AI Power Semis: BofA names TXN, ADI, and ON as top picks for 800V era

Bank of America raised price targets on Texas Instruments to $370 from $320 and on ON Semiconductor to $138 from $115, while highlighting Analog Devices as a top AI power semiconductor pick. TXN’s data center business is cited at about $1.5B in CY25 sales, with potential to reach $4.5B by CY28; ADI’s AI exposure could rise to $3.3B by CY28; and ON’s AI sales may grow to $1.6B by CY28 from a $250M+ run-rate. The note also points to improving China EV trends as a tailwind for ON, with the stock benefiting from 55% share in that market.

Analysis

The first-order read is obvious: the market is rewarding AI exposure, but the second-order issue is broader semiconductor mix re-rating. What matters is that these names are no longer being valued just as cyclical analog/industrial proxies; they are being underwritten as data-center power infrastructure beneficiaries, which can compress the usual duration discount in their multiples. That creates a more durable earnings multiple bridge than a one-quarter AI upside surprise, especially if industrial and auto end markets stop being headwinds and become incremental contributors.

TXN looks like the cleanest secular compounder because its AI-driven content expansion is being layered onto a very large installed revenue base, so even modest share gains in data-center power can move consolidated growth meaningfully. The risk is not demand discovery but estimate ratcheting: once Street models catch up to the content ramp, the stock will need fresh evidence from design wins and conversion rates to sustain the re-rate. In other words, the stock can keep winning even if revenue growth normalizes, but only if investors keep believing the TAM is expanding faster than consensus.

ADI is the highest-quality “hidden leverage” story because the mix is less obviously AI-branded and therefore more likely to remain underappreciated. The market may be missing that power protection/control tends to travel with infrastructure buildouts and has better persistence than pure compute-cycle spend; that should support a longer earnings revision cycle than the AI tape alone implies. ON is the most reflexive and therefore the most vulnerable to disappointment: it has the cleanest operating leverage to both AI and China EV recovery, but that also means any slip in EV demand or digestion in hyperscaler orders can compress the multiple quickly.

Contrarianly, the move may be less about AI fundamentals than about investors rotating into names where the next two years of content gains are easier to model than in software-adjacent AI beneficiaries. The crowd is likely still underestimating how much of this upside is coming from power, protection, and interface silicon rather than headline accelerator demand. If that thesis is right, the best risk/reward is not chasing the most extended name outright, but owning the laggard with the most credible revision runway and hedging the most crowded winner against a broad semis de-risking.

Key Decisions for Investors

• Long TXN on a 3-6 month horizon on any 3-5% pullback; thesis is estimate catch-up plus durable data-center mix shift. Risk/reward: lower beta than peers, but best visibility into multi-year content gains.
• Long ADI versus short a basket of high-multiple AI infrastructure names for a 2-4 month relative-value trade; ADI has less headline AI hype but more persistent power/protection revenue exposure. Hedge the pair with equal dollar weights.
• Buy ON only with defined risk: initiate via 3-6 month call spreads rather than stock, targeting upside from AI and China EV recovery while limiting downside if either theme stalls. Best used after any post-run consolidation.
• Pair trade: long TXN/ADI basket, short a stronger-consensus AI semi beneficiary, to express the view that the market is underpricing content gains but overpaying for crowded AI narratives. Horizon: 1-2 quarters.
• Take profits incrementally if TXN/ON continue to outperform without upward estimate revisions within the next earnings cycle; the trade is vulnerable to multiple compression once the market recognizes the growth is real but not unlimited.