293 Million Explains Why Intel Shared Its x86 Core

What Intel is actually shipping

Reuters reported this week that Intel is providing its Atom processor technology to Rosaic, a startup incorporated in Delaware in May and led by Amarjit Gill. The detail that separates this from an ordinary licensing deal is the format. Intel aims to ship register-transfer level code, the description engineers use to synthesise a working chip. An architecture licence lets a company design its own core to Intel's instruction set. RTL hands over Intel's core.

Atom is Intel's low-power x86 line, the part that goes into industrial controllers, edge gateways, point-of-sale terminals and in-vehicle computers rather than servers or laptops. Intel has not widely licensed pieces of x86 and has not previously provided Atom technology this way. Rosaic has no public website and no LinkedIn presence. An amended Delaware filing on 24 July added an equity structure that points to a seed round of around 10 million dollars.

Intel declined to comment. Gill and the company's finance executive did not respond to messages. That silence is worth measuring against Intel's own publishing habits. In the same fortnight the company announced a packaging collaboration with Lens Technology, the completion of the RAMP-C programme, and a 5 billion euro manufacturing expansion in Ireland. Intel announces its partnerships. This one it did not.

The number behind the motive

Read the deal against Intel's own second-quarter results, published on 23 July, and the reading changes. Intel Foundry booked 5.8 billion dollars in the quarter, up 31 percent. External revenue inside that figure was 293 million dollars. Divide one by the other and the foundry earns about 5 percent of its revenue from customers who are not Intel.

That is the constraint. Intel's problem in 2026 is not a shortage of designs. Group revenue rose 25 percent to 16.1 billion dollars, the fastest growth in almost fifteen years, with data centre and AI up 59 percent. The problem is that the foundry Intel has spent years building is still overwhelmingly serving Intel. Lip-Bu Tan's own framing of the quarter named the CPU franchise, ASICs, advanced packaging and what he called a vast wafer foundry network as the assets to grow from. A wafer network needs wafer customers.

An independent design house built on Intel's Atom RTL is a customer of a specific kind. It is not obliged to fabricate at Intel, but the code it starts from was written for Intel's process, and porting a core to another foundry is expensive and slow. Seeding a design company creates demand that lands, by default, in the fabs whose utilisation Intel needs to lift. That is a coherent commercial motive, and it explains the decision better than the conflict-of-interest framing the story arrived with.

Two for two, and both went elsewhere

The governance question is still real, and a good motive does not answer it. Gill and Tan are longtime co-investors. Gill helped Tan assemble the founding team at Rivos. Intel has said nothing publicly about how the transfer was priced or who approved it, and no board disclosure has appeared.

The precedent is what an owner should read. The two previous chip companies these two built together were both acquired, and neither by Intel. Qualcomm bought Nuvia. Meta bought Rivos last year, outbidding Intel for it. On a two-for-two record, the base rate for a Gill-founded CPU startup is that it ends up owned by one of Intel's competitors.

That is the asymmetry in the arrangement. If Rosaic succeeds, the likeliest outcome on past form is an acquisition, and the acquirer inherits Atom RTL that Intel has never before let out of the building. Intel can write transfer restrictions into a contract. It cannot write them into the engineers who will have learned the design.

What a second x86 source is worth to you

For anyone specifying low-power x86, this is the first credible prospect of a second source in that class, and second sources move prices well before they ship. European industrial buyers have lived with a single supplier for embedded x86 for as long as the category has existed, on procurement cycles that run ten years or longer.

The honest limit is that Rosaic has no product, no announced roadmap, no website and no confirmed funding beyond a filing indicating roughly 10 million dollars. Nothing here belongs in a design-in decision this year. Treat it as a negotiating fact rather than a sourcing option.

The practical move is small and costs nothing. At the next embedded silicon renewal, ask the incumbent whether its roadmap pricing assumes it remains the only source of Atom-class x86 through the contract term. The answer, or the reluctance to give one, tells you what the supplier thinks this deal is worth. That is more useful than waiting for a company with no website to ship something.

Read next: Intel's Foundry Grew 31%, and You Still Depend on TSMC  |  Cadence And Synopsys Tuned Their Tools For Nvidia