Let’s talk about the money, because in AI, the money is absolutely insane right now.
The Numbers Are Getting Ridiculous
In the first two months of 2026 alone, 17 US-based AI startups closed rounds of $100 million or more. Seventeen. That’s not a typo. Anthropic pulled in a $30 billion Series G at a $380 billion valuation. Nscale raised $2 billion in what became the largest VC round in European history. Ayar Labs, backed by Nvidia, closed $500 million for optical interconnect technology.
We’re not in “AI is promising” territory anymore. We’re in “AI is eating the entire venture capital industry” territory.
Where the Money Is Actually Going
Here’s what’s interesting: the funding isn’t spread evenly across AI. It’s concentrated in a few specific areas, and understanding where the money flows tells you where the industry thinks the real value is.
Infrastructure and compute. This is the biggest bucket by far. Companies building data centers, custom chips, networking hardware, and cloud infrastructure for AI workloads are pulling in the largest rounds. Nscale’s $2B round is infrastructure. Ayar Labs’ $500M is infrastructure. The market has decided that whoever controls the compute layer controls AI.
Foundation models. Anthropic, OpenAI, Mistral, and a handful of others continue to raise massive rounds to train bigger models. But here’s the thing — the number of companies that can realistically compete at the frontier model level is shrinking, not growing. The capital requirements are just too high. You need billions, not millions, to train a competitive frontier model in 2026.
Vertical AI applications. This is where things get more interesting for smaller startups. Companies applying AI to specific industries — healthcare diagnostics, legal document review, financial modeling, drug discovery — are raising healthy Series A and B rounds in the $20-80M range. The thesis is simple: general-purpose AI is a commodity, but domain-specific AI that actually solves real problems is valuable.
AI safety and alignment. A newer category that’s attracting serious money. Investors are betting that as AI gets more powerful, the companies building guardrails, evaluation tools, and safety infrastructure will become essential. It’s a smart bet, especially with regulation tightening globally.
The Uncomfortable Truth About AI Valuations
Here’s where I have to be honest: some of these valuations don’t make sense.
Anthropic at $380 billion? That’s roughly the GDP of Denmark. For a company that, while technically impressive, is still figuring out its business model beyond API access. OpenAI’s valuation has reportedly crossed $300 billion. These numbers assume that AI will capture an enormous share of global economic value, and that these specific companies will be the ones capturing it.
Maybe they will. But history suggests that the companies that dominate a technology wave are often not the ones that raised the most money during the hype phase. Remember, Yahoo raised more money than Google in the early internet era.
The more realistic scenario: most of the $100M+ rounds being raised right now will produce mediocre returns. A few will produce spectacular ones. The problem is that nobody knows which is which yet.
What This Means for Smaller Startups
If you’re building an AI startup and you’re not raising $100M rounds, don’t panic. The mega-rounds are mostly going to infrastructure and foundation model companies. The opportunity for smaller teams is in the application layer.
The playbook that’s working: Pick a specific industry. Find a workflow that’s painful and expensive. Build an AI solution that’s 10x better than the current approach. Raise a focused seed or Series A. Get to revenue fast.
Companies like Simile (AI for modeling human decision-making, $100M Series A) are showing that you don’t need to build a foundation model to raise serious money. You need to solve a real problem.
The playbook that’s failing: “We’re building an AI platform for everything.” Investors have seen enough of these to be skeptical. The generalist AI platform play is increasingly a losing bet unless you have something genuinely differentiated at the model level.
The Geographic Shift
AI funding is no longer a purely Silicon Valley story. Nscale’s record round came from Europe. Middle Eastern sovereign wealth funds are pouring billions into AI infrastructure. China’s AI startups, despite export controls on advanced chips, continue to raise significant rounds domestically.
The US still dominates in absolute dollar terms, but the gap is narrowing. And some of the most interesting AI applications are coming from markets that Silicon Valley tends to ignore — agricultural AI in India, financial AI in Southeast Asia, manufacturing AI in Germany.
My Prediction for the Rest of 2026
The funding pace will slow in the second half of the year. Not because AI is less promising, but because the easy money has been deployed. LPs (the people who fund VCs) are starting to ask harder questions about when these investments will produce returns.
We’ll see more down rounds and flat rounds for companies that raised at peak valuations but haven’t shown proportional revenue growth. The “AI premium” on valuations will compress.
The winners will be companies with real revenue, real customers, and real moats — not just impressive demos and big parameter counts.
The AI startup funding boom isn’t over. But the era of raising $100M on a pitch deck and a prototype? That’s ending fast.
🕒 Last updated: · Originally published: March 12, 2026