The AI funding landscape has fundamentally shifted in 2026. Just two years ago, investors backed AI companies on team strength and vision alone. Today, with $300 billion flowing into roughly 6,000 startups globally in Q1 2026 alone, venture capitalists are demanding something far more concrete: defensible technology built on proprietary data, deep domain expertise, and a clear path to profitability.

The change reflects a sobering reality. OpenAI raised $122 billion at an $852 billion valuation, Anthropic closed a $30 billion Series G, and France's Mistral AI built a $14 billion empire. Meanwhile, former DeepMind researchers announced Europe's largest-ever seed round in April 2026, raising $1.1 billion for their startup Ineffable Intelligence. The capital is flowing, but the scrutiny has intensified dramatically.

Why Horizontal and Vertical AI Startups Face Different Investor Expectations?

Not all AI companies are evaluated the same way. Socheat Chhay, Managing Director of Sopra Steria Ventures, the corporate venture arm of one of Europe's largest IT services groups, explains that investors separate AI into two distinct categories with vastly different funding criteria.

Horizontal infrastructure companies, like OpenAI and Anthropic, compete on speed and scale. Investors back these firms based on their team's ability to move fast, capture users before competitors do, and maintain customer loyalty. The playbook is straightforward: raise massive capital, grow aggressively, and dominate the market.

Verticalised applied AI, where most European founders compete, operates under completely different rules. Here, investors demand depth, defensibility, and proof that the founder understands the problem better than anyone else in the world. The expectations are higher, and the scrutiny is sharper.

"We separate AI into two main types. One is very horizontal: foundational models and infrastructure, the OpenAIs, the Anthropics. These are the largest deals. Then there's verticalised applied AI, which is AI for specific business use," said Socheat Chhay.

Socheat Chhay, Managing Director at Sopra Steria Ventures

How to Build a Data Moat That Investors Actually Believe In?

The single most important question investors ask vertical AI startups is deceptively simple: what data do you own that nobody else can replicate? This is the difference between a fundable company and a cautionary tale.

Chhay points to a well-funded legal tech startup built largely on top of an existing foundational model without proprietary data or custom algorithms underneath. "They're touching the surface of what legal contracting automation could be. But there's no proprietary custom data. With a new free Claude legal plug-in released in the market, we don't know what's going to happen to them," he explained.

Before pitching to serious investors, founders need to answer three critical questions:

• Data Ownership: What proprietary data does your company own that competitors cannot easily access or replicate?
• Data Acquisition Strategy: How are you systematically acquiring and reinforcing this data advantage over time?
• Competitive Defensibility: Why can't a well-funded competitor simply copy your approach with a new foundational model?

The mistake many founders make is bolting AI onto an existing product and calling it an AI company. "Every founder now sprinkles AI as a powder over their company to enter into AI investment thesis and gain valuation premium," Chhay noted.

"We differentiate between a SaaS company augmented by AI, an AI wrapper and a company with a genuine AI-first moat," explained Socheat Chhay.

Socheat Chhay, Managing Director at Sopra Steria Ventures

A traditional SaaS business with an AI layer isn't automatically disqualified. But the defensibility must come from somewhere: deep workflow integration, high switching costs, proprietary data, or a dominant position in a niche vertical. Founders who've thought carefully about AI's limitations and positioned themselves accordingly stand out to investors.

Which Verticals Are Attracting Serious AI Investment Right Now?

Chhay references a thesis from Sequoia Capital partner Julien Bek that gained significant attention on LinkedIn: the world's next trillion-dollar company won't sell software; it will sell outcomes, using AI to deliver them. Instead of selling legal tech licenses per seat, AI-backed agents working alongside humans will output contract reviews and due diligence. This structural shift is already being confirmed by OpenAI and Anthropic's recent moves to acquire AI services companies.

This framing shapes how Chhay invests. He's targeting verticals where services companies currently handle work that AI can now replace without human intelligence. The sectors attracting the most capital include:

• Manufacturing and Supply Chain: AI agents automating operational workflows that traditionally required human oversight and decision-making.
• Fintech Infrastructure: AI-native solutions replacing traditional financial services workflows and compliance processes.
• Legal and Compliance Automation: KYC (know-your-customer) and AML (anti-money-laundering) workflow automation, plus contract and due diligence review.
• DevOps and Engineering Services: AI replacing human execution in infrastructure management and deployment workflows.

"These are verticalised, highly replicable, and we've reached the stage where agents can replace human execution. That's where we're looking to put our needle," Chhay stated.

Why the Old Startup Playbook No Longer Works for AI Companies?

The economics of AI-native companies have fundamentally changed how investors evaluate teams, valuations, and capital efficiency. A traditional SaaS company needed five to ten years to reach unicorn status, often without profitability at exit. AI-native companies are reaching that milestone in half the time, with smaller teams and strong unit economics.

This shift changes everything about what investors look for. Founding teams no longer need complementary technical skills; those are now commoditised. Instead, investors want ultra-specialists in their domain who are also AI adopters. There's no time for the old iteration cycles that defined the 2010s startup playbook.

"The people building these companies need to be ultra-expert in their specific domain and be AI adopters. There's much less time for the old iteration cycles. That model is gone," noted Socheat Chhay.

Socheat Chhay, Managing Director at Sopra Steria Ventures

Because AI is replacing parts of headcount, from engineering to growth, founders need less capital than a late 2010s Series B playbook would suggest. Investors are now evaluating a new metric: productivity of capital within a compressed timeframe. Founders requesting smaller rounds because they're replacing human tasks with AI tokens are demonstrating they understand the new economics.

What Does Legal's AI Transformation Reveal About Enterprise Adoption?

The legal industry is serving as a proving ground for how AI will transform white-collar work across the economy. Legal AI startups are generating approximately $500 million in aggregate revenue, the second-highest of any industry after coding, according to venture capital analysis.

Less than half of law firms and corporate legal departments are currently using generative AI, according to a Thomson Reuters Institute survey from February 2026, yet adoption is accelerating rapidly. OpenAI invested in Harvey, a legal AI startup, just after ChatGPT launched. Anthropic rolled out lawyer-specific tools for Claude Cowork, with a recent webinar on legal AI attracting 20,000 attendees.

The legal industry's early adoption matters because it demonstrates how AI can integrate into existing workflows while maintaining strict accuracy and privacy standards. If AI can succeed in a highly regulated industry with multiple checks on effectiveness, it signals the technology can work in less stringent sectors.

"Legal is probably the first real professional industry to really move on an enterprise level on AI, besides coding," said Kimberly Tan, an investing partner at Andreessen Horowitz.

Kimberly Tan, Investing Partner at Andreessen Horowitz

Clifford Chance's chief technology officer Paul Greenwood noted that the legal industry's exacting standards have historically made it a proving ground for new technologies. "You get it right for law, you can get it right for everyone. I think that's also partially true with AI as well," he explained.

The broader implication is clear: if an AI company can navigate the legal industry's strict requirements around accuracy, privacy, and regulatory compliance, it has demonstrated the robustness needed to succeed in adjacent professional services like accounting, tax, and consulting.