These 3 AI Stocks Are Way Off Their Highs. Is the Pullback a Buying Opportunity?

September 6, 2026 - 09:42

Vertiv, Applied Optoelectronics, and Innodata all started September trading far below their 2026 peaks, even though their underlying businesses continue to post strong growth numbers. For investors watching these names, the big question is whether the recent sell-off is a genuine entry point or a warning sign. The truth is that these three companies are not interchangeable bargains, and each one carries a different set of risks.

Vertiv Holdings sells power and cooling equipment for data centers. Its second quarter sales were up 24 percent, driven by demand for high-density computing infrastructure. Applied Optoelectronics makes optical transceivers, which are critical for connecting servers inside AI clusters. The company has seen a surge in orders from major cloud providers. Innodata provides data engineering and model evaluation services, essentially helping companies clean and test the data used to train large language models.

The pullbacks have been steep. Vertiv is down roughly 40 percent from its high. Applied Optoelectronics has fallen even further, and Innodata has given back a large portion of its massive 2025 rally. The reasons for the declines vary. Some of it is profit taking after huge runs. Some of it is concern about customer concentration, especially for Applied Optoelectronics, which relies on a small number of buyers. There is also broader worry that AI capital spending could slow if the economy weakens.

None of these companies are cheap in the traditional sense. Their valuations still assume significant future growth. That means the market is pricing in execution, not just survival. If any of them stumble on delivery timelines or lose a key customer, the downside could be substantial. For investors with a long horizon and a tolerance for volatility, the current levels might offer a reasonable risk-reward. But for anyone looking for a safe way to play AI infrastructure, these stocks are not that. They are high-beta bets on a boom that is still real but increasingly crowded.