Prediction: 2 Unstoppable Artificial Intelligence (AI) Hardware Leaders That Will Join Alphabet in the $4 Trillion Club by 2028

The article argues TSMC and Broadcom have AI-driven growth catalysts that could push both into a rare $4T market-cap tier by 2028–2029. It cites TSMC demand visibility through 2029–2030 and a further $100B Arizona investment, while projecting 2028 EPS of $24.46 and ~71% upside to an implied ~$733 share price at a 30x P/E. For Broadcom, it highlights custom AI chips with hyperscalers and a forecast of AI semiconductor revenue exceeding $100B in 2027, alongside projected 2028 EPS of $21.28 and ~52% upside at a 30x P/E.

Analysis

The cleaner read is that AI capex is shifting from a pure GPU story to a broader infrastructure stack trade. TSM gets paid first on wafer starts and advanced packaging, but the real second-order winners are the toll-takers with scarce capacity and the customers that can arbitrage lower compute cost into better unit economics; that makes GOOG/GOOGL a stealth beneficiary if custom silicon materially lowers inference costs. The loser is not just NVDA share in custom workloads, but any supplier whose growth depends on hyperscalers keeping merchant-GPU spend at the same mix and urgency.

The market may be underestimating how much of this is already crowded in the valuation, and overestimating the speed at which incremental AI demand converts to FCF. TSM’s capex expansion is bullish on demand, but it also signals that capacity is still the bottleneck, so margins and returns could be less linear if utilization wobbles. For AVGO, the bullish case hinges on a few large design wins becoming a durable annuity; the bearish case is customer concentration and the possibility that 2027 revenue targets look more like revenue substitution than net new demand.

Catalyst path: near term, watch hyperscaler capex commentary and any evidence that ROI discipline is tightening; over 1-3 months, that matters more than the long-dated market-cap narrative. Over 6-18 months, the key falsifier is slower AI infrastructure spend, delayed custom-chip ramps, or TSM execution issues in Arizona/geopolitics. The consensus is too focused on TAM and not enough on who captures margin, so the better trade is relative-value rather than chasing both names outright.

Key Decisions for Investors

• Long AVGO / short NVDA as a 6-12 month pair: thesis is custom ASIC share gain and mix shift away from merchant GPUs; cut the short if NVDA networking/software attach offsets chip share loss.
• Accumulate TSM only on 5-8% pullbacks over the next 1-3 months; the stock benefits from capacity scarcity, but fresh highs likely need another capex inflection before re-rating.
• Add GOOG/GOOGL as an indirect beneficiary of lower AI compute cost; 6-18 month upside is in operating leverage, not headline AI revenue. Falsify if cloud capex or AI inference spending slows.
• Avoid chasing the group with outright longs after a strong AI tape; implied enthusiasm is likely ahead of realized earnings revision, so wait for post-catalyst volatility.