Closelook
NVIDIA architecture roadmap for investors: the Rubin Era (2026-2027, simulation and memory bandwidth) transitioning to the Feynman Era (2028-2030, edge actuation and latency)
Closelooknet
Home ACTION Weekly Signal Trading Portfolios ENGINE
Dossiers
ResearchAI Chip Supply ChainNVIDIA EcosystemAgentic ApplicationsSoftware-Credit Nexus
Indices
FrameworksABR FrameworkIndicesRubin Build-Out 100Euro-AI Sovereign 50
101 About Newsletter ↗
101
AI Infrastructure 101
Action
Weekly SignalTradingPortfolios
Engine
ABR FrameworkRubin IndexDossiers
AboutNewsletter ↗
Home / 101 / NVIDIA Investor's Timeline
TimelineNow 2026Closelook

NVIDIA Investor's Timeline: Hopper to Blackwell to Rubin and Beyond

NVIDIA releases new GPU architectures on roughly annual cadence: Hopper (2022), Blackwell (2024-2025), Rubin (2026), and a next-generation architecture (2027+). For investors, what matters is not the architecture specs but the revenue impact per generation, the supply chain implications (which suppliers win new design wins?), and what's already priced into NVIDIA's stock at 50x+ forward earnings. Closelook tracks each generation through the Functional Index and Sentinel Tickers, focusing on supply chain readiness rather than NVIDIA's stock price.

Hopper (H100/H200) — Revenue Peak, Mature Phase

Hopper is NVIDIA's cash machine. The H100 drove NVIDIA from $15B to $130B+ annual revenue. By 2026, Hopper is mature — production is well-ramped, supply constraints are easing, and average selling prices are declining as used H100s enter the secondary market. Investment implication: Hopper is the present, not the future. Companies that benefited from Hopper ramp (TSMC CoWoS, SK Hynix HBM3) are now looking at Blackwell and Rubin for growth.

Blackwell (B200/GB200) — Current Ramp, High Expectations

Blackwell is ramping in 2025-2026 with higher compute density, higher power consumption (1,000W+), and higher HBM requirements (HBM3E, 192GB per GPU). The supply chain challenge: Blackwell requires more advanced packaging (CoWoS capacity), more HBM (memory constraint), and more cooling (liquid cooling mandatory). Blackwell is where most 2026 revenue growth expectations sit — and it's largely priced in.

Rubin (2026-2027) — The Next Cycle

Rubin will use HBM4, next-generation NVLink, and potentially new packaging technologies. The supply chain implications are significant: HBM4 requires new bonding techniques, NVLink evolution affects networking equipment demand, and higher power per chip tightens the cooling constraint. Closelook's Rubin Functional Index is named after this architecture because it maps the supply chain that builds it.

What's Priced In

At 50x+ forward P/E, NVIDIA's stock prices in near-perfect Blackwell execution and strong Rubin outlook. Any supply chain disruption, demand slowdown, or competitive threat (custom ASICs gaining share) creates downside risk. The investment question is not "is NVIDIA a good company?" — it's "is NVIDIA's stock a good risk-adjusted investment at this price?"

Key Companies

NVDA
NVIDIA
Architecture owner — Hopper/Blackwell/Rubin
TSM
TSMC
Fabrication + CoWoS packaging for all generations
000660.KS
SK Hynix
HBM supplier — transitions with each generation
AVGO
Broadcom
NVSwitch networking for GPU clusters

Closelook View

Closelook tracks NVIDIA's roadmap through the supply chain lens, not the stock price lens. The Rubin Functional Index maps which companies win or lose design wins with each architecture transition. NVIDIA itself is just one of 100 constituents.

Rubin Functional Index →NVIDIA Ecosystem Dossier →NVIDIA 2030 Thesis →

Related Entries

FrameworkFunctional Index→Framework6-Layer Model→ConstraintPackaging Bottleneck→ConstraintMemory Wall→ThemeCapEx Cliff→

© 2026 Closelooknet · Thomas Look · Substack · LinkedIn · X

Not financial advice. All content is for informational and educational purposes. Past performance does not guarantee future results.

Privacy · Terms · Imprint