Why this $6.5-billion money manager sold Accenture to buy Nvidia and Microsoft

Portfolio manager Daneshvar Rohinton says 'a lot of quality companies are on sale' and has been buying Visa, Microsoft and Nvidia while selling Accenture. He argues Visa is at a 10-year valuation low, Microsoft remains the default enterprise platform with about 20% expected earnings growth over the next three years, and Nvidia is still well positioned to benefit from AI-driven GPU shortages. The article is primarily a stock-picking interview, but it underscores continued investor interest in AI and high-quality dividend-paying technology franchises.

Analysis

The important signal here is not simply that these are high-quality franchises, but that the market is temporarily re-rating “duration of demand” in the wrong direction. Payment rails and office software are being marked as if AI-native workflows immediately disintermediate incumbents, yet the nearer-term effect is more likely monetization migration: the incumbents absorb more volume, then re-price, bundle, or route it through adjacent products. That creates a window where secular winners with near-monopoly distribution trade at multiples more typical of ex-growth utilities than compounding platforms.

Visa is the cleanest expression of this mispricing. Even if some agent-to-agent transactions bypass cards, the larger second-order effect is that more commerce shifts online and into machine-readable, authenticated flows where trusted networks, fraud controls, and global acceptance become more valuable, not less. The real risk is not volume collapse; it is mix pressure and a slower take-rate expansion path over the next 12-24 months. If that concern fades, multiple re-expansion could do more for returns than incremental growth.

Microsoft’s setup is similar but with a different catalyst profile: capital intensity is the visible headache, while the hidden benefit is that the company can subsidize and distribute AI at scale better than any peer. The market is underestimating how aggressively product bundling can flatten standalone competition in collaboration and workflow software over the next 6-18 months. Accenture is the clearest loser because the value chain is being compressed upward into platform owners and downward into narrower implementation work; consulting leverage is being squeezed from both ends.

The contrarian view is that investors are treating AI as a binary threat to incumbent software and payment infrastructure, when the more likely transition is a redistribution of economics rather than destruction of demand. Nvidia remains the purest beneficiary of that redistribution, but the stock is increasingly a supply-constrained earnings compounding story rather than a simple multiple expansion story. That makes the best setup a barbell: own the platform winners being sold on disruption fears, fund it with shorts in services models whose relevance erodes as AI standardizes deployment.

Key Decisions for Investors

• Go long V on a 3-6 month horizon; add on weakness as the market overprices disintermediation risk. Base case is mid-teens upside from multiple normalization, with downside limited unless alternative rails show real payment share gains.
• Go long MSFT vs. short a basket of legacy workflow/point-solution software over 6-12 months. The risk/reward favors MSFT because distribution and bundling can offset AI capex, while smaller software names face margin compression and slower seat expansion.
• Short ACN on a 6-9 month horizon or pair it against MSFT. The thesis is structural margin pressure as AI reduces billable implementation hours faster than consulting firms can repackage services; upside for the short if enterprise buyers shift spend from services to platforms.
• Maintain/add to NVDA on pullbacks, but size for event risk over the next 1-2 quarters. This is still a fundamentals-led long, but the asymmetry is lower than in V/MSFT because expectations are already high; prefer buying only on weakness rather than chasing strength.
• If expressing the theme more efficiently, use a pair trade: long V or MSFT / short ACN for 6 months. That isolates the winner-from-distribution effect while reducing market beta and avoids needing a broad AI tape to work.