Texas Instruments posted its strongest quarter in years on July 22, with revenue up 23% year-over-year and EPS beating by more than ten percent, yet TXN shares dropped roughly 4% in after-hours trading. The market's indifference is itself the story.
The numbers were unambiguous. Revenue came in at $5.46 billion against analyst expectations of around $5.24 billion, EPS hit $2.14 versus a consensus of $1.95, gross margin held at 61%, and management guided Q3 revenue to a range of $5.65 billion to $6.15 billion, comfortably above what the Street had penciled in. Industrial revenue grew 30% year over year. Data center revenue doubled. Automotive came in at mid-teens growth. As Benzinga noted, TXN delivered a double beat and raised guidance, and the stock still slipped. That's not a bad quarter. That's a verdict on how investors are thinking about the chip sector right now.
Texas Instruments is the world's largest analog chipmaker, and analog is not glamorous. It doesn't make the GPUs that run inference workloads or the high-bandwidth memory that sits next to them. What TI makes are the power management chips, signal converters, and embedded processors that every electronic system needs to function: the unglamorous plumbing of AI infrastructure rather than the headline component. Data center revenue doubling sounds impressive until you realize TI is capturing that growth one power rail at a time, not one Blackwell rack at a time. The market knows the difference.
Nvidia's stock moves on AI narrative as much as earnings because its chips sit at the center of the most capital-intensive buildout in recent memory. Every hyperscaler capacity announcement, every sovereign AI fund, every new model training run feeds directly into GPU demand. TI benefits from that same buildout, but one step removed. You can't run a data center without TI's power chips. You just can't talk about them at a cocktail party. That adjacency discount is real, and the after-earnings selloff made it visible.
There's also a valuation problem that a strong quarter can't paper over. As GuruFocus reported, TXN was trading at a P/E of around 50x heading into the print, a premium that historically doesn't hold for a company whose end markets include industrial and automotive, both of which remain cyclical. Even a 23% revenue pop doesn't feel like it justifies that multiple when the analog cycle can turn as quickly as it turned down in 2022 and 2023. Insiders haven't been buying into the optimism either: GuruFocus noted roughly $88.9 million in insider share sales over the three months leading up to the report.
Then there's depreciation. TI has been building out domestic wafer manufacturing at a pace that's impressive on paper and expensive in practice. The Sherman, Texas fab and other capacity expansion projects carry a projected $350 million step-up in depreciation for 2026, according to data from the earnings call. That's a real drag on free cash flow even as operating profit climbed 48% to $2.31 billion. The company is building for a future where it controls its own manufacturing destiny, but investors are paying that construction bill today.
On the call, TI also disclosed it has started rolling out customer-by-customer price increases, with more impact expected in Q4 and into 2027. That's a smart move for margins longer term, but it introduces uncertainty about whether customers absorb those increases or start qualifying competing suppliers. Frankly, it's one of those disclosures that sounds like good news and gets processed as a risk.
What the selloff tells chip-sector investors #
The real signal here isn't that TI had a bad quarter. It didn't. The signal is that capital in the semiconductor sector is concentrating around the names with direct, irreplaceable exposure to AI compute: Nvidia, Broadcom to some extent, and a handful of memory suppliers. Everything else, even companies posting the kind of numbers TI just reported, is getting measured against that benchmark and found to be insufficiently exciting. Industrial recovery, automotive stabilization, power management demand from data centers: all good stories, none of them as good as
Also read: OpenAI's own AI models broke out of a test sandbox and autonomously hacked Hugging Face • Hyundai's 35,000 striking workers just forced the first real test of who controls humanoid robots on the factory floor • Substack's new AI scanner tells paying readers exactly how much of their newsletter a human actually wrote