By Serafina Lalany

 

I used to think of early-stage rounds as modest bets: $2 million here, $5 million there. A handful of hopeful experiments. In 2025, those days feel quaint. Today, it’s not uncommon to see seed rounds of $200–500 million, and even $2 billion “first checks.” It’s as though the venture world decided the only way to play now is with heavy artillery.

These mega-rounds don’t just change the math of capital; they change expectations. A startup that raises $250 million on Day One isn’t expected to move fast; it’s expected to deliver the extraordinary. That warping of expectations matters because it will shape which founders can survive under the pressure, and which investors have the patience to stay in the game.

In dissecting year-to-date early stage deals across five top venture funds (Andreessen Horowitz, Sequoia, Founders Fund, General Catalyst, and Greylock) , a few clear patterns emerge, and they aren’t obvious. We’ll see not just which AI bets are winning, but why they’re winning, who’s making them, and where.

The New Terrain: Mega Rounds at Early Stages

From our sample of those funds’ deals:

• The median early-stage deal is roughly $15–20 million, varying by investor and geography.
• The median post-money valuation has lifted accordingly, often landing in the $120–180 million range for many “hot” AI or platform plays.
• But those medians disguise the tail-end bets that dominate headlines: deals valued in the high hundreds of millions or even billions are now appearing at the seed/Series A stage.

This isn’t just a shifting baseline. It changes risk profiles, hiring cadence, and the narrative pitch to the market. Founders now compete not to raise hundreds of thousands, but to justify why they deserve hundreds of millions on day one.

What Kind of AI Gets Mega Capital? A Deeper Look

Labeling something simply as “AI” isn’t enough. The data reveals several sub-theses capturing the most investor attention in 2025:

Why infrastructure leads: Investors are placing their biggest bets on the scaffolding, the models, tooling, and compute layers, because success there amplifies every downstream AI use case. You might not see an immediate end-user product, but every other AI application will depend on that foundation.

Why robotics and physical AI are resurging: As AI capabilities scale, attention is turning to making it act in the real world. It’s one thing to generate text or images; it’s another to make a robot do something useful. For instance, a startup in the Midwest is building vision models for industrial automation, combining local manufacturing know-how with frontier AI. These hybrid approaches require heavy capital but promise deep moats in sectors that haven’t been fully automated yet.

Why biotech AI gets special treatment: Drug discovery is famously expensive and risky. conventional venture often avoids it. But when an AI biotech pitch is strong, a few big funds will break protocol and fund it with “insane conviction” because the upside (new drugs, multi-billion-dollar blockbusters) is massive. In our sample, one biotech AI company quietly accumulating over $1 billion in backing illustrates that even in today’s AI rush, investors will pour money into hard science domains if the potential payoff is big enough.

Why vertical AI matters: Generic, one-size-fits-all AI will soon saturate the market. The most defensible plays are now in customizing AI for specific industries with high domain friction. Logistics, legal tech, energy grid optimization– these all get attention. Founders in the middle of America often have direct access to domains like manufacturing, agriculture, or energy. That’s a latent advantage many coastal founders lack, and investors are noticing.

Lessons from the Outliers

Some early-stage deals are so far out on the high end, they almost defy belief. Consider a few extreme cases that illuminate how far the boundaries have stretched:

• Thinking Machines Lab – a stealthy AI startup founded by former OpenAI CTO Mira Murati that raised about $2 billion in its initial round with no product or revenue. This bet was made entirely on the strength of the team and a long-term vision of building safer, more versatile AI.
• Slate Auto – an Indiana-based EV maker that secured roughly $700 million (across its Series A and B, backed by Jeff Bezos’s fund and others) to build a low-cost, no-frills electric pickup truck. Slate’s truck is intentionally stripped-down—crank windows, basic knobs, no touchscreens—a stark contrast to the software-laden EVs from Tesla or Rivian.
• Somnia – a blockchain infrastructure project (backed by Improbable’s venture arm) that lined up $270 million to develop a high-speed Layer-1 network for on-chain gaming and finance applications. The goal is to reimagine blockchain scalability for things like real-time games and DeFi by processing hundreds of thousands of transactions per second.

