What the RDNA 5 rumor actually says
According to iXBT.games, Leadbetter referred to behind-the-scenes information suggesting Sony had “some kind of exclusivity” for RDNA 5 in handheld devices. Notebookcheck also attributes the remark to a Patreon-exclusive episode of Digital Foundry Direct. That is an attributed industry rumor, not an announcement from Sony or AMD.
There is also a verification limit: Frandroid reports that the complete excerpt was not publicly available, leaving the surrounding discussion unclear. Its coverage says neither Sony nor AMD has confirmed the claim and notes that Sony has not officially announced the next-generation handheld described in these reports. Canis therefore remains a reported codename, not a confirmed product specification sheet.
A custom chip is not exclusive ownership of an architecture
The important distinction is between having a customer-specific processor and preventing competitors from using the underlying graphics architecture.
AMD’s annual report, filed February 4, 2026, describes its semi-custom business as developing customer-specific solutions around CPU, GPU and multimedia technologies. It identifies PlayStation 5 and Xbox Series X|S as products powered by these designs. Tailoring silicon to a console maker’s requirements is an established business—not, by itself, evidence of architectural exclusivity.
AMD also explains that this business generates both design-and-development fees and revenue from subsequent chip sales. That provides a concrete economic reason why a customer’s commitment matters: the relationship involves engineering work as well as manufacturing processors. It does not verify a particular minimum order, a Sony handheld agreement or restrictions on other customers.
Notebookcheck’s interpretation is that the rumored advantage could reflect AMD’s selectivity about undertaking handheld chip projects, rather than a conventional exclusivity contract. That explanation remains unconfirmed. The earlier KeplerL2 claim about production-volume requirements is likewise not an announced AMD policy.
What this means for Xbox and Steam Deck
For Xbox and handheld PC readers, the distinction changes the story considerably. Sony potentially commissioning a bespoke RDNA 5 processor would not, on the evidence available, demonstrate that Microsoft, Valve or other manufacturers are prohibited from obtaining RDNA 5-based hardware. The reports establish neither competitors’ final chip choices nor the alleged arrangement’s duration or boundaries.
The sensible conclusion is conditional: Sony could gain a distinctive custom-silicon option if the rumor proves accurate. Calling that a guaranteed performance victory—or declaring future Xbox and Steam Deck hardware locked out—goes beyond the evidence. For now, the interesting question is who can commission which chip, not who has already won the handheld race.
Update: Digital Foundry clarifies RDNA 5 comment was not confirmation of Sony exclusivity (October 26, 2026)
Digital Foundry has now directly clarified that its earlier discussion should not be read as confirmation that Sony has blocked AMD from supplying RDNA 5-based chips to competing handhelds. The outlet says the “exclusive of sorts” wording referred only to the currently rumored leaked AMD handheld SoCs: Canis, said to use RDNA 5, and Gainsborough, said to use RDNA 4m.
Crucially, Digital Foundry says it is working from broadly the same leak-based information available elsewhere and that its deductions are not confirmation of Sony’s handheld plans or AMD’s customer restrictions. That explicitly undercuts interpretations that AMD has stopped making custom chips for devices such as a future Steam Deck.
For Windows handheld buyers and IT observers tracking the market, the practical takeaway is unchanged but firmer: neither a 2027 Xbox handheld nor a possible Steam Deck successor can be considered ruled out of RDNA 5 hardware based on this rumor. Their eventual silicon choices—and whether Microsoft or Valve would use an existing AMD design or commission something different—remain unknown.