Broadcom vs. ON Semiconductor: Which Technology Stock Is a Better Buy in 2026?

Broadcom (AVGO) is positioned as the stronger 2026 pick on earnings/cash flow, with FY2025 revenue up ~23.9% to ~$63.9B, net income ~ $23.1B (net margin ~36.2%), and free cash flow ~$26.9B, but it carries a valuation premium (P/S ~27.2x; forward P/E ~31.5x) and customer concentration (40% from top five). ON Semiconductor (ON) generated far weaker FY2025 results (revenue down ~15.3% to ~$6.0B; net margin ~2.0%; free cash flow ~$1.4B), and is tied ~51% to autos, though it is set to diversify via its planned all-stock ~$7B Synaptics acquisition. Overall, the piece frames Broadcom as higher-quality/less cyclical, while ONsemi is the more recovery/cycle-dependent opportunity with integration risk.

Analysis

This is less a “which chip stock is cheaper” debate than a split between two very different cash-flow regimes. AVGO is the higher-quality compounding asset: its AI and networking mix gives it operating leverage that analog peers cannot match, but the market already prices in continued hyperscaler capex intensity, so upside now depends on re-acceleration rather than normalization. ON is the opposite: a cyclical recovery/turnaround story with balance-sheet flexibility, but its earnings power is hostage to auto and industrial inventory turns; that makes the stock more sensitive to a few quarters of end-market data than to any long-duration AI narrative.

Second-order, the likely relative winner in the semiconductor complex is not the most obvious AI name but the vendor that can attach itself to system-level bottlenecks. If AI infrastructure remains spend-heavy, AVGO’s networking/interconnect franchise should continue to pull share from more generic compute-adjacent suppliers, while TXN and STM face a tougher mix because their analog/power content has less direct exposure to hyperscaler budgets. ON’s edge-AI pivot is strategically sensible, but it is still a story about optionality; until automotive bookings inflect, any M&A-driven repositioning is more likely to compress execution multiples than expand them.

The main risk to the consensus “AVGO wins” view is multiple fragility: at these valuation levels, a modest slowdown in AI orders or a customer capex pause can trigger 15-20% downside without any fundamental collapse. For ON, the reverse risk is that the market is underestimating how long auto/industrial weakness can persist; if that recovery slips into late 2026, the stock can look optically cheap for months while EPS revisions keep drifting lower. The cleanest falsifier for an AVGO bull case is a guide-down tied to top-customer digestion; for ON, it is another quarter of stagnant automotive content and no evidence the Synaptics deal creates incremental growth rather than distraction.

Key Decisions for Investors

• Prefer AVGO over ON as the core long in semis: initiate or add on 3-5% pullbacks, but only if next quarter commentary confirms AI/networking demand remains a 2H-2026 catalyst; upside is slower but more durable than ON’s cyclical re-rating.
• Pair trade: long AVGO / short ON for the next 1-3 months as a quality-vs-cyclical expression. Thesis: AVGO’s earnings visibility should hold while ON remains exposed to auto/industrial downdrafts and integration noise; stop if ON prints a clear sequential revenue inflection or raises guide.
• If you want a cheaper AI-enabled laggard, use SYNA as a watchlist name rather than chasing ON outright. The missing variable is whether the acquisition actually expands addressable content; if it does not, the market may punish deal dilution before any synergy benefit appears.
• Fade TXN/STM versus AVGO on rallies if hyperscaler capex data stays firm. Those names are more vulnerable to a relative-multiple reset if investors keep paying up for AI infrastructure scarcity rather than broad analog exposure.
• Set an alert on AVGO downside below a market-implied ‘AI digestion’ threshold: any 10%+ drawdown on unchanged guidance would be a signal to reduce long exposure, because the stock’s downside skew is driven by multiple compression, not just fundamentals.