Broadcom vs. Micron: Which Semiconductor Stock Is the Better AI Buy over the Next 3 Years?

Micron reported fiscal Q4 2026 revenue of $54.23B, up 379% year over year and above the $51.07B consensus, while adjusted EPS of $33.42 exceeded the $31.61 estimate; it guided current-quarter revenue to $61.5B and EPS to $38.15, both above consensus. Broadcom expects AI revenue to rise 186% to $58B in fiscal 2026, then reach $115B in 2027 and $230B in 2028, supporting projected EPS above $30 in fiscal 2028. The article favors Micron over Broadcom based on its lower valuation, faster anticipated growth, and continued tight AI-memory supply conditions.

Analysis

The reported Micron figures require independent verification before any capital deployment: the stated quarterly revenue and margin levels are not economically plausible relative to Micron's historical scale, creating a high risk that decimal placement, period labeling, or sourced estimates are wrong. The investable signal is directionally valid but narrower: incremental HBM and high-density DRAM tightness can drive disproportionate EPS upside because memory pricing flows through a largely fixed-cost manufacturing base. That operating leverage is also the principal risk—MU should be valued on normalized mid-cycle earnings and inventory discipline, not peak-cycle spot pricing; a weakening in contract DRAM/NAND pricing would compress the multiple before reported earnings roll over.

AVGO's upside depends less on aggregate AI spending than on whether custom ASIC programs move from design wins to recurring volume ramps without impairing its networking attach rate or gross margin. Successful hyperscaler ASIC deployments are a competitive headwind for NVDA's accelerator share, but also expand AI cluster demand for AVGO switches, optics, and connectivity; the more relevant pair is AVGO versus NVDA rather than a standalone AI-beta trade. Over the next 1-3 months, earnings revisions and customer capex commentary are catalysts; over 6-18 months, the key question is whether customer concentration converts AVGO's projected AI pipeline into durable revenue or exposes it to a small number of delayed internal-chip programs.

Consensus may be underestimating the cyclical asymmetry in MU and overestimating the linearity of AVGO's custom-chip ramp. MU can outperform in the early phase of a memory upcycle even if AI unit growth moderates, but it is the weaker hold once capacity additions and customer inventory rebuilding normalize. AVGO offers higher earnings durability through diversified infrastructure content, yet its premium can de-rate sharply if a major ASIC customer changes architecture, delays a tape-out, or brings more design capability in-house.

Key Decisions for Investors

• Do not act on the cited MU financial figures until the earnings release, 10-K, and consensus data are reconciled; set an alert for verified HBM bit-growth, DRAM contract-price changes, and gross-margin guidance. A downward revision to next-two-quarter gross margin or DRAM pricing would falsify the near-term long thesis.
• For a 1-3 month tactical expression, favor long MU / short SOXX in modest size after verified guidance confirms HBM allocation and contract-price strength. This isolates memory-cycle acceleration from broad AI multiple risk; exit if MU underperforms SOXX by 10% following results or if management signals incremental capacity materially ahead of demand.
• For a 6-18 month structural position, accumulate AVGO on post-earnings volatility rather than chase AI-revenue projections. Pair long AVGO with a smaller short NVDA only if disclosed custom-ASIC volume commitments or customer capex plans validate share substitution; unwind if NVDA retains accelerator share while AVGO's AI networking growth decelerates.
• Monitor GOOG and META capex guidance as leading indicators for AVGO's customer concentration risk. A synchronized reduction in 2027 AI infrastructure spending would impair AVGO's revenue conversion before it is visible in reported results and would argue for reducing both AVGO and MU exposure.