Two well-regarded GPU leakers now agree: Nvidia’s next gaming architecture isn’t landing in 2027 after all. Kopite7kimi, described by TechSpot as one of the most renowned leakers covering graphics cards, posted on September 21 that the Rubin-based RTX 60 series has moved to 2028, walking back an earlier prediction of a second-half-2027 launch.
TechSpot puts the RTX 50 series, the current generation led by the RTX 5080 and RTX 5090, at almost 20 months old already. If Rubin lands in the second half of 2028 as now predicted, the gap between generations stretches to somewhere between 41 and 47 months, well past Nvidia’s average cadence of around 26 months.
A Second Leaker Says the Same Thing About AMD
Kopite7kimi’s post echoes a claim made days earlier by Kepler_L2, another leaker with a track record on both companies’ roadmaps, who said neither the RTX 60 series nor AMD’s RDNA5-based Radeon 10000 series would ship before 2028. Two separately sourced predictions converging on the same year, for two different companies, is what makes this worth taking seriously as more than a single rumor.
Kepler_L2’s claim carves out one exception: AMD’s “AT2” GPU, reportedly aimed at RTX 5080-class performance, is still said to be on track for a 2027 launch even as the rest of RDNA5 slides. Nvidia’s own roadmap has a comparable near-term item, an RTX 50 Super refresh that TechSpot says multiple reports earlier this year described as ready to ship but held back specifically over the cost of 3GB GDDR7 memory modules.
The Memory Crunch Is the Reason, Not a Coincidence
Neither leaker needed to explain the delay: everyone tracking the GPU market already knows why. AI accelerator demand has pulled DRAM and GDDR7 production capacity away from consumer graphics cards for most of 2026, and TechSpot notes the RTX 5090’s street price has run as high as $7,000 to $9,000 in parts of the US market this year. TechSpot also cites SK hynix among several firms warning the shortage could get worse before it improves, persisting as late as 2030.
OC3D, sourcing the same Kepler_L2 and Kopite7kimi posts, instead points to a more optimistic industry expectation that the DRAM shortage peaks around mid-2027 as new production comes online, positioning 2028 as a calmer year to launch into. The two outlets aren’t citing the same underlying forecast, so treat the exact recovery date as unsettled; what both agree on is that memory supply, not engineering readiness, is the reason either company would wait.
AMD’s Driver Work Says the Chips Are Still Coming
A delayed launch date doesn’t mean paused engineering. AMD sent Linux kernel graphics driver patches on September 21 that add “GDDR7” as a recognized video memory type inside the open-source AMDGPU driver, alongside new support for the IH 8.0 interrupt handler and NBIF 7.10 bus interface IP blocks. None of that is surprising on its own, current RDNA4 cards use GDDR6 while RDNA5 has been expected to move to GDDR7, but it confirms AMD’s engineers are actively laying groundwork for silicon that, per Kepler_L2, most of the lineup won’t ship until 2028. The patches follow other recent AMDGPU work introducing Display Core Next 6 and GFX 13.0.x support, both read as RDNA5-generation IP.
None of this is official. Nvidia and AMD have confirmed no roadmap dates for either RTX 60 or RDNA5, and a leaker’s account, however consistent its history, is not a company statement. What’s changed is that two separately sourced predictions, from people who don’t appear to be relaying the same tip, now point at the same year for both of the market’s GPU makers, for the same underlying reason.