Texas Instruments just did something Wall Street usually hates and income investors should love. In its Q1 2026 results, trailing free cash flow jumped 154% to $4.35 billion while trailing capital expenditures fell to $4.1 billion from $4.7 billion a year earlier. That is not an AI bubble headline. That is what happens when the spending binge starts giving way to cash extraction.

The market is still acting as if the only AI trade worth discussing is whichever company can spend the most, fastest, with the fewest adult supervisors in the room. Fine. Let momentum traders fight over that circus. For dividend investors, the interesting part of the 2026 semiconductor story is the opposite: which companies no longer need to spend like maniacs to stay relevant.

The Part of the AI Trade That Actually Pays You

AI spending has not collapsed. Anyone telling you that is either lazy or trying to sell a dramatic macro chart. What has changed is the market's tolerance for endless capital intensity. Hyperscalers still need compute, networking, and power. We covered the infrastructure side of that in The AI Power Crunch and the model-efficiency angle in The DeepSeek Shock. But inside semiconductors, 2026 is starting to separate two groups:

• Companies still trapped in a capex arms race
• Companies finally harvesting the returns from capex they already made, or avoiding heavy fab spend entirely

That distinction matters because dividend safety is not a story about intentions. It is a story about how much cash is left after management pays for the next factory, the next equipment cycle, and the next strategic panic attack.

Why Lower CapEx Can Be Bullish for Yields

When capital expenditures come down and revenue merely holds together, free cash flow can snap higher with almost indecent speed. That is the part growth investors keep missing. A semiconductor company does not need explosive top-line growth to produce a strong dividend year. It needs:

• Decent demand
• Stable gross margins
• Less money disappearing into fabs and equipment

That is exactly what Texas Instruments is showing. In Q1 2026, management said revenue rose 19% year over year, led by industrial and data center. More important for income investors, quarterly capex dropped to $676 million from $1.123 billion a year earlier. On a trailing-12-month basis, capex fell to $4.1 billion from $4.7 billion, while free cash flow jumped to $4.35 billion from $1.72 billion.

That is the post-hype AI dividend setup in one line: less incremental spending, more cash left for owners.

Texas Instruments also declared a new quarterly dividend of $1.42 per share. The company returned $6.0 billion to owners over the trailing 12 months. That is not management promising to become shareholder-friendly someday. That is management already writing the checks.

A Historical Reminder Nobody Enjoys

This pattern is not new. After every semiconductor expansion phase, the best dividend stories usually emerge from the companies that stop playing offense with their balance sheet and start acting like landlords. The spending cycle crests, utilization stabilizes, and suddenly the same stock that looked "boring" during the boom becomes a cash-return machine during the cleanup.

The smartphone buildout produced versions of this. The auto-chip shortage produced versions of this. AI is not special enough to repeal corporate finance.

Three Semiconductor Names That Fit the 2026 Income Thesis

Texas Instruments: The CapEx Comedown Is the Whole Point

Texas Instruments is the cleanest version of the thesis because you can see the handoff directly in the numbers. For years, the company leaned into 300mm manufacturing and broader internal capacity. That spending was defensible. It was also a nuisance if you were buying the stock for near-term cash return.

Now the math is starting to look much better. TI's latest results show:

• Trailing 12-month cash flow from operations of $7.8 billion
• Trailing 12-month free cash flow of $4.35 billion
• Trailing 12-month capex of $4.1 billion, down from $4.7 billion
• Trailing 12-month dividends paid of $5.05 billion

No, this is not a screaming current-yield story. It is better. It is a dividend growth story tied to improving cash conversion.

If you are the kind of investor who obsesses over whether every AI dollar goes to the obvious winners, TXN will probably look too dull. That is exactly why it is interesting. It sells into industrial, automotive, and increasingly data-center power and edge applications. You are not paying for a superhero narrative. You are getting a company whose AI relevance sits one layer below the hype and one layer above the payout.

