Jensen just put Nvidia’s Blackwell and Vera Rubin sales projections into the $1 trillion stratosphere

Nvidia CEO Jensen Huang projected at least $1.0 trillion of orders for the company’s Blackwell and Vera Rubin chips through 2027, up from about $500 billion in demand cited through 2026. Rubin, announced in 2024, is claimed to run ~3.5x faster on training and ~5x faster on inference (up to ~50 petaflops); Nvidia expects to ramp production in H2. The guidance implies a materially larger TAM for Nvidia’s AI hardware, supporting stronger revenue and capex outlooks and positive implications for suppliers in the AI ecosystem.

Analysis

The headline projection crystallizes two durable structural changes: hyperscalers re-architecting data centers around accelerator density, and the upstream supply chain (foundry, advanced packaging, HBM, power/networking) becoming the binding constraint on realized revenue. Expect incremental dollars to flow disproportionately to suppliers that can expand capacity on a 6–24 month cadence (TSMC/ASML/advanced substrate partners) and to high-margin networking/telemetry components that enable rack-scale AI clusters.

Key near-term risks are executional rather than market signaling: wafer/packaging yield curves, HBM wafer allocations, and assembly test turnaround times will govern shipment cadence over the next 6–12 months and can create multi-quarter volatility even if secular demand holds. Medium-term threats (12–36 months) include rapid adoption of model compression or sparsity techniques and any meaningful tightening of export controls, each of which can materially reduce hardware intensity per inference and reroute demand geography.

From a positioning lens, the most convex exposure is via long-dated, concentrated exposure to the lead accelerator vendor balanced with targeted supply-chain longs (foundry, memory, networking) and a volatility-hedge that protects against delivery squeezes or a demand re-rating. Tactical opportunities exist to monetize headline-driven rallies via covered-call overlays or to establish asymmetric long-call structures where theta decay is limited but upside remains large.

Contrarian risk: the market is likely underestimating demand concentration and overestimating linearity of hardware spend. A handful of cloud customers account for the bulk of incremental orders; any shift in their internal architecture choices or model economics (e.g., 4-bit quantization adopted broadly) will compress TAM assumptions rapidly. Treat consensus capacity commitments as directional signals, not guaranteed revenue, and size positions accordingly.

Key Decisions for Investors

• Long NVDA LEAPs (24-month calls) sized 3–5% of book; hedge with a 12-month put (buy-put + long-call) to cap downside. Rationale: highest convexity to the acceleration narrative; target 3:1 upside/downside if production ramps on schedule; unwind 50% on a 30% realized-shipment beat.
• Pairs trade: Long TSM (TSM) or 12–18 month call spreads (benefit from foundry capacity expansion) / Short INTC (12-month puts) to express shift from CPU to GPU-centric builds. Timeframe 12–24 months; reward: capture structural foundry re-rating vs legacy x86 exposure; key risk is TSMC ramp delay or Intel foundry progress.
• Supply-chain exposure: accumulate MU and ASML over next 6–12 months in tranches (20–30% of target each month). Rationale: memory and lithography tighten on advanced-node and HBM demand; risk/reward positive if yields and allocations normalize; beware cyclical inventory risk—use stop-losses at 20% drawdown.
• Volatility management and tactical monetization: sell short-dated NVDA covered calls into spikes (>20% move) to harvest implied vol premia, and maintain a tail hedge (buy 6–12 month NVDA puts sized to 1–2% of NAV) to protect against execution or geopolitical shocks. This captures upside while limiting single-name blowups.