Prediction: Broadcom and Nvidia Will Be 2027's Best Performing AI Stocks

Nvidia expects 70% revenue growth in 2027, driven by higher AI-hyperscaler spending and its next-generation Rubin architecture, while Broadcom forecasts its AI semiconductor revenue will double in both 2027 and 2028. Despite 2026 share-price gains of roughly 17% for Nvidia and less than 5% for Broadcom, versus about 11% for the S&P 500, the article argues both remain undervalued as earnings have increased. Using a 30x trailing P/E assumption, the author estimates more than 100% upside for Nvidia and more than 50% for Broadcom by the end of the next fiscal year, contingent on forecasts being achieved.

Analysis

The relevant setup is not a generic AI re-rating but a divergence in compute architecture. AVGO’s custom-ASIC and networking exposure can take incremental share of hyperscaler workloads where cost per inference matters more than software portability, while NVDA retains the highest sensitivity to frontier-training demand and accelerated product cycles. This makes AVGO the cleaner beneficiary if AI capital expenditure shifts from initial cluster build-outs toward scale-out networking and inference optimization; TSM and HBM suppliers (TSM, MU, SK Hynix proxy exposure where available) are secondary beneficiaries, but their upside is constrained by capacity allocation rather than end-demand alone.

The market should discount management long-range revenue commentary until hyperscaler capex guidance, purchase obligations, and lead-time data corroborate it. Over the next 1-3 months, the catalyst is calendar-year budget disclosure from MSFT, AMZN, GOOGL and META plus NVDA/AVGO order visibility; the key risk is that capex remains high but shifts toward internally designed silicon, lowering NVDA content per dollar of AI spend. Over 6-18 months, power availability and data-center construction bottlenecks—not chip demand—are the most plausible constraint, creating a risk of order timing slippage and inventory digestion despite intact multiyear demand.

Consensus likely overstates the usefulness of assigning a fixed terminal P/E to both companies. NVDA deserves a higher multiple only if its platform economics preserve gross-margin resilience through the next architecture transition; AVGO’s multiple can expand if custom silicon proves repeatable across several customers rather than concentrated in one program. A broad AI rally is therefore less attractive than a relative-value expression: AVGO offers a more differentiated catalyst, while NVDA requires sustained execution against unusually elevated revenue and supply-chain expectations.

Key Decisions for Investors

• Initiate a 3-6 month long AVGO / short SOXX pair at roughly beta-neutral sizing. The thesis is custom ASIC plus Ethernet/networking content gains versus a semiconductor basket with more cyclical analog, PC and handset exposure; exit if AVGO’s next reported AI backlog or AI revenue run-rate misses management’s prior trajectory.
• Maintain NVDA as a tactical long only into independently verifiable hyperscaler capex updates over the next 1-3 months, rather than underwriting a full-year multiple expansion now. Use a defined-risk call spread 6-9 months out if implied volatility is below its post-earnings range; cut exposure if a major customer signals AI-capex deceleration or if gross-margin guidance falls meaningfully on the next platform ramp.
• Express the architecture-share view through long AVGO / short NVDA only after a relative-strength breakout following earnings, with a 5-8% relative stop. Risk/reward is favorable if custom-silicon revenue visibility improves, but the trade is invalidated by evidence that NVDA’s full-stack software advantage is sustaining dominant inference share.
• Add TSM selectively on confirmation that advanced-packaging capacity expansion remains sold out into the following two quarters. Do not chase memory suppliers solely on AI demand: monitor HBM contract pricing and inventory days, as an easing supply bottleneck could compress margins before end-market AI spending weakens.