The AI Gold Rush Meets a Hard Budget Ceiling
The pullback in AI spending across SaaS companies is a market correction in which executives are slashing budgets, restructuring teams, and narrowing model portfolios after discovering that experimental AI initiatives bring heavy costs and uncertain enterprise ROI, forcing a shift from hype-led expansion to disciplined, selective investment. This is not a retreat from AI; it is a retreat from the fantasy that every workflow needs a frontier model and every software product needs an AI sticker. Corporate buyers have figured out they can get acceptable performance from cheaper systems, and the vendors that overbuilt for an endless AI bonanza are now paying the price. The correction is painful, but overdue, and it will separate AI projects that solve clear business problems from those that only looked good on earnings calls.
Monday.com: SaaS Layoffs in Service of an AI-First Story
Monday.com’s decision to lay off around 630 employees—about 20% of its workforce—is a blunt example of software company restructuring driven by AI overreach. The company is cutting deep to concentrate resources on its AI Work Platform, aiming for a “leaner, more focused operating model” centered on AI development. Earlier this year it redesigned its product lineup around the idea that enterprises want AI agents working alongside staff, bundling a no-code app builder, customizable AI agent, workflow automation, and a chatbot that can generate reports and update dashboards. In practice, this is a bet that a tighter AI story will please investors more than a broader product footprint. But when more than 122,000 tech roles have already been eliminated in 2026, with many cuts tied to AI-related restructuring, Monday.com’s move looks less like bold innovation and more like joining a painful pattern.
The company expects USD 45–55 million (approx. RM207–RM253 million) in restructuring charges, a number that underscores how expensive the AI pivot has become even before considering ongoing development costs. The uncomfortable truth is that many SaaS layoffs AI decisions are masking an earlier miscalculation: hiring and building for a level of AI demand that never fully materialised. Instead of a steady stream of high-margin, AI-powered upsells, vendors are facing customers who question incremental gains and count every GPU-hour against their budget.
Amazon’s Nova Retreat: Frontier Bets Over Big Portfolios
Amazon’s overhaul of its AI plans—winding down most in-house Nova models and focusing on a single frontier effort—is another sign that the era of sprawling, unfocused AI portfolios is ending. The company is deprecating prominent Nova offerings such as Premier, Omni, the Reel video generator, and the Canvas image generator, keeping them in a maintenance state for existing customers but ceasing serious development. Resources are being redirected to a frontier-model project led by Pieter Abbeel, with a debut expected at the re:Invent conference later this year. This is the sober logic of scarcity: engineers and compute are limited, and spreading them across overlapping models is a luxury Amazon no longer thinks it can afford. As one report notes, labs are now competing on the economics of daily use rather than raw capability alone, which aligns with buyers who care more about predictable spend than benchmark scores.
It would be easy to misread this as Amazon losing faith in AI; the reality is the opposite. The company remains a major investor in external model makers and a key cloud and training partner, choosing to concentrate on infrastructure and one frontier push while letting others run the broader model race. In a sense, Amazon is acknowledging that enterprise AI ROI is about matching capability to use case, not offering every conceivable model. The AI spending pullback is forcing large vendors to decide which bets they truly believe in—and to shutter the rest.
From Tokenmaxxing to Thrift-Maxxing: ROI Finally Matters
The real story behind these moves is that corporate spending on AI is slowing because finance teams have started calling bluff on unsustainable burn rates. Companies across the United States have come to the “radical idea” that they do not have to blow their budgets on AI, and many are now mixing lower-priced models—including some built in China—alongside offerings from OpenAI and Anthropic. As one executive put it, using the most powerful and expensive AI model for mundane tasks is like “driving a Lamborghini to go to the grocery store to pick up milk.” That blunt reality is eroding the case for limitless AI experimentation. US firms have flipped from “tokenmaxxing” to “thrift-maxxing,” blending cheaper options with premium models, which threatens sky-high valuations for model labs built on assumptions of constant volume growth.
This shift exposes a core misalignment of the first AI wave: vendors optimised for maximum consumption, while customers now optimise for minimum necessary capability. Enterprise AI ROI is no longer measured by how futuristic a demo looks, but by whether a given system delivers more value than it consumes in compute and integration costs. For SaaS providers, the AI spending pullback translates directly into tighter scrutiny of features, pricing, and staffing. Software company restructuring around AI is less about embracing the future and more about surviving a present in which buyers want proof, not promises.
The Correction Will Hurt Now—and Help Later
The current wave of SaaS layoffs AI, portfolio cuts, and budget discipline is painful and will likely intensify before it stabilises. Yet it also marks the moment AI becomes a normal enterprise technology rather than a mystical growth engine. By shedding overbuilt teams, deprecating little-used models, and switching from tokenmaxxing to thrift-maxxing, the industry is starting to respect the basic arithmetic of enterprise software: costs must be justified, and features must earn their keep. The coming winners will be those who treat AI as a tool, not a story—matching model choice to task, prioritising reliability and clear savings over spectacle. The losers will be the firms that still believe hype can outrun spreadsheets.
This correction is more than belt-tightening; it is a reset of expectations. Software company restructuring that once read as breathless AI evangelism now looks like hard-nosed triage. In the long run, that is good for customers and for serious AI builders. When the dust settles, the projects that survive will be the ones that prove their value, line by line, on the budget sheet.
