Understanding where demand is forming is one of the more actionable things a domain investor can do, and funding events are among the better forward signals available. They are public, they are timestamped, and each one tells you two things: which industries are gaining momentum, and which naming constructions companies in those industries are actually using. Funding rarely coincides with naming, since most companies are already operating on something by the time they announce. What it shows is where the next concentration of buyers, and the pressure to improve their names, is forming.

That is why we started tracking funding events at Atom, initially as a seller feature: alert sellers when a newly funded company matches a name in their inventory. That remains the primary use. What we did not expect was how much the same data would show in aggregate. So we took 1,000 rounds, mostly announced between April and July 2026, representing $186.8 billion in disclosed capital, and read them the way a domain investor would.

The finding worth reading for is at the end: two names can be equally attractive and have completely different odds of finding a buyer, and the data tells you which is which.

What the data shows

• 72% of funded companies chose an outright brandable name. One out of 715 owns a category keyword domain
• 59% of verified domains are .com
• 14% are .ai, now the second startup extension by a wide margin
• 36% of companies do not operate on a literal exact match of their name
• 47% → 66% .com share from seed to Series B and later

1. Volume demand and value demand are different businesses

The fourteen largest verticals, covering 871 of the 1,000 rounds. The remainder is 71 untagged companies plus nine smaller verticals, none above ten rounds.

Industry	Rounds	Total Raised	Median Round
AI	209	$105.5B	$19.0M
Other/unclassified	144	$39.9B	$10.0M
Fintech	87	$7.6B	$16.8M
Robotics	23	$4.9B	$32.0M
Foodtech	33	$4.8B	$8.0M
Healthcare	86	$4.1B	$11.5M
Biotech	56	$3.5B	$8.4M
Crypto	22	$2.7B	$52.5M
SaaS	103	$1.9B	$5.2M
Spacetech	17	$1.9B	$10.0M
Hardware	19	$1.6B	$23.2M
Climate	31	$1.2B	$19.5M
E-commerce	19	$1.1B	$8.7M
Cybersecurity	22	$0.8B	$25.0M

AI took 21% of rounds and 56% of capital. Everyone knows this. The useful part is the two right-hand columns, which describe different kinds of demand.

AI, SaaS, fintech and healthcare are half of all rounds: constant naming events at modest round sizes. Crypto runs 22 rounds at a $52.5 million median, robotics 23 at $32 million, cybersecurity 22 at $25 million. Thin markets, better capitalized companies. You can see it in who shows up. The four biggest crypto rounds are Kraken on kraken.com, Crypto.com, Polymarket on polymarket.com and Digital Asset on digitalasset.com, every one on a clean expensive .com. In SaaS, at a $5.2 million median, companies concatenate their way onto whatever was available.

Round size is not naming budget. But breadth pays in the volume verticals because the naming never stops, and depth pays in the concentrated ones, where you may wait a year for an inquiry and then find yourself talking to someone who just raised $50 million. Also worth noting: 36% of rounds were pre-seed or seed, the cohort likeliest to still be deciding.

2. Real words carry a premium

Naming approach	Companies	Share	Median Round
Invented brandable	438	44%	$10.0M
Brand + modifier (Pearl Health, Catena Labs)	209	21%	$12.0M
Evocative real word	169	17%	$20.0M
Compound brandable	113	11%	$11.0M
Multi-word descriptive	71	7%	$6.0M
Pure category name	0	0%	n/a

Companies with a single real-word brand raised roughly twice the median of everyone else, and the gap survives removing outliers. Causality runs both directions: a team that can raise $20 million can also afford a real-word .com. That is less an objection than a description. Real-word .coms are what well-capitalized companies buy once money stops being the constraint.

The winners look like Cursor, Alan, Nourish, Current, Farther, Hark, Redo. Ordinary words, mostly four to six letters, none describing the product. Cursor is a coding tool, Nourish is healthcare, Current is a bank. The word does no explanatory work, which is the point.

