Memory Chips Get Pricier, Software Downgrades Pile Up: Wall Street's AI Story Is Splitting in Two
As one Wall Street bank slashes price targets on Salesforce, Adobe and Workday while another calls a discounted Oracle a bargain, the AI trade is quietly separating winners who make the picks and shovels from everyone who has to buy them.
Something notable happened in equity research this week that has nothing to do with a single stock and everything to do with how the AI economy is being sorted into haves and have-nots. A major Wall Street bank took an axe to price targets across enterprise software — Salesforce, Adobe, Intuit, Workday — citing a maturing industry facing what amounts to an existential threat: the same artificial intelligence tools these companies sell could eventually make their own products unnecessary. Meanwhile, chipmakers and infrastructure suppliers are getting the opposite treatment, with analysts scrambling to defend memory and semiconductor names even as prices for the physical components rise across the board.
The pattern is worth pausing on, because it cuts against the simple story most people have been told about the AI boom: that it's good for tech, bad for nothing. In reality, the same forces reshaping the economy are creating clear winners and losers within the same sector. Software companies that sell subscriptions to run payroll, manage customer records, or handle accounting are being asked an uncomfortable question — if an AI model can eventually do what your software does, at a fraction of the cost, why does anyone need to license your platform at all? That fear, sometimes called terminal value risk, is showing up in analyst models as lower long-term growth assumptions and shrinking valuation multiples, even for companies with no current signs of customer defection.
At the same time, the physical infrastructure underpinning AI — memory chips, networking gear, cooling systems — is seeing the opposite dynamic. Demand for the components that make AI models run is outstripping supply, which is why memory prices have been climbing all year and why that cost is now working its way into everything from laptops to cloud computing bills. Chip capacity is described by industry analysts as effectively sold out for the next several quarters. That's a very different kind of scarcity than software faces, because you cannot download more silicon.
For the average consumer and small business owner on Long Island, this divergence matters more than it might seem. If AI genuinely erodes the value of enterprise software subscriptions over time, the businesses that pay for that software — accountants, payroll processors, retailers running point-of-sale systems — could eventually see lower costs, assuming competition among AI providers actually passes savings through rather than just replacing one subscription fee with another. But if chip scarcity keeps pushing up the cost of the hardware everything runs on, those savings could get eaten by higher costs for the servers, laptops, and cloud capacity that make any of this software possible in the first place. It's not obvious which effect wins.
There's also a policy angle that deserves more attention than it's getting. The disruption fear rattling software investors is closely tied to open-source AI models emerging from China, which are forcing American AI labs to reconsider their competitive position. Some strategists have floated the idea that U.S. policymakers could move to restrict Chinese open-source models from the American market — not on national security grounds alone, but because they threaten to commoditize AI capability that American companies have spent hundreds of billions of dollars building. If Washington does move in that direction, it would be a rare case of AI policy directly propping up software company valuations, and it's worth watching for anyone trying to understand why a stock's fortune might turn less on its own earnings and more on a decision made in Washington.
None of this should be mistaken for a verdict on any individual company. But the split is real, and it says something uncomfortable about where the easy money in the AI economy is likely to keep flowing: toward the physical bottlenecks — chips, power, cooling — and away from software layers that AI itself might eventually swallow. For anyone trying to read the broader economy rather than pick winners, that's the signal worth carrying forward: this isn't one boom, it's two very different ones happening at the same time, and they don't move together.
You Might Also Be Interested In


