Who raised on Friday 28 August 2026

How we would run these

Both of today's raises are credit, and both were announced like credit. Lambda took $1B in private debt to buy Nvidia chips it leases to Microsoft. USD.AI took a $100M facility from Bullish to lend against AI infrastructure. Two balance sheet announcements, aimed at an audience that spends all day arguing about who can actually get GPUs.

Lambda, $1B in private debt

JPMorgan arranged it. Second facility this week after a $926M term loan B, and it lands while the company is reported to be in talks for a $3B pre-IPO round.

Nobody quote-tweets a credit facility. What this money buys is capacity, and capacity is the only question a buyer has: can you get chips, and can you turn them on. A billion dollars of Nvidia hardware with Microsoft already on the other side of the lease answers that more convincingly than a Series D ever does.

So we would run it as a supply claim. The founder posts a count of GPUs and the date they go live, because that is a promise a buyer can hold you to and it is the kind of specific that gets screenshotted. Then 80 to 100 creators who actually rent compute, not generalist tech accounts, saying what three months of waiting on capacity somewhere else cost them. The Microsoft lease carries its own credibility, so the copy leaves it alone and lets people find it.

Then the second half, which is where most infrastructure launches stop early. X is where the mindshare gets acquired. LinkedIn is where it converts, because the person who signs an infrastructure contract lives there. Same claim, longer, written for the buyer instead of the crowd, going out while the X thread is still moving.

USD.AI, $100M credit facility

Bullish provided it, and the capital funds lending against AI infrastructure. A lender that finances compute just got financed, which puts this two steps away from anyone actually buying a GPU.

That distance is the story. A credit layer forming around compute is what a maturing market looks like from the inside, and whoever runs a company sitting in that position can post a view of where the money is going that nobody else is placed to write. The facility is the permission slip. The argument is the post.

This is the version of use case seeding that works for a company with no product to demo. Hand five people who already write about AI infrastructure the same underlying data on what is being lent against and at what terms, let them each reach their own conclusion in their own voice, and the read becomes consensus instead of a company line. It reads as consensus because it is.

What both of them are leaving

A round is the one week a company holds attention it did not pay for, and the half-life is about 48 hours. Both of these went out as financial news, which reaches the people who already read financial news and stops.

The follow-on matters more than the day. Spending the whole budget on the announcement post is the mistake we talk clients out of most often, because a hundred thousand dollars of amplification on a single post reads as bought and everybody can tell. The version that works is the drop, then a fixed bench holding the feed every week afterwards at a fraction of the spend. Out of sight, out of mind, and nobody buys infrastructure from a company they have stopped thinking about.