Which AI Stocks to Keep — and Which to Dump — as Dalio Cries Bubble
Billionaire investor Ray Dalio is warning the AI trade is nearing a breaking point, with a debt-funded infrastructure boom and climbing interest rates pushing the bubble closer to bursting. Yet even as the S&P 500 and Nasdaq notched fresh closing records this week, investors are being forced to pick winners from losers: a handful of AI names have rallied more than 200% in 2026, but not all are built to hold up if Dalio’s call proves right.
Debt-Funded AI Build-Up Nears Tipping Point
Dalio, founder of Bridgewater Associates, said Wednesday that AI infrastructure is being financed by a massive wave of borrowing, and with rates expected to climb further, the risk of a popping bubble is mounting. In an interview with Bloomberg News on Tuesday, Dalio pointed to heavy borrowing to buy assets — or a potential wealth tax that forces investors to sell — as factors that could turn the tide.
The AI trade pushed the S&P 500 and Nasdaq to new closing records on Tuesday, but the rally has been concentrated in stocks directly benefiting from the boom. Less than half of S&P 500 constituents closed above their 200-day moving average that day, the Wall Street Journal reported, citing Dow Jones Market Data. The record highs also stand in contrast to warning signs in the Treasury market, where a bond selloff has pushed yields to their highest levels in two decades.
Three to Keep, Three to Dump
Against that backdrop, one analysis has sifted through the year’s biggest AI winners and flagged three names to hold — and three to sell.
Keep
- Advanced Micro Devices (AMD) — Riding two powerful trends: AI inference and agentic AI, with a pending acquisition of World Labs poised to position it in physical AI. Large deals with OpenAI, Meta Platforms and Anthropic are set to drive rapid growth, while it remains a leading server CPU supplier benefiting from surging demand tied to agentic AI.
- Marvell Technology (MRVL) — A player in both custom AI chips and optical networking, having helped Amazon develop custom AI accelerators and Microsoft build its own AI chip, and recently winning a deal with Alphabet for components attached to its TPU ecosystem. It also benefits from AI data centers shifting from copper to optical interconnects. At its analyst day, it forecast fiscal 2031 revenue of $70 billion to $90 billion.
- Lumentum (LITE) — One of the few makers of high-power indium phosphide (InP) lasers used to convert electricity into light for high-speed data transmission, and a player in optical circuit switches (OCS) and co-packaged optics (CPO). It holds roughly 60% of the electro-absorption-modulated laser (EML) market, with a deep patent portfolio that gives it staying power once components are certified.
Dump
- Micron Technology (MU) — A beneficiary of the memory price surge, but not a technology leader, trailing South Korea’s SK Hynix and Samsung in advanced memory. The current market is driven by high bandwidth memory (HBM) demand, but Micron is the big DRAM maker least exposed to HBM — it has instead benefited from conventional memory prices rising as peers focus on HBM. Earnings could fall sharply when ordinary DRAM and NAND prices start to decline.
- Sandisk (SNDK) — The S&P 500’s best performer this year, also riding the memory boom, but arguably more of a commodity player than Micron, focused purely on flash memory. The market has lower barriers to entry and more participants, and is likely to return to balance faster than the DRAM market.
- Intel (INTC) — A sharp rebound this year was more about being in the right place at the right time, as server CPU demand took off with the rise of agentic AI and hyperscalers scrambled for chips. But Intel has been losing share in that market — a trend expected to accelerate — while its foundry business continues to lose significant money.
AI
Semiconductors
Telecommunications
U.S. stocks
