Nvidia Will Soon Make More than Apple and Microsoft Combined

Bank of America’s Vivek Arya says Nvidia could generate more than $400 billion in free cash flow over the next two years, yet the stock still trades at under 20x forward earnings and under 20x EV/FCF, a roughly 50% discount to Magnificent 7 peers. Arya argues Nvidia could raise shareholder returns via buybacks and a dividend increase, potentially lifting yield from 0.2% to 0.5%-1.0% at a cost of $26 billion to $51 billion. The article is constructive on Nvidia’s long-term fundamentals, though it flags rising competition and the challenge of sustaining valuation premium.

Analysis

The market is treating AI capex as a hardware story, but the more durable edge is turning into platform lock-in and cash conversion. If customers are standardizing model training and inference around a single software stack, the economic moat shifts from chip performance to switching costs, which tends to support higher terminal margins and less cyclical demand than the sell-side is currently embedding. That matters because the next leg of upside is less about unit growth and more about whether capital returns become the proof point that this is not a transient buildout.

The second-order winner is the ecosystem around AI deployment: networking, power, and data-center infrastructure should continue to compound even if chip growth moderates, because installed AI fleets create ongoing spend on interconnect, cooling, memory, and power management. The biggest competitive pressure is likely to show up first in gross-margin mix, not unit shipments, as custom silicon and lower-cost accelerators win narrow workloads while the incumbent keeps the high-value training and enterprise software layer. That suggests the market may be overestimating the speed of share loss but underestimating the pace of price competition in inference over the next 12-24 months.

The key risk is not demand disappearing; it is multiple compression if capex growth normalizes before free cash flow distribution ramps. If management does not materially increase repurchases or dividends within the next 1-2 quarters, investors may start to question whether excess cash is being hoarded for optionality rather than returned, which can cap rerating even in a strong fundamental tape. A secondary risk is that mega-cap ownership saturation limits incremental fund flows, meaning the stock may need a visible capital-return catalyst, not just another earnings beat, to sustain outperformance.

The contrarian takeaway is that the market may be too focused on whether the incumbent remains the best chip vendor and not focused enough on whether it becomes the best capital-return story in semis. If cash flow stays on track, the valuation gap versus slower-growing megacaps looks hard to justify, particularly on EV/FCF. But if free cash flow is even modestly delayed by customer concentration or export/supply constraints, the stock could de-rate quickly because expectations are now anchored to perfection.

Key Decisions for Investors

• Long NVDA vs short AMD over the next 3-6 months: the spread should favor the incumbent if software lock-in keeps pricing power intact; stop if AMD wins visible hyperscaler share or NVDA guidance implies margin erosion.
• Add a call spread in NVDA with 6-9 month tenor to express rerating from capital-return surprise rather than pure earnings growth; upside is driven by dividend/buyback expansion, downside limited to premium paid.
• Long NVDA / short a basket of custom-ASIC enablers that monetize substitution risk indirectly, including AVGO on any strength, if management commentary signals inference workloads shifting to lower-cost alternatives.
• Pair long NVDA with short MSFT only if the market starts rewarding capital efficiency over absolute AI spend; the trade works best if NVDA announces a materially larger buyback authorization within 1-2 quarters.
• Watch for a tactical entry after the next earnings print: if the stock sells off on 'capex normalization' headlines while cash-return language improves, use the weakness to build a 6-12 month long.