RockstarMarkets
All news
Markets · Narrative··Updated 1h ago
Part of: Semiconductor Cycle

Memory Chip Valuations Hit 2025 Highs Despite Frothy Multiples; NVDA, AMD, AVGO Extended on AI Demand

NVDA, AMD, and AVGO have surged to fresh record valuations as insatiable AI infrastructure demand for VRAM, HBM, and networking chips overwhelmed supply. Despite trading at elevated multiples (price-to-earnings ratios), semiconductor stocks continue outperforming as Cerebras IPO surge and Hyperledger ecosystem growth validate that AI capex cycle remains in early innings, not peak.

R
Rocky · RockstarMarkets desk
Synthesised from 8 wires · 40 mentions in the last 24h
Sentiment
+70
Momentum
75
Mentions · 24h
40
Articles · 24h
22
Affected sectors
Related markets

Key facts

  • Cerebras Systems IPO raised $5.55B on May 15, surging on debut; validates AI chip demand
  • Memory chip backlogs extending to 2027; Broadcom, Micron signaling sustained strong demand
  • AI semiconductor names trading at 40-60x forward earnings, near 2018 peak multiples
  • NVDA gained another 4.4% on China H200 approval, validating $5-10B revenue upside
  • Distributed capex thesis: NVDA + AMD + AVGO + Broadcom all gaining share, not zero-sum

What's happening

One of the market's most persistent fears, that AI capex is nearing a peak, was dealt a blow on May 15 by the blockbuster IPO of Cerebras Systems, an AI semiconductor company that raised $5.55 billion and surged on debut. Simultaneously, memory chip makers like Micron Technology, SK Hynix indications, and chipmakers Broadcom and AMD all signaled that orders for high-bandwidth memory (HBM), VRAM, and AI networking silicon remain backlogged well into 2027. The narrative flipped from "is capex slowing?" to "capex is doubling down."

The engine is straightforward: every AI model training cluster, every inference farm, and every enterprise deployment requires not just GPUs (NVDA's dominion) but also the DRAM, HBM, and networking glue that ties it all together. Broadcom has become the "picks and shovels" play, supplying the switching and interconnect fabric for data centers. AMD's EPYC CPUs and Instinct accelerators are gaining share from Intel. Micron is scrambling to expand HBM output. Even older chip names like Rambus are getting bids from AI-hungry customers. The conventional wisdom that one stock (NVDA) would capture all the upside has given way to a more distributed narrative: AI capex benefits the entire semiconductor stack.

Valuations are stretched, however. Many of these names trade at 40-60x forward earnings, a level that left the broader market in 2018. The justification is that AI is a multi-year, multi-trillion-dollar infrastructure rebuild, not a one-year bump. Orders are real, not vapor. But the gap between price and fundamental value has widened, and any disruption to demand, whether from an economic slowdown, a shift to more efficient models, or regulatory intervention, could trigger a sharp repricing. Enterprise cloud budgets are being redirected toward AI, but consumer spending remains soft, raising questions about whether capex can sustain without a broader economic reacceleration.

For now, the momentum is squarely with the bulls. Hyperledger decentralized AI initiatives, Solana's AI token ecosystem, and traditional OEMs' custom silicon programs all suggest the capex cycle is fractal: it is not just NVDA, it is NVDA plus 50 other companies in a supply chain race. The risk is concentration; the opportunity is diversification. Traders who bought the whole semiconductor stack in May are outperforming those who went all-in on a single name.

What to watch next

  • 01Micron or SK Hynix earnings and HBM production guidance: June 2026
  • 02Broadcom data center revenue growth and backlog update: June earnings
  • 03Supermicro SMCI earnings and AI server shipment trends: June 2026
Mention velocity · last 24 hours
Coverage from these sources
Previously on this story

Related coverage

More about $NVDA

Topic hub
Semiconductor Cycle: AI Capex, Memory and the SOX Trade

Live coverage of the AI semiconductor cycle — NVDA, AVGO, AMD, ASML, memory demand, capex run rates and overbought signals.