Billionaire Chase Coleman Just Increased His Stake in These 3 Top Semiconductor Stocks. Are They Still Buys?

Tiger Global's Chase Coleman added to Nvidia, Broadcom, and Taiwan Semiconductor in Q1, with these three names ranking among his top 10 holdings. The article argues all three are major AI infrastructure beneficiaries: Nvidia remains the LLM training leader, Broadcom is a custom ASIC/TPU and data center networking winner, and TSMC is the key foundry that benefits regardless of whether hyperscalers use GPUs or ASICs. The piece is bullish on their long-term AI-driven growth, but it is primarily investor commentary rather than new company-specific financial news.

Analysis

The market is increasingly pricing AI as a three-layer stack: compute architecture, custom silicon design, and foundry capacity. That is structurally bullish for the infrastructure complex, but the second-order winner is the least visible one: TSMC’s bottleneck power means every dollar of capex shifted from one chip architecture to another still routes back through the same scarce manufacturing node. In practice, that makes competitive intensity in GPUs vs ASICs less important than expected; the more hyperscalers diversify, the more aggregate wafer demand and packaging demand accrue to the same chokepoints.

Broadcom’s setup is the most underappreciated because its growth is less about one product cycle and more about becoming the outsourced silicon arm of the hyperscalers. The key risk is not demand but concentration: once a few customers validate custom inference economics, adoption can scale fast, but revenue visibility can also be lumpy if a single hyperscaler pauses design wins or digestion kicks in. Over the next 6-18 months, the relevant catalyst is not headline AI spend but evidence that custom chips are taking share in inference workloads where power efficiency matters most.

Nvidia remains the cleanest training leader, but the market may be underestimating how much of the next leg of growth has to come from non-training attach rates—networking, CPUs, and server-level integration—because pure GPU mix alone is unlikely to sustain prior growth rates indefinitely. That creates a subtle margin risk if the company has to defend share with broader platform pricing while customers simultaneously develop more bespoke solutions. The contrarian view is that the ecosystem is not winner-take-all at the chip level; it is winner-take-most at the platform and manufacturing layers, which is why relative outperformance may shift from NVDA to AVGO/TSM as the cycle matures.

Key Decisions for Investors

• Stay long TSMC into AI capex revisions over the next 3-6 months; best risk/reward is that it monetizes every architecture path. Use pullbacks to add, since the downside is limited by structural capacity scarcity.
• Rotate a portion of NVDA exposure into AVGO on strength: AVGO has a better medium-term payoff if inference custom silicon adoption accelerates over the next 12-18 months, while NVDA is more exposed to sentiment swings around training growth deceleration.
• Pair trade: long TSM / short a basket of smaller semiconductor equipment or foundry laggards if you expect AI wafers to remain concentrated in the highest-end process nodes. The thesis is share gains plus pricing power over 6-12 months.
• For event-driven exposure, consider AVGO Jan-2026 calls financed by selling higher-strike calls; upside is convex if hyperscaler custom-chip rollouts surprise, while premium selling helps if digestion delays the next leg.
• Tactically reduce chase-buying in NVDA after sharp AI headline rallies; the trade works best on pullbacks unless there is fresh evidence that networking/CPU attach is accelerating materially.