The record year is not the market you are raising into

H1 2026 set a $510B record and two companies took nearly half of it. Almost none of that is a comparable for your round. Here is how to build the set that is.

fundraisingautomation

Global startup funding hit $510 billion in the first half of 2026. That is a record. It passed the whole of 2025 in six months.

Two companies took $217 billion of that. OpenAI and Anthropic, between them, 43 percent of the half.

Every founder I speak to has read the first number. Almost nobody has done anything with the second one.

$510B
Global funding, H1 2026
A record. More than all of 2025.
70%+
Of Q2 dollars went to AI
Just under half a year earlier
43%
Went to two companies
OpenAI and Anthropic, $217B in H1
Source: Crunchbase H1 2026 global funding data, reported 2 July 2026. It counts rounds that were publicly announced. Rounds nobody announced are missing from it, so read these as a floor rather than a total.

The average is the least useful number in the report

A market with two outliers that size has no usable average. The mean stops describing anything you can act on.

The headline says capital is abundant. Your inbox says nobody replied. Both are true.

They are describing different markets.

I use one rule to tell them apart. A number is a comparable only if I could plausibly have been the company in it.

By that test, $510 billion is not data. It is weather.

Your actual market is about forty rows long

Strip the record year down to rounds that could have been yours.

Same stage. Same sector. Same country or funding corridor. A cheque size within reach of your ask. Closed in the last twelve months.

Run that filter and you do not get a market. You get maybe forty rows.

Forty is enough. It gives you the going ask, the going dilution, and the funds that actually wrote at your stage this year.

Two ways to read the same news

Headline reading

A record $510B year

  • Capital is abundant, so raise more
  • AI took 70% of Q2, so add AI to the deck
  • Valuations are up, so anchor high

Comp set reading

Forty rounds that could have been mine

  • Median cheque at my stage, in my country
  • The eleven funds that wrote one this year
  • What each company had on the day they raised

Build the rows before you build the list

Most founders build the investor list first and the evidence never. Do it the other way round.

Each row needs eight fields. Company, date, round, amount, lead investor, sector, country, and one more.

The last field is the one everyone skips. What did they have on the day the money landed.

Revenue, users, pilots, contracts, or nothing at all. Whatever the announcement actually states.

That column decides whether a round is a precedent or a fantasy. Without it you are collecting logos again.

One row

Eight fields, and only one of them is hard

  1. Step 01

    The facts

    Company, date, round, amount. Straight off the announcement, no interpretation needed.

  2. Step 02

    The people

    Lead investor and the others in the round. This becomes your target list later, earned rather than guessed.

  3. Step 03

    The evidence

    What the company had on the day. If the source does not say, write UNKNOWN and leave it.

Build forty of these before you write a single outreach email.

The prompt

This is the part I automate. Paste one funding announcement, get one row or a refusal.

You are turning a funding announcement into one row of a comparables table.

Here is the announcement:
[PASTE THE ARTICLE OR PRESS RELEASE TEXT]

My company: [ONE LINE ON WHAT YOU DO]
My stage: [PRE-SEED / SEED / SERIES A]
My sector: [X]. My country: [Y]. I am raising: [AMOUNT].

Return exactly these fields, one per line. Write UNKNOWN wherever the
source does not say it. Never estimate. Never fill a gap from the
company's website or from what you already know about them.

COMPANY:
DATE:
ROUND:
AMOUNT_RAISED:
LEAD_INVESTOR:
OTHER_INVESTORS:
SECTOR:
COUNTRY:
EVIDENCE_AT_RAISE: (revenue, users, pilots or contracts stated in this
  source only; UNKNOWN if it states none)

Then answer in one line: is this a comparable for my round?

Say COMPARABLE only if stage, sector, country and amount are all within
reach of mine. Otherwise say NOT COMPARABLE and name the single field
that rules it out.

The refusal matters more than the row. Run twenty announcements through it and most come back NOT COMPARABLE.

That rejection pile is the real finding. It is the market you kept reading about and were never in.

What forty rows change

Your ask stops being a guess. You can see what companies at your stage asked for and got.

Your target list stops being a logo list. It becomes funds that wrote a cheque at your stage in the last twelve months.

And your story stops arguing with the market. You are no longer explaining why you deserve a number the market has not paid anyone like you.

The prompt above is on the resources page. Free, no email required.

Written by Mridul Sharma. Field notes on fundraising, automation, and the unglamorous work behind the raise.

Follow on LinkedIn · More posts