynix dominates high-bandwidth memory, which GPUs require to reduce latency.
All three stand to benefit as cloud companies lock in long-term spending on AI infrastructure.

3 Key Points

What happened
Broadcom, AMD, and SK Hynix have each fallen 15% or more from recent highs. Broadcom is down 23% from its spring peak; AMD about 18% from its high; SK Hynix around 20% following its IPO this year.
Why it matters
Cloud companies are seeing strong returns on AI infrastructure investments with quick payback periods and locked-in contracts, signaling that robust AI spending will continue. Broadcom projects more than $100 billion in ASIC revenue in fiscal 2027 (with Citigroup estimating $180 billion in fiscal 2028), while AMD sees a $220 billion market opportunity in agentic AI CPUs. SK Hynix controls the HBM supply chain critical to GPU performance, with the market expected to remain imbalanced until at least 2030.
What to watch
Broadcom is trading below 20 times fiscal 2027 analyst earnings estimates; SK Hynix has a forward P/E around 6 times. Both valuations suggest entry points if the AI capex cycle holds.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Get news like this every morning — free

Context & Analysis

AI chip spending has remained volatile this year as investors question whether infrastructure capex is sustainable, but the evidence from cloud companies themselves suggests it will persist. Alphabet, Meta, OpenAI, and other hyperscalers have demonstrated that custom AI chips and specialized memory deliver strong returns with predictable payback periods, and they have locked in long-term supply contracts to secure capacity. This shift from relying solely on Nvidia's general-purpose GPUs to building custom silicon (as AMD and Broadcom enable) and securing dedicated memory supplies (SK Hynix's domain) reflects a maturation of the AI infrastructure market—moving from proof-of-concept to capital-intensive, contractually-binding deployment. The pullback in all three stocks has coincided with industry-wide concern about chip valuations, not a collapse in underlying demand. Broadcom's projection of more than $100 billion in ASIC revenue by fiscal 2027 and AMD's identification of a $220 billion agentic AI CPU market in the next few years are grounded in stated customer commitments rather than speculative forecasts.

FAQ

Which three chip stocks are down 15% or more?: Broadcom (down 23% from its spring highs), AMD (about 18% off its highs), and SK Hynix (around 20% following its IPO this year).
What is driving demand for these chip makers?: Cloud computing companies are seeing strong returns on AI infrastructure investments with locked-in contracts. Broadcom benefits from Alphabet's aggressive capex on custom TPUs and from other hyperscalers building their own AI chips. AMD is capturing inference work (where AI produces answers after training) through partnerships with OpenAI and Meta. SK Hynix supplies specialized high-bandwidth memory (HBM) that GPUs need to optimize performance.
How long is the AI spending cycle expected to last?: SK Hynix sees the HBM market remaining imbalanced until at least 2030, though the company notes there is a good chance the supercycle lasts much longer.

Yahoo Finance AIRead Original Article