20.4% of Berkshire Hathaway's $306 Billion Portfolio Is Invested in 3 Artificial Intelligence (AI) Stocks

Berkshire Hathaway's $306 billion public-equity portfolio includes three AI-levered holdings: Apple (18.4% portfolio weighting; remaining stake worth >$56B after selling ~75%), Alphabet (1.8%), and Amazon (0.2% after selling ~77%). AWS generated $128.7B in 2025 revenue and carries a $244B order backlog; Amazon also developed Trainium2 chips with ~40% better price/performance. The piece is analytical rather than news-driven, highlighting structural AI tailwinds across cloud, devices, and search that could benefit these large, long-held positions.

Analysis

AI adoption is reshaping economics across three layers: hyperscaler infrastructure, silicon suppliers, and downstream software monetization. The non-linear part is in the supply chain—power, cooling, interconnects and wafer capacity—which create bottlenecks that concentrate pricing power among a handful of suppliers and can sustain gross-margin upside for those with scale even if unit demand softens.

A key second-order effect is demand reallocation within large-cap tech: capital intensity of AI favors vertically integrated incumbents that can capture both hardware and recurring software/service revenue, pressuring pure-play hardware vendors and low-margin retailing/fulfillment businesses. This bifurcation magnifies index concentration risk and raises the probability that outperformance will be concentrated in a smaller cohort of names over the next 12–36 months.

Risks cluster around three catalysts that could reverse the current drift: rapid GPU/accelerator oversupply (6–18 months) driving price deflation, a sharper-than-expected enterprise pause in AI projects (quarters), and regulatory actions that compress ad/search monetization windows (12–24 months). Each of these would compress multiples quickly because much of the current valuation premium is forward-looking and tied to multi-year service monetization assumptions.

The consensus misses asymmetric on-device monetization and service bundling as a low-friction way to convert installed bases into SaaS-like revenue, which is underappreciated in valuations for large consumer OEMs. Conversely, the market may be over-pricing perpetual hardware scarcity; if foundry capacity scale-up accelerates, the high-multiple hardware names suffer faster multiple compression than the software beneficiaries.

Key Decisions for Investors

• Long NVDA call spread (6–9 month, 20–30% OTM buy/sell) to play continued accelerator tightness; allocate 2–3% portfolio, target asymmetric 2:1 upside vs premium at risk. Use tight stops at 50% premium loss given earnings/capacity announcements can gap prices.
• Pair trade: long NVDA / short INTC equal notional (3–12 months) to express structural share shift in AI silicon and execution diferenças; size short smaller (0.5–1x) due to execution/ turnaround risk. Reward: captures valuation rerating; Risk: Intel execution surprises or large foundry investments re-rate quickly.
• Long GOOGL (buy 9–18 month LEAPS or 12 month call ladder) to play AI-driven ad/search monetization and enterprise deployments; position size 1–2% of portfolio. Hedge with short-dated put protection if you enter post-earnings to limit drawdown on short-term ad volatility.
• Buy protective puts on AAPL or sell covered calls (6–12 months) depending on portfolio exposure—if long existing AAPL exposure, sell calls to monetize time premium; if adding exposure, prefer call spreads to limit premium burn. Objective: capture on-device AI upside while protecting against a macro or regulatory shock that compresses multiples.