The 30-Year Software Pricing Limit Has Been Reached
There’s a key piece of information that isn’t mentioned in any press release but explains everything: when Microsoft’s CEO states that 30% of the company’s code is now generated by a machine, he’s not announcing a productivity achievement. He’s admitting, without saying it outright, that the scarcity that justified decades of premium pricing has just evaporated.
This is what’s happening in the sector that Wall Street refers to as Big Software, and the market valuation loss of Oracle—over $463 billion since September 2025—is not an anomaly. It’s the first visible signal of a revaluation that isn’t finished yet.
The Moat Was Not Technology, It Was Difficulty of Construction
For thirty years, the central argument of any enterprise software company was simple: building what we create requires hundreds of engineers, tens of millions of dollars, and years of iteration. That entry cost was the protective moat. Not the brand, not the technology itself, but the enormous friction involved in replicating it.
Generative artificial intelligence doesn’t compete with that software. Instead, it directly attacks the friction that protected it. When open-source models reach 90% of the capabilities of cutting-edge models, the cost of building a functional competitor collapses. A startup that previously needed five years and fifty engineers to build a minimally viable CRM can now cover that ground in a fraction of the time and capital. The moat didn’t disappear because someone built a better bridge; it disappeared because it dried up.
This radically changes what companies are buying when they hire a platform like Salesforce or a Microsoft cloud service. For decades, they were contracting execution certainty: the assurance that someone had already solved the problem of building the software, tested it with thousands of clients, and would keep it functioning. That argument is still partially valid, but its price is adjusting downward at a speed that financial analysts’ models did not anticipate.
When Productivity Becomes a Justification for Cuts
The most revealing movement of internal tension within these companies is not the investment in artificial intelligence. It’s the series of decisions that followed.
In May 2025, Microsoft announced a round of layoffs in which approximately 40% of the cuts—over 800 positions—affected software engineers at its Redmond headquarters. Oracle and Block did the same, explicitly citing automation as justification. The corporate narrative framed these moves as a reallocation toward higher growth areas. But the financial mechanics are more direct: if 30% of the code is now produced by a machine, maintaining the same number of human engineers undermines the return on investment argument in AI presented to the board of directors.
An operational paradox exists within this reasoning as the data begins to indicate. Code generation tools produce more bugs than code written by experienced engineers. This means that the speed gain comes with an increased cost of supervision and correction. Companies are sacrificing in-depth review for speed of delivery, and that will have consequences that won’t show up on financial statements for the next few quarters but will emerge in end-customer satisfaction.
The race to prove that investment in AI is generating returns—Microsoft committed $10 billion to OpenAI, Google invested $2 billion in Anthropic, Amazon another $4 billion—is creating internal pressure that distorts product priorities. Projects without an associated AI narrative are being canceled or deprioritized regardless of their value to the customer.
The Value Shift That Balances Have Not Yet Shown
The most useful question for an executive evaluating their exposure to this sector is not whether Salesforce or Microsoft will survive. They probably will, at least in some form. The more pertinent question is: where is the value that these platforms once captured shifting to?
Competitive architecture analysis points in a consistent direction: value is moving from software to physical infrastructure. Data centers, semiconductors, energy capacity. Companies that control those assets—not those selling software licenses—are consolidating their positions. Hyperscalers with their own infrastructure have a defense that SaaS pure players cannot replicate quickly.
Oracle’s case illustrates this shift with painful clarity. The company tried to position itself as a provider of infrastructure for the largest language models on the market. When OpenAI decided to redirect its capacity towards Microsoft and Amazon, Oracle lost that contract and, with it, the narrative that sustained its valuation. Over $463 billion in market capitalization evaporated in months. Not due to a scandal, not due to a poor accounting decision, but because the market recalibrated how solid Oracle’s position was in the AI value chain.
For executives evaluating which enterprise software providers will remain relevant in the next five years, the question is no longer whether they have AI functionality. It’s whether the AI layer they offer creates real dependency or is simply a wrapper around models that the customer could access directly. The difference between those two situations will determine which companies can maintain their margins and which will enter a downward spiral of price reductions to retain customers.
The Work Customers Were Always Hiring Wasn’t Software
There’s an underlying reading in all this that analyst reports tend to dodge because it complicates valuation models.
Companies that purchased Salesforce or Microsoft Dynamics, or any long-cycle enterprise software platform weren’t buying technology. They were purchasing the elimination of operational uncertainty: the certainty that their critical processes would work, that there would be support when something failed, that the provider would be there in three years. That’s what justified multi-year contracts and premium pricing.
Artificial intelligence does not eliminate that need. What it does eliminate is the perception that only established consolidated platforms can meet that need. When an internal team can build and maintain a functional CRM tool at a fraction of the previous cost, the argument of the consolidated provider ceases to be the only route to that operational certainty.
Enterprise software companies that understand this before their competitors will not compete on AI functionalities. They will redefine what specific certainty they are selling and to whom, and they will build their pricing model around that specific certainty, rather than around the technical complexity they can no longer protect.
The failure of high-priced models in this cycle demonstrates that the work that customers were always hiring wasn’t access to sophisticated software, but the reduction of operational risk in relying on technology they don’t control. Whoever can credibly sell that guarantee in an environment where building software is no longer the barrier will have solved the real problem. The rest will keep lowering prices until margins can no longer justify operations.
