One of the more persistent misconceptions to emerge during the current artificial intelligence cycle is the idea that these technologies must produce winners and losers among the very companies building them. Late last summer and into the fall, a narrative took hold that AI would "eat software" — that intelligent systems would automate away the value of the applications people pay for today. This claim deserves skepticism. Software is one of the largest customers for AI in the first place, so the more reasonable expectation is that both can prosper. They are not rivals so much as two expressions of the same growth story, and the relationship between them is now playing out in the markets in real time.
The Rotation Underway
For much of the past year, the verdict was simple: semiconductors were where the growth lived. Going back to October of last year, the "we love semiconductors" thesis made obvious sense — chips were the picks and shovels of the AI gold rush, and the picks-and-shovels trade is always the easy first call. The problem is that an obvious thesis eventually gets priced in. After a sharp chase higher over recent weeks, semiconductor names have begun to stutter, and within the span of just a few days a notable sell-off rattled investors who had grown accustomed to relentless gains.
That reversal is visible in the numbers. The benchmark semiconductor index, having reached new highs and climbed roughly 130% over a single year, gave back ground sharply. Meanwhile, software — which had carried a premium valuation in the middle of last year only to see that premium evaporate — has quietly outperformed semiconductors month to date. It is still early in the month, but the shift is real and worth watching. The logic behind it is straightforward: at some point, picks and shovels have to turn into actual earnings, and software is the easiest place for AI investment to convert into profit.
Demand Tailwinds Across the Board
None of this means the semiconductor story is broken. If anything, the demand picture remains broad and strong. Tailwinds appear almost everywhere across the hardware stack: analog chips benefit from the buildout of AI servers, memory is buoyed by strength in both NAND and DRAM, and overall AI server demand continues to surprise. Even traditional processor makers have signaled they cannot ship CPUs out the door fast enough. The demand is not concentrated in one corner of the market — it is pervasive, touching nearly every category of component.
The dramatic moves in individual memory and storage names underscore the point. Some of these stocks have risen extraordinarily — one storage name climbed on the order of 3,300% in a single year — and even after pulling back, analysts see meaningful additional upside, with one estimate calling for another 52% gain. That kind of optimism speaks to how convinced the market remains that the underlying buildout is durable.
A Prudent Way to Hold Both
The sensible response to this environment is not to abandon one side for the other but to manage exposure thoughtfully. A reasonable approach is to recognize when a position has outgrown its intended size. Semiconductor holdings that began as a modest portion of a portfolio can double or triple until they suddenly dominate it. Trimming those winners after a big run-up — and redeploying into a variety of areas, software among them — is not a bet against chips. It is risk management. One can continue to own the semiconductor space while declining to add to it at a moment when an aggressive chase has already played out and the options market shows heavy call buying.
Timing matters here. There is little appeal in buying semiconductors immediately after such a chase, but patience changes the calculation. Nothing has happened to suggest the AI build cycle is ending anytime soon, and it is entirely plausible that semiconductors return to new highs later in the year. Software and hardware can both win — simply on different time frames.
Earnings as the Real Test
The coming wave of earnings will reveal how much of this is hype and how much is substance. Several software names report soon, and one of the most instructive is the maker of widely used tax and small-business tools. It reports at the same time as the marquee chip designer, offering investors a software-side reading on the same economy. Its breadth is underappreciated: it owns a dominant consumer tax product reporting in what is seasonally its biggest quarter — roughly half its annual earnings — alongside a small-business accounting platform that doubles as a barometer of how small companies are faring, and a consumer-finance product that reveals the health of everyday borrowers. Taken together, that is a remarkably wide window into the broader economy.
This company has also been beaten down along with much of the software complex, even as it has worked aggressively on agentic AI. Because it is among the first major agentic-AI-focused firms to report, its commentary will be closely scrutinized for evidence that AI is genuinely converting into profit. Other software names tell a mixed technical story — some charts have traveled from high to low even as analysts set price targets well above current levels, in one case more than $100 higher — while at least one communications platform has bucked the trend, climbing steadily to a fresh 52-week high.
Conclusion
The defining mistake of this cycle would be to treat software and semiconductors as a zero-sum trade. They are complementary engines of the same expansion: chips supply the capacity, software monetizes it. Leadership will rotate between them as valuations stretch and reset, and disciplined investors will trim into strength rather than chase it. But the underlying thesis remains intact. The AI buildout shows no sign of exhaustion, and the most important confirmation will come not from price charts but from earnings — the moment when picks and shovels finally show up as profit.