The more instructive cases are the companies whose brand and domain do not match. Latent raised $80 million and calls itself Latent, but its website is latenthealth.com. Candid raised $105 million and is on candidhealth.com. Dream raised $260 million and sits on dreamgroup.com. In each case the company introduces itself with one word and then has to give out a longer address. Latent.com, candid.com and dream.com are all owned by someone else.

That mismatch, between the brand a company picked and the domain it ended up with, is the demand this market runs on. Nobody in that position is thrilled about it. Whether they ever do anything about it is a separate question, but the wanting is real and it is visible in the data.

Sixty percent of companies use a single-word brand with no modifier. About 31% end in a vowel, led by -a (69) and -o (61). The -ly and -ify pattern appears in only 7 of 1,000 names. Not proof the style is dead, since those domains still trade, but it has clearly fallen out of favor with founders raising institutional money. If you hold suffix inventory, your buyer is more likely a bootstrapped operator than a funded startup.

Takeaway: the premium sits on short “open-vessel” words that mean nothing in context. Words a company can pour meaning into.

3. Why .ai gets used and .health doesn’t

Forty percent of these companies use a multi-word name, almost always a brand paired with a modifier. Which modifier they choose is one of the more useful things in the dataset, because the modifier is the part of the name that has to go somewhere when they buy a domain.

Modifier	Companies	Median Round	Survives into the domain
AI	26	$8.7M	14%
Health	18	$16.0M	93%
Labs	14	$14.0M	82%
Therapeutics	14	$6.5M	38%
Energy	11	$6.6M	29%
Technologies	10	$15.0M	62%
Bio	9	$5.6M	25%
Robotics	8	$32.0M	83%
Medical	7	$12.0M	40%
Security	7	$13.0M	43%

The right-hand column is the interesting one. When a company is called Pearl Health, “health” ends up in the domain too, at pearlhealth.com. Same with Labs and Robotics. The modifier gets concatenated because there is nowhere else for it to go.

AI is the exception. Of the 26 companies branded “<Word> AI,” only 14% keep “ai” in the second-level domain. The rest move it into the extension: Venice AI on venice.ai, Luffy AI on luffy.ai, Rightbrain AI on rightbrain.ai. This is not about which extensions exist. Matching gTLDs are available for most of these modifiers, and a handful of companies here use them.

Modifier	Companies with domains	On the matching TLD
AI	22	19
Bio	4	3
Energy	7	2
Security	7	2
Health	14	1
Labs, Therapeutics, Technologies, Robotics, Medical	38	0

Health is the sharpest case. Fourteen companies branded themselves “<Something> Health,” .health has existed for years, and one used it. So the variable is legitimacy rather than availability. A company on venice.ai reads as owning its name. A company on pearl.health would not, yet, which is why pearlhealth.com wins. One group ends up with a two-word domain; the other with what reads as a one-word domain.

Whether other extensions follow .ai is open, and we would not bet on it. .ai had an unusual combination: a historic boom in its category, two characters, and an abbreviation everyone already reads as a word. “Venice AI” and “venice.ai” are the same string to the eye. “Pearl Health” and “pearl.health” are not, quite.

Takeaway: short .ai is the one alternative extension currently doing a .com’s job. Everything else in the modifier list is still a .com game.

Which compounds have end users

This has a direct consequence for two-word inventory, and the split is sharper than we expected. Counting trailing modifiers across all 1,000 companies:

Used: health (18), therapeutics (14), energy (11), bio (9), robotics (8), medical (7), security (7), aerospace (4), care (4)

Not used once: crypto, fintech, payments, pay, money, bank, software, data, app

Not one company branded itself “<Something> Crypto” or “<Something> Payments.” The pattern carries into the domains, meaning the end of the second-level domain rather than the extension. Eighteen companies operate on something like pearlhealth.com, six on a domain ending in “bio,” six in “robotics,” three in “security.” None on the equivalent ending in “crypto” or “fintech.”

