Practical finance

The great AI data center analysis Part 2 – Jan 2026

Link to part1

We’ll cover the core segments and key public companies within each, including their recent performance, valuation posture, competitive moat, and how they’d fare if AI spending slows.

🧠 1) AI Chips & Accelerators — The Core Compute Layer

NVIDIA (NVDA) — The Dominant Leader
  • 2025 Revenue: ~$130.5B (+114% YoY) with Data Center > $100B; Data Center revenue surging over 400% in the latest quarter.
  • Margins: ~75% gross, ~55% net — exceptionally high for hardware.
  • Growth Drivers: Blackwell/H200 GPUs, software ecosystem (CUDA), hyperscaler demand.
  • Valuation: Premium — forward P/E ~33–53× and EV/Sales ~19×; high multiples reflect expected long-term AI expansion.
  • Moat:
    • Ecosystem lock-in through CUDA (developers building on Nvidia libraries).
    • Brand leadership in performance and time-to-market.
  • Diversification: Low outside AI compute. High exposure means earnings swing with hyperscaler capex — making it powerful but also cyclical if AI capex slips.

Summary: Nvidia is the engine of AI compute with unmatched growth — but its valuation is priced for perfection. Any slowdown in hyperscaler spending can amplify volatility.

👉 If AI demand slips, NVDA’s high multiples make it more sensitive than diversified players.

AMD (AMD) — A Growing Challenger
  • Data Center Growth: Management forecasts data-center revenue growth of ~80% annually in coming years.
  • 2025 Trends: Rapid expansion in EPYC CPUs + Instinct GPUs; multi-GW commitments from OpenAI.
  • Valuation: ~41× forward P/E — lower than NVDA but still rich; reflects growth expectations.
  • Moat:
    • Architecture breadth (CPUs + GPUs).
    • Multi-customer footprint across AWS, Meta, Oracle, and others.
  • Diversification: Medium — enterprise + gaming CPUs balance cyclical data-center GPU demand.
  • What’s overlooked: AMD’s server CPU share bolsters non-GPU revenue — a diversification edge vs pure GPU vendors.

Comparison vs NVDA: NVDA’s compute share and margins are higher, but AMD’s multi-product mix gives an earnings cushion if AI growth decelerates.

Micron Technology (MU) — Memory Powerhouse
  • 2025 Performance: Record revenue ~50%+ YoY; HBM/DRAM revenue multiples expanding quickly.
  • Valuation: ~9–10× forward P/E — much cheaper than peers, even as growth accelerates.
  • Growth Drivers:
    • HBM (High-Bandwidth Memory): Essential for AI GPUs/accelerators.
    • Partnerships with NVDA + AMD.
  • Margins: HBM business boosted gross margins from ~22% to >50% recently.
  • Moat:
    • Scarcity in HBM capacity worldwide.
    • Deep technical integration with AI compute stacks.
  • Diversification: Medium — memory also sells into PCs and mobile, cushioning cyclical downturns.

Key Insight: MU’s combination of massive growth, valuation discount, and critical tech means it can outperform big-chip peers if memory pricing remains firm.

🌐 2) Networking & Interconnect — The Nervous System

Arista Networks (ANET) — High-Speed Switch Leader
  • 2025 Financials: ~30%+ revenue growth; earnings up ~38%.
  • AI Role: Switches that tie GPUs and servers into clusters are mission-critical for performance.
  • Moat:
    • Strong relationships with hyperscalers (Microsoft, Amazon, Google).
    • Product roadmaps tied to AI cluster needs.
  • Valuation & Risk: Premium valuation due to growth expectations. If AI builds slow, interconnect demand lags after compute, leading to temporary bumps.
Cisco Systems (CSCO) — Diversified Networking & Security
  • 2025 Revenue: ~$56.7B with double-digit product order growth.
  • Moat:
    • Incumbent in enterprise networking.
    • Recurring revenue from security & software.
  • Diversification: High — sells beyond AI into routers, security, and services.
  • AI Edge: Integrated AI networking products but not as dependent on hyperscaler AI builds as Arista.

Summary: Cisco is a safe networking play that benefits from AI indirectly but won’t crater if AI builds slow.

