For the better part of three years, artificial intelligence startups building software-as-a-service products could do almost no wrong in the eyes of venture capitalists. Valuations soared, term sheets materialized overnight, and founders with little more than a GPT wrapper and a pitch deck found themselves flush with capital. That era is now definitively over.
A growing chorus of investors is publicly and privately signaling that the bar for AI SaaS companies has risen dramatically — and that many of the business models that attracted tens of millions of dollars in funding as recently as 2024 are now considered uninvestable. The shift marks a significant recalibration in how Silicon Valley’s most influential capital allocators think about the intersection of artificial intelligence and enterprise software.
Thin Moats and Thinner Margins: The Core Complaint From VCs
According to a report from TechCrunch, multiple prominent venture capitalists have outlined what they are no longer willing to fund in the AI SaaS category. The overarching theme is defensibility — or rather, the lack of it. Investors say they have grown weary of startups whose primary value proposition amounts to a thin application layer built on top of foundation models from OpenAI, Anthropic, or Google. These companies, often referred to as “wrapper” startups, face an existential problem: the very platforms they depend on can replicate their features with a single product update.
One investor quoted in the TechCrunch piece described the situation bluntly, noting that if your entire product can be replaced by a new feature announcement from OpenAI, you don’t have a company — you have a feature. This sentiment has become pervasive across Sand Hill Road and among growth-stage investors in New York and London alike. The concern isn’t theoretical. Over the past 18 months, OpenAI, Google, and Microsoft have all shipped native capabilities that directly competed with funded startups in areas like document summarization, customer support automation, and code generation.
The Margin Problem That Won’t Go Away
Beyond defensibility, investors are increasingly fixated on the unit economics of AI SaaS businesses. Traditional SaaS companies have long been prized for their high gross margins — typically in the 70% to 85% range. AI-native SaaS companies, however, face a fundamentally different cost structure. Every API call to a large language model carries a compute cost, and for companies processing millions of queries per month, those costs add up quickly.
Several investors told TechCrunch that they are now passing on companies whose gross margins fall below 60%, a threshold that would have been acceptable for hardware businesses but is considered dangerously low for software. The problem is compounded by the fact that many AI SaaS startups have been subsidizing usage to drive adoption, creating an illusion of product-market fit that evaporates the moment pricing reflects true costs. As one venture partner at a top-tier firm put it, the “land and expand” playbook doesn’t work if every expansion seat loses you money.
What Investors Say They Actually Want Now
The investor backlash doesn’t mean capital has stopped flowing into AI entirely. Far from it — global AI investment remains at record levels, with billions being deployed into infrastructure, semiconductors, and vertical-specific applications. What has changed is where within the AI stack investors want to place their bets. According to the TechCrunch report, VCs are now gravitating toward companies that own proprietary data, have built domain-specific models fine-tuned on datasets that are difficult to replicate, or operate in regulated industries where compliance requirements create natural barriers to entry.
Healthcare, legal technology, and financial services have emerged as particularly attractive verticals. In these sectors, the combination of sensitive data handling requirements, industry-specific workflows, and regulatory oversight means that generic AI tools are insufficient. Startups that have spent years building relationships with hospitals, law firms, or banks — and have accumulated proprietary training data in the process — are finding that investor interest has actually intensified, even as their horizontal AI peers struggle to raise follow-on rounds.
The Graveyard of AI Copilots and Assistants
Perhaps no category has fallen further from grace than the AI “copilot” or “assistant” space. In 2023 and 2024, hundreds of startups launched products that promised to be an AI assistant for a specific profession — AI for salespeople, AI for recruiters, AI for marketers. Many raised seed and Series A rounds on the strength of impressive demos and early user traction. But investors now say the category is oversaturated and that most of these products have failed to demonstrate meaningful retention or willingness to pay.
The problem, according to multiple investors, is that these tools often deliver a “wow” moment during the first use but quickly become commoditized as users realize they can get similar results from ChatGPT, Claude, or Gemini directly. Customer churn rates for AI assistant products have been alarmingly high — some investors cited figures above 40% monthly churn for consumer-facing AI tools and 15% to 20% for enterprise versions. Those numbers are incompatible with the kind of recurring revenue growth that justifies venture-scale valuations.
A Broader Correction in Enterprise Software Valuations
The pullback in AI SaaS enthusiasm is occurring against a broader correction in enterprise software valuations. Public SaaS companies have seen their revenue multiples compress significantly from the peaks of 2021, and that repricing has cascaded into private markets. Late-stage AI SaaS companies that raised at 40x or 50x forward revenue are now finding that prospective investors in subsequent rounds are offering terms at 15x to 20x — or declining to participate altogether.
This dynamic has created a growing cohort of “zombie” AI startups: companies that raised substantial capital at inflated valuations but cannot grow into those valuations quickly enough to raise again without a painful down round. Some of these companies have significant runway remaining and continue to operate, but their strategic options are narrowing. Acqui-hires — where a larger company buys a startup primarily for its engineering talent rather than its product — have become an increasingly common exit path.
The Infrastructure Bet Gains Momentum
As application-layer AI companies face skepticism, infrastructure plays are absorbing a disproportionate share of venture dollars. Companies building tools for model evaluation, AI observability, data labeling, and inference optimization have seen strong investor demand. The logic is straightforward: regardless of which AI applications ultimately succeed, all of them will need infrastructure to run, monitor, and improve their models.
This “picks and shovels” thesis is not new — it was the prevailing wisdom during the cloud computing boom of the early 2010s, when investors who backed AWS competitors and DevOps tooling companies generated enormous returns. Whether the analogy holds perfectly for AI remains to be seen, but the capital flows suggest that many of the industry’s most sophisticated investors believe it does.
What Founders Should Take Away From the Shift
For founders building AI SaaS companies, the message from the investment community is clear: differentiation must come from somewhere other than the model itself. Proprietary data, deeply embedded workflows, network effects, and regulatory expertise are the new prerequisites for serious funding. The days of raising a $10 million Series A on the promise of applying GPT-4 to an industry vertical are over.
The correction is painful for many founders who entered the AI gold rush with genuine ambition and technical talent. But seasoned investors argue that this kind of shakeout is both inevitable and healthy. The companies that survive will be those that have built something genuinely difficult to replicate — not just a clever prompt chain on top of someone else’s model. For the venture capital industry, the AI SaaS correction is less a repudiation of artificial intelligence than a return to first principles: sustainable margins, defensible technology, and a clear path to durable competitive advantage.
WebProNews is an iEntry Publication