So a two-word .com ending in health, bio, security or robotics has demonstrable end users. The equivalent ending in crypto or fintech has none here, not because those industries are small but because founders in them do not name companies that way. Fintech and crypto brands are single words or invented, and when they add a modifier it is usually Labs or Technologies rather than the category.

Then there is “pay”, which is where this dataset shows its limits. No company in these 1,000 rounds branded itself “<Something> Pay”, so on the funding data alone you would conclude pay names have thin demand. That conclusion would be wrong. On Atom, “pay” is one of our better-selling roots and it moves consistently. We see those sales because they happen on our own marketplace, and they are not showing up here at all.

Both things are true. Venture-funded startups are not naming themselves FooPay, but the buyers for pay domains are payment processors, regional fintechs, bootstrapped operators and companies outside the US venture cycle, and none of them appear in a funding feed.

Treat that as a warning label on everything else here. Funding data is a good signal about one buyer pool and silent about the others, and silence is not absence.

Takeaway: before buying a two-word compound, ask which buyer pool it serves. Health, bio, robotics and security compounds are in active use by funded startups. Pay compounds sell too, but to buyers this dataset cannot see. Be careful with the construction where neither is true.

4. Startups are not buying category domains. One company is.

Across 715 verified domains, we checked every second-level domain against the standalone commercial keywords domain investors think of as category assets: the insurance.com, booking.com, hotels.com tier. Words that name a market rather than describe an activity.

Exactly one company came back. Crypto.com, which raised $400 million this quarter.

The story behind it is the point. Crypto.com was not founded as Crypto.com. It was Monaco, a Hong Kong crypto card company. In 2018 it bought crypto.com from Matt Blaze, a Penn cryptography professor who had registered the domain in 1993 and spent years publicly refusing to sell, for a reported $12 million. Then Monaco renamed the entire company to match.

That is not a startup picking a name. It is an operator making a strategic acquisition and reorganizing its identity around it, eight years before the round we captured.

Which is the lesson. Category .coms sell for real money, just not to companies at the naming stage. They sell to operators consolidating a position, to companies rebranding at scale, to international entrants buying credibility. Those buyers do not appear in a funding feed, and when one does it looks like this: the purchase happened years earlier, unrelated to raising a round. Two markets, side by side, on very different clocks.

Everyone else is on a brandable. Only 102 of 715 companies hold a domain that is even a single dictionary word, and 25 of those are single-word .coms: Anthropic, Cursor, Alan, Nourish, Current, Farther, Hark, Redo.

There is a reason beyond price. A category name tells the market what you do, useful at launch and constraining afterward, and most founders raising in 2026 are pitching a company that intends to outgrow its first product. Genius AI is the example in reverse: GlossGenius was legible in beauty and wellness, and when the company wanted the broader service economy, the name had to change.

One clarification, because we conflated these ourselves on a first pass. Describing what you do is not the same as owning a category. Bricklaying Robotics is descriptive and bricklayingrobotics.com is worth very little. Crypto.com is worth eight figures.

The brand-plus-modifier structure concentrates in spacetech (47%), biotech (45%), healthcare and cybersecurity (41%), and is rarest in fintech (10%) and AI (19%). Brandable share holds near 89% at seed and 82% at Series B and later, so it is not a habit that gets outgrown.

Takeaway: brandable demand is set by companies at formation, a large renewing pool. Category demand is set by a much smaller set of strategic buyers on a longer timeline. Neither is better, but confusing them is how people end up holding the right asset with the wrong expectations about when it moves.

5. The .com graduation effect, and a visible exception at the top

Stage	.com share	.ai share
Pre-seed / Seed	47%	20%
Series B and later	66%	10%

The .com share climbs nineteen points between seed and Series B, while .ai’s roughly halves.

This reframes an argument our industry has had for years about whether .com still matters to founders who grew up on .io. The question was probably never preference. It is affordability and timing: early on an alternative extension is a rational trade, and later, when hiring and selling to enterprises, the calculus shifts. One caution, this is a cross-section rather than a longitudinal study. We are comparing different companies at different stages, not the same companies over time.

