Nvidia Believes Artificial Intelligence (AI) Capex Will Reach $3 Trillion to $4 Trillion by 2030. Here's Where Its Stock Price Could Go If It's Right.

Nvidia is projected to benefit from a data center AI capex boom, with global spending potentially reaching $3T–$4T annually by 2030 (vs. ~$650B planned by the big AI hyperscalers this year). The article expects Nvidia revenue/earnings could rise up to ~4x by 2030, assuming data center chips take a larger share of capex and that a likely move to custom AI chips only partially offsets Nvidia share gains. Under a simplified valuation (20x earnings), the piece estimates a potential market cap of ~$12.8T, implying roughly a +172% gain to about $530/share—though it notes Nvidia faces competitive pressure from rivals and customer-designed chips.

Analysis

The market is still treating AI capex as a simple NVDA volume story, but the more durable winner is the toll collector on the entire stack. If hyperscaler and sovereign buildouts keep compounding, TSM captures the aggregate wafer-starts, packaging, and node mix regardless of whether the accelerator is NVIDIA silicon or a custom ASIC; that makes its earnings less exposed to share shifts than the headline suggests.

The key second-order effect is that custom chips do not kill demand for compute; they change the economics of who captures the margin. GOOGL is a beneficiary on two fronts: it monetizes internal TPU efficiency while still needing massive data-center buildout, and lower inference cost can expand usage faster than GPU spend decays. NVDA remains the best near-term beneficiary of supply tightness, but over 6-18 months the risk is that the market starts discounting a plateau in pricing power before unit growth slows.

Contrarian view: consensus is probably underestimating how quickly ASIC penetration can compress NVDA's mix and gross margin even if revenues stay high. That argues for viewing NVDA as a high-quality cyclical with a harder comp path after the next 2-3 quarters, while TSM looks like the cleaner structural compounder. The main falsifier is any 2026-2027 capex guide-down from hyperscalers or evidence that AI utilization fails to justify the spending ramp; that would hit TSM less than NVDA, but it would still delay the whole thesis.

Key Decisions for Investors

• Prefer TSM over NVDA on a 6-18 month horizon: long TSM / short NVDA as a hedge against ASIC share gains and NVDA margin normalization; target is relative multiple compression if NVDA growth slows from 'hypergrowth' to merely strong growth.
• Add GOOGL on weakness as a second-order AI infrastructure beneficiary: the stock should benefit if internal TPU economics translate into broader cloud demand, with the trade working over 6-12 months rather than days.
• For NVDA holders, trim into strength or pair against TSM ahead of next hyperscaler capex season; the risk/reward worsens if next earnings merely confirm rather than raise the 2026-27 spending trajectory.
• Set an alert on hyperscaler capex revisions and TSM advanced-packaging commentary: a downward revision would be the cleanest falsifier for the bullish semicapex chain, while any acceleration supports staying long TSM.