These outliers stretch the definition of what’s “fundable.” They show that when the vision is huge, investors will write the check. In each case, the founders’ ambition matches the capital behind them. These companies aren’t just lucky, they’re aiming to redefine their respective sectors.

Heartland Bets: Big Visions, Big Capital

It’s not just coastal startups getting big money. A number of notable early-stage companies in 2025 hail from outside the usual tech hubs. The examples below aren’t token regional plays, each has attracted serious funding and is going after a big problem

What jumps out from these heartland companies is how often they sit at the intersection of multiple domains (energy, health, manufacturing, etc.) where deep domain expertise is critical. Their locations give them close proximity to the problems and customers they’re addressing. For example, the nuclear startup in Oak Ridge benefits from being near national labs and a talent pool of nuclear engineers.

Importantly, the capital behind these startups isn’t small. These are full-throttle ventures, not just symbolic “flyover country” investments. In other words, the heartland is no longer a sideshow in tech. It’s becoming a bona fide proving ground – one of the primary deployment zones for innovations in hard industries like energy and advanced manufacturing (and the AI that powers them).

Coast vs. Interior: Capital’s Footprint

Even with the mega-round fervor, most capital still concentrates on the coasts but not exclusively. In our dataset, about 19% of early-stage deals (and roughly 18–20% of total dollars) went to startups based outside the traditional coastal tech hubs. In other words, roughly one-fifth of the venture action is now happening off the coasts. That’s not a majority by any means, but it’s enough to signal a meaningful broadening of venture geography, compared to a decade ago when Silicon Valley and NYC might have dominated nearly everything

What This All Means (for Founders, VCs, and the Heartland)

• Higher baseline expectations. If you pitch an AI or frontier-tech startup today, you’re not competing against notional “good ideas” in a vacuum – you’re competing against labs that raised nine-figure seed rounds with no product. The bar is higher. To stand out, you’ll need to demonstrate conviction, domain depth, or some unfair advantage that justifies serious capital. The era of the $3 million seed for a great team and a slide deck is largely behind us in certain sectors.
• “AI for X” is the winning playbook. In 2025, a generic AI demo isn’t enough. The startups commanding the biggest rounds are AI + something specific – AI + healthcare, AI + supply chain, AI + you-name-it. Linearity in AI is gone; to multiply your upside, you couple AI with a domain where incumbents are clumsy or some real-world friction exists. Venture capital is pouring money into these combinations. A startup that remains a generic AI platform with no vertical focus will struggle to raise anything beyond a modest round now.
• The heartland has a latent advantage. The places that make our steel, grow our food, treat patients, and produce energy aren’t just markets for tech, they can be testbeds. If you’re building an AI solution for manufacturing, being in Michigan or Ohio gives you daily access to factories and engineers in that field. Real-world feedback loops form faster. Partnerships come easier. Industry credibility builds sooner. As more VCs invest in these regions, innovation becomes more distributed. The next great AI company might truly be built in a Michigan, Ohio, or Tennessee. Not just used there.
• A call to alignment. There’s a tempting narrative in chasing whatever’s hot (be it massive AI labs or flashy consumer apps), but that shouldn’t be our only north star. Building transformative technology in the heartland isn’t just a feel-good story; it’s strategic. If you believe AI will revolutionize core industries, it makes sense to build it near the bones of those industries. That’s where the rich data lives, where the veteran domain experts are, and where early adopters can collaborate closely.

In the next few years, we’ll learn which of these bold bets actually survive and justify their valuations. But right now, we have a choice: either follow venture capital’s gravitational pull to ever-bigger, centralized bets, or help push that momentum outward. I believe the heartland is ready for its moment– to not just adopt technology, but to build it. And that’s something we should all be rooting for.