$CrowdStrike Holdings, Inc.(CRWD)$
Forget finding "the next ChatGPT"—top VCs are going all-in on AI Data Center bottlenecks.
On Sept 28, Seligman Ventures raised its fund target from $500M straight to $1B—just 8 months after launching.
• Original target: 6–8 deals in Year 1
• Actual result: 14+ deals done ($300M+ deployed)
• Deal flow running 10–15x higher than expected
💡 Where is the money actually going?
Not into AI chatbots or wrapper apps, but into physical hardware bottlenecks:
🔹 Chip Interconnects & Optics: Eliyan, Lumilens
🔹 Next-Gen Compute: SambaNova, Velaura AI (low-power AI)
🔹 Infrastructure: Power, Cooling & Cybersecurity
📊 The Macro Picture
According to Reuters / Crunchbase data, US & Canada startups raised a record $392B in H1 2026. Semiconductor startups alone pulled in $10.7B—on track to smash full-year 2025 numbers.
📈 Why stock investors need to pay attention:
Private VC flows are usually a preview of public market rotations.
Buying GPUs is just step one. Thousands of GPUs linked together create massive real-world problems:
How do you transfer data fast enough between chips?
How do you feed hundreds of kW of power to a single rack?
How do you dissipate extreme heat?
How do you secure the entire physical data center?
The narrative is moving from "Who sells the GPU?" to "Who solves the new problems caused by having so many GPUs?"
🎯 Tickers to watch on this thesis:
• Networking & Optics: AVGO, ANET
• Power & Liquid Cooling: VRT
• Cybersecurity: CRWD, PANW, NET
⚠️ The Risk Factor
Hardware requires higher Capex and longer dev cycles. If Big Tech hyperscalers ever scale back AI Capex, hardware suppliers will feel the impact much faster than software firms. Heavy VC funding proves where the market is betting, not who will ultimately win.
📌 Bottom Line
Phase 2 of the AI trade isn't about building a slightly smarter model. It's about owning the essential plumbing—the tech without which GPUs literally cannot run.
Comments