In 2025, a US AI startup raised a $1 billion seed round — roughly 1.4 trillion Korean won — before it had shipped a single product. When the deal broke, the press ran headlines calling it the largest seed round ever, and industry newsletters spent days dissecting it. But one group of people took the news in stride: the investors who have spent decades staring at venture return data.
They knew something the headlines didn't capture: startups that raise the largest early-stage checks in history tend to cluster at the bottom of venture portfolio returns. The more buzz a deal generates, the higher the entry price — and the lower the multiple an investor eventually walks away with. It would be easy to write this off as an LP-versus-VC squabble over returns. But look closely at how the pattern actually works, and it maps exactly onto the decisions solo founders and one-person operations in Korea make every day.
Why No One Saw These Numbers Coming — Again and Again
From the second half of 2024 through 2025, early-stage funding for AI startups climbed sharply. Seed rounds that used to average in the single-digit millions of dollars moved into the tens of millions for some AI and biotech companies, and in a handful of cases into the hundreds of millions. Single rounds topping $1 billion were reported several times in the first half of 2025 alone — some of them at companies less than a year old.
Two forces are behind the shift: a change in cost structure, and internal pressure inside the investment industry itself. One driver is the sharp rise in the cost of the GPU clusters needed to train AI models — which fed a narrative, spreading fast among both investors and founders, that competing in this space required massive capital from day one. At the same time, large VCs running funds worth trillions of won face pressure to deploy capital. Writing a handful of big checks is operationally easier than writing dozens of small ones, and a headline-grabbing investment doubles as a branding exercise aimed at LPs. Together, these two forces normalized an equation the market now takes for granted: raise big, grow fast, and you must be a good company.
The trouble is, this equation collides with historical data. Reports from multiple venture data research firms show that companies at the very top of venture return rankings rarely raised enormous sums early on — that pattern is far less common than the current narrative suggests. Instead, it's the companies that started small, confirmed product-market fit first, and only then brought in serious capital that sit much higher on a multiple-adjusted basis.
The Higher the Entry Price, the Lower the Ceiling on Returns
The math behind venture returns is straightforward: profit is the gap between the valuation at entry and the valuation at exit. Sign at a high valuation during the seed round, and even a company that goes on to perform brilliantly caps the multiple an investor can capture. Getting in at 100 and returning 10x looks like the same kind of success as getting in at 1,000 and returning 3x — but they produce entirely different numbers once you run the math across a whole portfolio.
Trace back the deals that top venture history's multiple rankings, and a pattern repeats: founders raised small, raised only what they needed, confirmed there were real paying customers in the market, and only then went out and raised serious money. A significant share of companies that later grew into enterprises worth tens of billions of dollars kept their first outside check well under a few million dollars. They didn't start small because they had no other option — they started small because they understood that doing so bought them better terms later.
So why do investors keep writing giant seed checks when they know the pattern? Because they have to regularly show LPs proof that "our fund is in the room for the defining deals of this era" — and a headline deal is exactly that proof. In a structure where maximizing returns and keeping the fund alive sometimes pull in different directions, big VCs will sometimes choose brand over multiple.
The Case for Starting Big Isn't Entirely Wrong, Either
There's a strong counterargument, of course. In capital-intensive fields like AI infrastructure, the view goes, you simply can't compete without enormous capital from the outset. Building a general-purpose AI model can require tens of millions of dollars just for GPU clusters. In a field where the company that executes the same idea faster and at greater scale wins the market, early capital effectively buys time. By this logic, raising small at the start means never getting to compete at all.
This counterargument holds up in certain domains. Fields like fusion energy, drug discovery, and semiconductor design require foundational research and infrastructure investment before anything else — starting without capital simply isn't possible. Telling founders in these fields to prioritize capital efficiency first can be a prescription that ignores reality.
But most of the mega seed rounds of 2025 happened outside these domains. A more accurate read is that they occurred not in fields that are genuinely capital-intensive, but in fields where the narrative of capital-intensity has taken hold. Looking at the SaaS, marketplace, and media/content companies that actually sit at the top of venture return rankings, cases where a company needed $1 billion at the seed stage are rare. The capital wasn't necessary because the costs were high — the story came first, and the money followed the story.
When Validation Comes First, Capital Carries a Different Weight
What stands out to me is that this structure repeats regardless of scale. Look closely at the decisions solo founders and one-person operations in Korea face, and the exact same logic behind mega seed rounds plays out at a fraction of the size.
The instinct to buy equipment, build out a space, and front-load a marketing budget before the business has proven anything all points in the same direction: spend the capital first, validate later. Whether it's trillions of won going into a seed round or tens of thousands of dollars coming out of a personal bank account, the structure — resources committed before results — is identical.
The data points the other way. Proving what can be proven without capital first, and committing resources only on the strength of that evidence, is the sequence that favors businesses built to last. That sequence gets harder to hold to as the capital gets bigger — investors expect results on a timeline, and meeting that timeline creates pressure to spend quickly. The less capital involved, the less that pressure, and the more room there is to keep the validation sequence intact.
Venture investors in Korea tend to describe the same pattern: their best investments were the ones made when nobody else was paying attention, before a price had even formed. Stories of huge returns coming out of headline deals are rare; far more common are stories of spotting something that looked like a 1 percent chance, in a corner no one else was watching. This observation isn't unique to venture capital. It's the same logic structure that governs how solo operators find opportunities and stake out market space before anyone else has noticed it.
A competitor moving big is no reason to follow the same playbook. A competitor who has taken on large amounts of capital carries pressure proportional to that capital — the urgency to prove metrics fast and show revenue growth scales with the size of the check. A player with less capital can move at a different speed and choose a different order of operations.
The starting point is figuring out whether the belief that your business must start big comes from your actual cost structure, or from a narrative that's simply in circulation. That distinction changes the direction of your early resource allocation.
The deals that drew the most attention ended up at the bottom of the returns table because the price got set before the results did. The ones who spotted an unnamed possibility in a corner nobody was watching ended up in a far more comfortable position later. That sequence didn't look much different whether it played out in a $1 billion seed round or a one-person business that started with a few million won — a few thousand dollars — in capital.