At the top of the market the pattern breaks. Prometheus is valued at $41 billion on prometheus.ai. Fireworks raised $1.5 billion at a $17.5 billion valuation on fireworks.ai. DeepSeek, Mistral and Helsing all fly .ai. Companies on .ai account for more than $31 billion of capital here.

The strongest evidence is Genius AI: a decade-old business doing nearly $200 million in revenue that rebranded from GlossGenius onto genius.ai this year at a $1.15 billion valuation. Not a seed-stage company taking what it can get.

So the read is two-sided. The broad market still trends toward .com with scale. The AI frontier increasingly treats short .ai as a destination rather than a waypoint.

TLD	Count	Share	Median Round
.com	424	59%	$16.0M
.ai	97	14%	$17.5M
.io	33	5%	$23.2M
.in	17	2%	$6.0M
.org	16	2%	$4.3M
.co	15	2%	$16.0M

Two details. .io carries the highest median round of any major extension at $23.2 million on only 5% share, still the developer-infrastructure signal. And .ai holders run shorter brands than .com holders, a median of 6.5 characters against 9.0.

Takeaway: a seed-stage company on .ai or .io is a plausible .com buyer in a couple of years, and that pipeline is measurable. Do not assume it applies at the top. The companies with the most money are the least likely to move.

6. A third of companies are on something other than their name

Two definitions, because they give different answers. A literal match means the second-level domain spells the full company name (pearlhealth.com for Pearl Health). A semantic match means the domain spells it across the dot (venice.ai for Venice AI).

TLD	Literal	Semantic
.com	64%	64%
.ai	64%	82%
.io	73%	79%
.in	82%	82%
.co	94%	94%

Read literally, .ai and .com are identical. Read semantically, .ai leads by eighteen points. That gap is the whole point of the extension: .ai companies are not better at getting their exact name, they are using an extension that carries part of the name for them.

Overall, 36% of companies do not operate on a literal match of their name. They append a word: labs (11), ai (8), group (5), plus technologies, pay, app, health. Relay is on relayfi.com, Catena Labs on catenalabs.com. They prepend one: get (5), use (3), then try, my, meet, go. Or they move extensions: Convey raised $38 million from a16z and sits on convey.dev, Dapple closed $30 million on dapple.co, Sable raised $45 million from Sequoia and 8VC and operates on withsable.com, which does both. Another 171 carry a compound name with a strippable modifier.

We cannot see whether these companies want a different domain, and some are probably happy. What we can say is structural: a third of recently funded startups are on something other than their name, and that share renews every cycle.

Takeaway: the compromise rate is the closest thing this data has to a demand floor. Not a list of buyers, but evidence that the wanting does not go away.

7. Roots, and which ones are actually attached to money

Roots are how most investors think about inventory, so a frequency count on its own is not much use. Below is every root appearing in at least eight of the 1,000 brands, with what those companies raised and how many are still at seed.

Root	Companies	Median round	Seed share	Read
bio	31	$7.4M	29%	Ubiquitous in biotech, but the smallest checks on the list
ai	28	$13.6M	43%	Still being adopted heavily by brand-new companies
health	24	$21.0M	38%	High volume and high median. The strongest combination here
ver-	22	$17.0M	32%	Trust and verification. Quietly well funded
ther-	22	$4.8M	32%	Common but cheap. Therapeutics rounds are small and early
gen	21	$22.0M	19%	Genetics and generative at once. Mostly later-stage companies
energy	14	$10.0M	36%
robo-	12	$36.0M	42%	Highest median on the list, and still forming
ora	12	$7.7M	17%	A classic brandable root, now mostly in mature companies
space	10	$28.7M	30%	Few companies, large rounds
arc	9	$4.9M	56%	The newest root here. Over half still at seed
nex	8	$4.4M	38%	High usage, low funding. Crowded rather than valuable

The seed share column is the one to study. High means the root is still being picked by brand-new companies, which is where naming demand starts. Low means the companies using it named themselves years ago.