⚡ 3) Infrastructure — Power, Cooling, Chassis

Vertiv (VRT) — Critical DC Infrastructure
  • 2025 Growth: Stock up ~43% on AI data-center demand. Q3 revenue ~$2.7B, earnings +63%, guidance still strong.
  • Product: UPS, cooling (liquid & precision), racks.
  • Moat:
    • Specialized hardware essential for high-density AI clusters.
  • Diversification: Medium — also serves non-AI DCs and enterprise.
  • Risk: More cyclic because demand tied to new builds.

Vertical Reflection: A pure infrastructure play — high beta to data-center construction cycles. Best choice when capex is booming.

nVent Electric (NVT) — Electrical Distribution
  • 2025 Results: ~25–28% revenue growth, net income +84% on DC-related projects.
  • Moat:
    • Niche electrical components critical for power delivery.
  • Diversification: Better than Vertiv — sells components across industries beyond data centers.

Bullish Angle: Often under the radar — strong growth but smaller cap. Good way to play electrical buildouts.

Amphenol (APH) — Connectivity & Cabling
  • 2025 Performance: Stock +95%; Q3 revenue ~$6.2B, earnings +86%.
  • Moat:
    • Massive scale and broad product lineup.
    • Essential connectors/fiber for servers & networking.
  • Diversification: High — sells into aerospace, automotive, mobile, and telecom.
  • Undervalued Aspect: Connectivity gear is critical but overlooked by many AI investors.

🏢 4) Cloud & Data Center Owners

Equinix (EQIX) & Digital Realty (DLR) — Data Center REITs
  • Equinix: Long-term demand driven by colocation + cross-connect services.
  • Issue: 2025 revenue and growth forecast disappointed, with slower near-term growth expectations.
  • Capital Intensity: High — data center expansions require big capex.
  • Moat:
    • Global footprint, connectivity, long-term leases.
  • Risk: If hyperscalers slow synthetic demand, leasing might lag — but diversified tenant bases reduce risk.

📊 5) Big Cloud Providers (AI Hosts + Brokers of Demand)

Microsoft (MSFT) / Amazon (AMZN) / Alphabet (GOOGL)
  • All three are amid gigantic cloud and AI capex cycles; Azure and AWS have sustained growth rates ~30–40%+.

Moats & Risk:

  • Very diversified: software, ads, e-commerce (for AMZN), productivity tools (MSFT), search (GOOGL).
  • Cloud is just one piece of the business — less risky if AI demand cools.
  • These are demand generators, not infrastructure suppliers, making them core players but less direct “picks & shovels.”

🤝 Valuation & Diversification Summary

CompanySegmentGrowthValuationDiversificationMoat Strength
NVDAGPU🚀 Very HighPremiumLowVery Strong
AMDCPUs/GPUs🚀 HighHighMediumStrong
MUMemory🚀 HighCheapMediumStrong
ANETNetworkingHighHighMediumStrong
CSCONetworkingModerateReasonableHighModerate
VRTDC InfrastructureHighMidMediumNiche
NVTElectricalModerateAttractiveMediumNiche
APHConnectorsHighMidHighDiversified
EQIX/DLRData Center REITModerateUtility-likeMediumGlobal footprint
AMZN/MSFT/GOOGLCloudVery HighPremiumVery HighVery Strong

🧠 Key Takeaways for Investors

📌 Biggest Bottleneck Remains Power + Memory

  • Memory (HBM) demand is tight; Micron’s price and production trajectory suggests potential supply shortages — a pricing tailwind for earnings.

📌 Compute Market Is Consolidated

  • Nvidia still dominates the AI chip space but AMD is climbing. Both have different risk profiles (NVDA high growth, AMD better diversification).

📌 Infrastructure Players Offer Defensive Exposure

  • Amphenol and Cisco may outperform pure AI names if growth slows because they sell into multiple industries.

📌 Cloud Providers Are Demand Base, Not Pure Infra

  • MSFT/AMZN/GOOGL benefit as demand drivers — but AI slowdown would affect their cloud growth less severely than it would chip or infrastructure suppliers.

Leave a comment