
99% of onchain agentic commerce last quarter ran on $USDC.
97% settled through x402.
90% moved over @Base.
@Coinbase owns all three rails and earns almost nothing from them yet.
But Q2's headline numbers buried it (save this)...
Here's the part everyone already saw:
Revenue came in at $1.22 billion against Street expectations near $1.29 billion. The GAAP net loss was $359 million. The stock fell about 5% after hours.
But look at what the quarter was. Total market crypto spot volume fell 25%, total crypto market cap dropped 11%, and volatility compressed to multi-year lows.
Crypto is in a tough spot, and you can't ask a crypto-native company to print records through one. Even Robinhood's crypto revenue fell 38% to $100 million.
So the question is now about what Coinbase built while the market was dead.
First, the Bitcoin dependency is basically gone:
88% of net revenue now comes from something other than Bitcoin spot trading. Subscription and services alone hit 48% of net revenue.
They also cut hard into the downturn:
Adjusted expenses down 9% quarter over quarter, headcount from 4,988 to 4,321, and the full year expense outlook trimmed to $4.2 to $4.45 billion.
That combination produced a 14th consecutive quarter of positive adjusted EBITDA ($208 million) - in a market like this one!
But the balance sheet discipline isn't what we're excited about.
This is:
Coinbase is quickly becoming the settlement layer for AI agents paying each other.
According to the Artemis data in the Q2 deck, 99%+ of onchain agentic commerce completed in $USDC, 97%+ of onchain agentic transactions ran through x402 (Coinbase's protocol), and 90%+ of agentic stablecoin volume settled on Base.
Brian Armstrong's framing on it went something like this:
There will soon be more AI agents than humans, they'll need to hire and pay each other, and they can't open bank accounts.
@KyleReidhead, our Head of Research, thinks this is the single most interesting thing at Coinbase right now, and a big reason as to why he keeps holding it.
They have a real shot at owning that infrastructure.
The catch, which Brian pointed out on the call: it isn't driving meaningful revenue yet, and there's no guidance attached to it.
Kyle's fine with that. Free options don't come with guidance.
The part he's less happy about...
Coinbase appears to be losing the everything-app race.
E.g. Robinhood's event contracts generated $156 million in Q2 revenue, more than its equities and crypto lines.
Coinbase's prediction markets grew 106% quarter over quarter and crossed $100 million annualized - which is real growth off a much smaller base, but still 2nd place.
Meanwhile, $USDC (the other revenue engine) has stalled.
Average USDC market cap sat roughly flat at $77 billion, and stablecoin revenue slipped to $292 million from $305 million.
Kyle's view of things: Coinbase is struggling to expand outside its crypto bubble, and that's what's holding it back in this bear.
(He's watching Robinhood closely, including its push onchain)
But here's why he's still long...
Coinbase is the number one exchange in the US for both retail and institutions, so it wins the recovery by default.
Kyle thinks crypto has already bottomed.
And the CLARITY Act, which Brian says is at the one yard line, is the unlock for stablecoins and tokenization that Coinbase is built to capture.
He holds it as a call on the CLARITY Act and the crypto rebound, with the agentic rails as the free upside.
Long term he backs @Brian_Armstrong and trusts him to figure the rest out (though the clock is ticking).
Kyle has been long $COIN for years now, and has given multiple profitable entry calls to Milk Road PRO members in that time.
If you want Kyle's real time trade signals - try Milk Road PRO for $1, link in bio: @MilkRoadDaily
P.S. We went live with Coinbase's CBO @ShanAggarwal to talk Q2 earnings yesterday - check it out below. 👇