Forming: arc (56% seed), ai (43%), robo- (42%), health (38%). Robotics is the standout, pairing a high seed share with the largest median round in the study. New companies, still naming, with money.

Established: ora (17%), gen (19%), bio (29%). Widely used, but mostly by companies that named themselves a while back.

Crowded: nex is the caution. Eight companies, $4.4 million median, the lowest on the list. A root can be popular precisely because it is generic enough that everyone reaches for it, which is what makes it hard to sell at a premium. The same applies to ther- and bio, both very common and both near the bottom on round size. Biotech names literally and raises modest early rounds: plenty of buyers, limited budgets.

Takeaway: frequency tells you a root is in use, median round tells you whether the users can pay, seed share tells you whether the naming is still happening. You want all three. Health, robo- and AI currently carry all three at once.

Where the judgment actually happens

There is a version of domain investing that requires no skill. Everyone can see that relay.com or cursor.com is valuable. Those names are obvious, priced accordingly, and mostly out of reach. Knowing they are good is not an edge.

The judgment shows up one tier down, in the two-word compounds and the alternative extensions, where names that look equally good on paper have very different odds of finding a buyer. That is where this data helps, because it separates aesthetics from demand.

Meridian Health and Meridian Payments are, as objects, about equally attractive. Same syllable count, same register, both easy to say. An investor going on instinct would price them similarly.

But across a quarter of funding, 18 companies named themselves “<Something> Health” and not one named itself “<Something> Payments.” Zero. The same is true of Crypto, Fintech, Money, Bank, Software, Data, and App. Meanwhile Therapeutics appeared 14 times, Energy 11, Bio 9, Robotics 8, Security 7.

That does not make Meridian Payments a bad name. It makes it a name whose buyers sit somewhere this data cannot see, which means a different thesis and probably a longer timeline. Meridian Health has demonstrable buyers in the venture cycle. Meridian Payments needs a different story about who shows up, and you should be able to tell it before you buy.

The same logic runs through extensions. venice.ai and pearl.health are structurally identical: a brand on its category’s own TLD. One reads as a company that owns its name. The other is a construction that 13 of the 14 eligible companies did not choose, concatenating onto .com instead. Nothing about the names explains the difference. Adoption does.

The exceptions prove it. The one company here sitting on a category keyword paid around $12 million and then changed its own name to match. That is available to almost nobody.

That is the whole discipline in one sentence. A good name and a name with buyers are different things, and only one of them pays.

The reason to watch funding events is not that any given round produces a sale. It is that a thousand funded companies a quarter show you which constructions are actually in use by businesses operating under real constraints, and which only look like they should be. Comps tell you what a name was worth to someone who already bought. This tells you where the next concentration of buyers is forming.

Most of the inventory that disappoints investors was not necessarily bad. It was well chosen for a demand that was not there.

We are going to keep running this every quarter. If the constructions shift, we would rather find out from the data than from a portfolio that stopped selling.

Methodology

_{Sample. 1,000 funding rounds from Atom’s funding event tracking, 91% announced April to July 2026, $186.8B across the 914 with stated amounts. Rounds, not unique companies: Atoms, CuspAI and Syntetica each appear twice and were deduplicated. Three press-wire filings claiming exactly $100M apiece were excluded. Medians throughout. The industry table covers 871 rounds; the rest are untagged or in verticals with under ten rounds.}

_{Coverage. Naming findings cover all 1,000 companies. Domain findings cover 715 domains verified individually against a company announcement or press source. That group skews larger (median round $15.0M vs $6.0M), since bigger rounds get more coverage.}

_{Limits. This is one buyer population. It says nothing about incumbent consolidation, rebrands, or strategic acquisitions, which is where category domains actually trade. We see brands and current domains; whether a company wants a different domain is inferred. Round size is not naming budget. And as the “pay” case shows, absence from this data is not absence of demand.}