Two AI-infrastructure startups describe different ways to pay for compute: $125 million of equity and a $400 million debt facility

Two young companies that sell computing capacity to AI software teams announced funding within six days of each other, and their own posts describe two ways of paying for hardware: Blacksmith has raised $125 million in equity, while General Compute combined a $100 million Series A with a debt facility of up to $400 million, according to the companies' Blacksmith Series C post and General Compute Series A post.

What happened

Blacksmith, which runs continuous integration (CI) jobs for software teams, said on October 21 that Sequoia Capital is leading its Series C at a $1 billion valuation, bringing total funding to $125 million. FinanceInTime covered the round. The company said the capital goes toward scaling its team and growing its compute footprint, and that it manages just under 10 megawatts of compute across data centers in the US and Europe.

General Compute, which sells AI inference capacity, said on October 27 that it raised a $100 million Series A co-led by Blackbird and Abstract, with participation from Fuse Venture Partners, Village Global and others. The company said its first major deployments go live in the first quarter of 2027, with more than 50 megawatts of capacity in the pipeline. It said the round builds on a $400 million debt facility from Upper90 and on partnerships with chipmakers.

Two business models

The two companies sell different products, but both described owning or financing hardware and renting out the result. Blacksmith said in its Series C post: "Companies bring us their CI workloads, and we run them faster than they could anywhere else, with zero operational overhead."

General Compute said in its Series A post that it finances, deploys and operates GPUs and purpose-built inference chips from Cerebras, SambaNova and Positron AI, matching workloads to the hardware that runs them best. Customers get dedicated bare metal or a guaranteed tokens-per-second service level, without having to buy or manage the infrastructure themselves, the company said. A token is the unit of text an AI model reads or writes.

How Blacksmith's money stacked up

Blacksmith's Series B, a $45 million round led by Peak XV, valued the company at $550 million, the company said in an August 12 post. The company said it raised that round in March 2026 and announced it in August. The Series C valuation of $1 billion is about 1.8 times the Series B valuation (computed: 1,000 divided by 550).

Asked in the August post what it would do with the money, the company answered: "Compute. We're building a cloud and that ain't cheap." It said it managed hundreds of thousands of cores of compute and planned to grow that by an order of magnitude in the following months. The Series C post gave a different measure, just under 10 megawatts, and a target of five times that footprint in 2027.

Over the same period, usage rose. Since January, the number of CI jobs running on Blacksmith each week has grown from 5 million to more than 90 million, and annual recurring revenue has grown 17 times, the company said. The posts do not give revenue in dollars, so the data does not show how much of the capital is covered by sales.

How General Compute's money stacked up

General Compute's debt came first. In July it announced a committed facility of up to $400 million from Upper90 Capital Management, according to a press release republished by StorageNewsletter. The financing begins at $100 million and scales with customer demand, the release said, and Upper90 is also an equity investor in the company.

The facility is four times the size of the Series A (computed: 400 divided by 100), and the two together come to $500 million of committed capital (computed: 400 plus 100), although the debt begins at $100 million and is drawn as demand grows. The company said it had more than $300 million of secured, price-protected supply of chips.

On September 29, General Compute said it had signed a multi-year agreement with Cerebras Systems, which it called "the largest single hardware commitment we have made" and "the first deployment drawn against the $400M debt facility we secured with Upper90 in July." On October 20 it said it would buy Positron AI's Atlas systems, available to customers beginning in the first quarter of 2027, and had committed to buying next-generation Titan systems.

The companies' own explanation

General Compute describes financing as the core of its business. The company wrote in its Cerebras post: "Chips do not win markets on architectural merit. They win on deployed capacity, and deployed capacity is a financing problem before it is an engineering problem." It added that the fastest-growing infrastructure companies in AI are, in practice, extensions of NVIDIA's balance sheet, funded with NVIDIA equity and collateralized by NVIDIA GPUs. That is the company's view, not an independent finding.

The company also pointed to power. In the July release it said the newest GPU racks demand upwards of 120 kilowatts of power and need liquid-cooling retrofits, while its SambaNova-based racks run on only 20 kilowatts per rack with no water cooling. Those figures are the company's claims and were not tested by an independent party.

Blacksmith described the demand side. Its CEO, Aditya Jayaprakash, wrote that CI is the spikiest workload a company runs: a single customer's demand regularly jumps from zero to tens of thousands of CPUs and back again. He also cited a figure from Anthropic that CI job volume grew 25 times in six months, which FinanceInTime did not check against Anthropic's own publication.

What the numbers do not show

The posts do not say what share of either company's hardware is owned, leased or financed, what the debt costs, or how much compute is in use. They do not report profit. The data therefore does not show whether a company with debt-funded hardware and one with equity-funded hardware face different costs; the comparison here is limited to what each said it raised and plans to build.

What to watch

General Compute said its first major deployments go live in the first quarter of 2027. Blacksmith said it is working to grow its compute footprint five times in 2027. Both statements are company plans, and the companies' later posts and filings, if any, would show whether they were met.