Nscale lines up $3.5B in pre-IPO cash with Nvidia at the table

Share
Nscale lines up $3.5B in pre-IPO cash with Nvidia at the table

The neocloud that went from obscurity to a $45 billion Anthropic contract in under a year is now selling that story to IPO investors — and one of its suppliers is helping fund the pitch. Also: Roland finally ships a generative music tool, and it is deliberately not a song machine.

Nscale is in talks to raise as much as $3.5 billion ahead of a planned IPO, and Nvidia is on both sides of the deal. Bloomberg reported Friday that the London-based AI cloud firm is seeking to sell up to $1.5 billion of convertible notes to a group of investors, while separately aiming to secure about $2 billion in financing from Nvidia — the chipmaker whose GPUs its entire business is built on. Daniel Loeb's Third Point is set to lead the convertible investment, with Goldman Sachs working the raise. The notes are being offered at the IPO price minus a double-digit percentage discount, with that discount adjusted up to a $30 billion valuation, above which the conversion price stays put.

The pitch to investors rests on a number that has roughly doubled in a month: Nscale is briefing that the total value of its contracts sits at about $103 billion, following the $45 billion agreement to supply computing power to Anthropic that we covered last week — Anthropic commits $45B to rent 460MW of Nscale AI compute. It is also floating annual revenue of about $18.1 billion and adjusted earnings before interest, taxes, depreciation and amortization of roughly $13.6 billion — figures the company stresses are illustrative rather than formal guidance. That distinction matters more than usual here: a $13.6 billion EBITDA projection on a $14.6 billion private valuation from March is either the most underpriced asset in AI infrastructure or a model that assumes every contracted megawatt gets built and paid for on schedule.

The structure is the real story. Nvidia supplying $2 billion of financing to a customer that buys Nvidia chips is the circular-financing pattern that has defined this cycle, now arriving at the pre-IPO stage with a hedge fund anchoring the equity-linked tranche and a bank running the books. Nscale has already closed roughly $3 billion in debt for two US campuses this week and signed a $3.5 billion compute deal with humanoid robot maker Figure that includes an equity stake in the customer. Bloomberg has reported the IPO itself could raise another $3 billion. A company briefing $103 billion in contracts while raising $3.5 billion in private cash is telling you something about how much capital it takes to convert a backlog into powered steel.


Roland shipped Melody Flip, its first generative AI product, and pointedly built it to output sketches rather than songs. Released Thursday as a plug-in built with Sony's Computer Science Laboratories, Melody Flip analyzes an imported track, loop, or beat for structure, tempo, key, chord progression, genre, and mood, then recommends from a library of more than 250 "Palettes" — each a connected set of melody, chords, bass, and drums. You pick a direction, generate variations, and export the result as audio or MIDI into a DAW. It runs as VST3 and AU on macOS and Windows 11, and it is included with every Roland Cloud membership tier, including the free one.

The design choice is what makes it interesting. Roland says the tool does not copy the melody of an imported song but uses its analysis as reference points for new directions, and the sound selection is deliberately unglamorous — mostly general MIDI tones of the sort you would find in a 1990s video game, which is a fairly direct way of saying you are meant to export the MIDI and finish the track somewhere else. That is a different bet than the one Suno and Udio made. Roland is selling a starting point to people who already produce, not a finished artifact to people who don't, and it is doing it under the industry's "AI For Music" principles with language about amplifying rather than replacing the artist.

It is also a defensive move. Roland watched a year of AI-music copyright fights, DJ marketplaces banning fully generated tracks, and artists suing over voice and style, and shipped a tool whose output is intentionally unfinished. The circular question for the industry: if the major instrument makers all ship AI as a sketchpad, does that normalize the technology for working musicians — or does it just make the fully generated competitors look more convenient by comparison? Melody Flip being free at the Roland Cloud entry tier suggests Roland thinks distribution, not licensing, is the fight.

What to watch: whether Nscale's $30 billion conversion cap holds as the IPO prices — the gap between a $14.6 billion private mark and a $103 billion contract book has to be resolved somewhere, and the note terms are where it gets resolved first.

Does a chipmaker financing its own customer change how you read an AI infrastructure IPO, or is that just how the buildout gets paid for? Tell us in the comments.

Read more

Akhetonics says its all-optical CPU reaches a customer in 2026

Akhetonics says its all-optical CPU reaches a customer in 2026

Light-based computing keeps promising more than it delivers — but one Munich startup has just put a date on its bet, and the interview laying it out is doing the rounds on Hacker News this week. Akhetonics says it will deploy its first commercial machine with a major customer by the end of 2026, with several more planned for 2027. The company, founded by Michael Kissner and Leonardo Del Bino, is building a computer where data enters as light, is switched as light, and circulates through memory

The Week in AI — October 5–11, 2026

The Week in AI — October 5–11, 2026

Every big claim this week turned out to rest on fine print more interesting than the headline: revenue only the company reporting it can define, safety tests sandboxed while the product keeps the web, and a Pentagon phase-out nobody would confirm until reporters kept asking. The week's top 5 1. OpenAI's revenue was $20 billion below the numbers everyone quoted — and the gap was definitional. The Financial Times reported Thursday that OpenAI's annualized revenue runs roughly $20 billion unde

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Russia's largest tech company is learning what the AI era's infrastructure war looks like from the receiving end — three data centers in four days, and with them much of the cloud layer Russian businesses run on. A Ukrainian drone strike knocked out Yandex's data center in Vladimir early Sunday morning, the third of the company's facilities hit since October 8. The site — reported at roughly 50 MW and designed for about 2,880 server racks — stopped operating completely after the attack, Yandex

Agent teams cost up to 5x more, barely score higher

Agent teams cost up to 5x more, barely score higher

The multi-agent hype train hit a benchmark this weekend — and the grid and the trucking regulators had quiet weeks of their own. Vals AI put agent teams head-to-head with single agents on its Vibe Code Bench, and the teams cost between 1.8 and 5.1 times more for almost no extra quality. The evals company ran GPT-6 Sol and Claude Opus 5.5 solo and in teams across 50 apps at two reasoning efforts; out of four comparisons, only one was statistically significant — Sol at medium effort, where the t