Broadcom: AI Revenue Without the Fab Headache

Broadcom is not "legacy" in the old analog sense, but it belongs in this discussion because it monetizes AI demand without carrying foundry-style capex baggage. In its fiscal 2025 results, Broadcom reported:

• Q4 revenue of $18.0 billion, up 28%
• AI semiconductor revenue up 74% year over year
• Q4 free cash flow of $7.47 billion on just $237 million of capex
• Full-year free cash flow of $26.9 billion
• A 10% dividend increase, taking the annual target to $2.60 per share
• A 15th consecutive annual dividend increase

Hock Tan said AI semiconductor revenue should double year over year to $8.2 billion in Q1 2026, driven by custom accelerators and Ethernet AI switches. That is the part the market loves. The part income investors should love more is that Broadcom is doing it with a capital base that looks positively civilized compared with any company trying to build cutting-edge fabs itself.

This is the stealth appeal of the name. Broadcom gets paid on the networking and custom silicon architecture that keeps AI clusters useful. Someone else gets the joy of writing the gigantic manufacturing checks.

Qualcomm: AI at the Edge, Dividends in the Mail

Qualcomm is still treated by too many investors like a smartphone orphan. That is stale thinking. In its fiscal 2025 results, Qualcomm said full-year revenue reached $44.3 billion, with record QCT revenue. Management highlighted:

• 18% year-over-year growth in total QCT non-Apple revenue
• 27% growth in combined Automotive and IoT revenue
• Expansion into data centers and advanced robotics

That is not the profile of a company waiting around for handset replacement cycles to save it.

The dividend side is just as relevant. Qualcomm increased its quarterly payout from $0.85 to $0.89 per share in March 2025, raising the annualized payout to $3.56. Cristiano Amon also reiterated a balanced capital-return policy with anti-dilutive buybacks as the baseline.

The bull case here is not that Qualcomm becomes Nvidia. That fantasy belongs on social media, preferably with all the other adult supervision failures. The case is that edge AI, automotive, and low-power intelligent computing are large enough growth vectors to support continued dividend growth without forcing the company into a capex war it cannot win or should not want.

The Name That Does Not Fit the Thesis

Intel is useful here for one reason: it shows what happens when capital intensity overwhelms the income story.

In August 2024, Intel told employees it would suspend its dividend, cut capital expenditures by more than 20% in 2024, and shift toward capital efficiency and more normalized spending levels. As of April 26, 2026, Intel lists no upcoming dividend. The same investor page shows the last quarterly dividend was $0.125, with no future payout presently declared.

That does not make Intel uninvestable. It does make it a poor example of the "post-hype AI dividend play."

If you want a blunt rule, here it is: do not confuse strategic importance with dividend safety. Governments may want Intel fabs. Analysts may want Intel to recover. None of that forces the cash to appear on schedule for shareholders.

If you bought Intel for yield and Texas Instruments for discipline, you already learned the difference the hard way.

What Actually Matters for Dividend Safety From Here

If you want to separate real semiconductor income plays from yield cosplay, watch three things over the next week or two:

• Capex direction, not just revenue guidance. If TXN keeps stepping down quarterly capex while industrial and data-center demand remain constructive, the dividend-growth case gets stronger.
• AI revenue mix without balance-sheet abuse. Broadcom can keep looking excellent as long as AI networking demand grows faster than any need for heavy incremental investment.
• Non-handset diversification at Qualcomm. Automotive, IoT, and edge AI need to keep growing fast enough that investors stop treating the dividend as hostage to smartphones.

And one more thing, because somebody has to say it plainly: if you are shopping for "AI dividend stocks" by current yield alone, you are setting yourself up to own the wrong names. We already covered that trap in our guide to spotting dividend cuts and our broader dividend stock analysis framework. In semiconductors, yield without free cash flow discipline is usually just a warning label wearing makeup.

The next useful datapoint is not another breathless headline about trillion-dollar AI spending. It is the next batch of earnings calls where management teams quietly admit how much less they need to spend to keep participating. When those numbers come out, which would you rather own: the company still begging the market for patience, or the one that can afford to keep raising the payout?