A new leak is rattling the PC hardware world just as buyers were starting to plan their next GPU purchase around 2027. According to a report published by Wccftech on September 16, 2026, hardware leaker Kepler_L2 claims that both AMD and Nvidia are pushing their next-generation gaming GPU architectures back to 2028, with a single exception: an AMD chip codenamed AT2, reportedly built on the RDNA 5 architecture, which is said to be the only new GPU die launching in 2027.
The claim, if accurate, would mark one of the longer gaps between flagship GPU generations in recent memory. It also lands at a moment when memory pricing and supply have already forced chipmakers across the industry to rework their 2027 and 2028 roadmaps, a dynamic shattered.io covered in August when Intel’s CEO warned memory costs would not ease until 2028. Here is what the leak actually says, what is confirmed versus speculation, and what it could mean for anyone waiting on a new AMD RDNA 5 or Nvidia Rubin graphics card. For the latest on GPUs, CPUs, and the broader chip market, see our hardware coverage hub.
What the Leak Claims: AMD and Nvidia GPUs Pushed to 2028
Wccftech’s report centers on a claim from Kepler_L2 that the bulk of AMD’s and Nvidia’s next-generation gaming GPU lineups will not reach the market until 2028. That covers AMD’s RDNA 5 architecture broadly and, in related coverage of the same leak wave, Nvidia’s Rubin architecture tied to a future GeForce RTX 60 series. The single carve-out named in the leak is AT2, an AMD die said to be RDNA 5-based, which Kepler_L2 describes as the only GPU launching next year, meaning 2027.
It is worth being precise about what is and is not established here. The existence of Kepler_L2’s post, and the fact that Wccftech and other outlets reported on it, is confirmed. The underlying claim, that AMD and Nvidia have actually decided to delay their next-gen GPU stacks to 2028, is not confirmed by either company. Neither AMD nor Nvidia has issued a statement addressing the report as of publication. That distinction matters, because leak-driven roadmap stories have a habit of hardening into “fact” in comment sections long before any company confirms a date.
Leaker: Kepler_L2 (posted to hardware forums, incl. AnandTech)
Core claim: AMD and Nvidia next-gen gaming GPUs slip to 2028
Named exception: AT2 (AMD, reported RDNA 5-based) ships 2027
Nvidia reference: Rubin architecture / future RTX 60 series
Confirmed by AMD or Nvidia: No
First reported by: Wccftech, September 16, 2026Who Is Kepler_L2, and Why This Leak Is Getting Attention
Kepler_L2 has built a reputation in enthusiast hardware circles as a source who posts early architectural and roadmap details ahead of official announcements. Wccftech and other outlets covering this specific report describe Kepler_L2 as a prominent, long-time leaker with a track record in the space. That framing, however, is characterization applied by the outlets covering the claim, not a credential verified by AMD or Nvidia. Leakers in this space operate without any formal relationship to the companies whose roadmaps they discuss, so their accuracy varies from leak to leak.
That caveat does not mean the claim should be dismissed outright. Leak-driven reporting has correctly called major GPU roadmap shifts before, sometimes months ahead of any corporate confirmation. But it does mean readers should treat the 2028 timeline, and the AT2 specifics attached to it, as a claim under active reporting rather than a settled product date. Tom’s Hardware and other GPU-focused outlets have historically flagged similar early leaks as directional rather than definitive, and that framing applies here too.
The AT2 Chip: AMD’s Lone RDNA 5 GPU for 2027
The most concrete piece of the leak is the AT2 die itself, and even that comes with heavy caveats. Reports describe AT2 as an AMD GPU built on the RDNA 5 architecture, positioned by most coverage as a mid-range part rather than a halo flagship. Beyond that broad description, the specifics diverge depending on the outlet. Some coverage of the same leak describes AT2 as roughly comparable in scope to a successor for AMD’s current RX 9070 XT tier, while other reporting attaches specific compute unit counts, ranging from figures in the 40-CU range to numbers closer to 64 CUs depending on the source.
None of those compute unit figures, clock targets, or memory configurations have been confirmed by AMD. There is no official AMD product page, roadmap slide, or press release that lists AT2 as an announced die, product name, or SKU. What is consistent across the coverage is the framing that AT2 would be the only new GPU AMD ships in 2027 if the broader delay claim holds, which would make it a notably narrow product launch for a full calendar year from a company that has historically shipped multiple SKUs per generation within months of each other. This isn’t the first RDNA 5 leak to circulate this year either; an earlier report pointed to clock speed targets for the architecture well before AT2 entered the conversation.
What AT2 Would Replace
AMD’s current RDNA 4 lineup, led by the RX 9070 XT, launched in the mid-range and upper-mid-range segments in 2025. If AT2 is positioned as its successor, as some of the reporting suggests, that would mean AMD skips releasing a new halo flagship GPU in 2027 entirely, leaving RDNA 4 or a refreshed variant of it to hold the top end of AMD’s stack for an extended stretch. That is a meaningfully different cadence than AMD’s recent generations, where flagship and mid-range parts typically launched within the same product wave.
Nvidia’s Side: Rubin and the RTX 60 Series
The Nvidia half of the leak is thinner on specifics than the AMD half. Coverage of the same Kepler_L2 claim ties Nvidia’s delay to its Rubin architecture, expected to eventually underpin a future GeForce RTX 60 series. Unlike AT2, no exception has been reported on Nvidia’s side, meaning the claim as circulated suggests Nvidia does not have an equivalent stopgap chip landing in 2027. Nvidia currently sells its Blackwell-based RTX 50 series, led by the RTX 5090, as its newest consumer gaming lineup, and if the leak holds, that generation would need to carry Nvidia’s desktop GPU stack for a longer stretch than the roughly two-year cadence the company has followed for most of the past decade.
Nvidia has not commented on the report. The company’s public GPU roadmap communications tend to arrive close to launch windows rather than years in advance, so the absence of an official response this early is not unusual on its own. Readers can track Nvidia’s current shipping lineup directly on the GeForce graphics card page, which as of this writing still centers on RTX 50-series Blackwell hardware with no Rubin-based consumer product listed.
Why Chipmakers Might Push Next-Gen GPUs to 2028
The leak does not include an official explanation from either company, so any reasoning has to be inferred from broader industry conditions rather than stated as company confirmation. Several factors are plausible contributors, and multiple have already shown up in this outlet’s own 2026 coverage of the chip industry.
- Memory pricing and supply: DRAM and HBM shortages have already pushed up costs across the PC hardware stack in 2026, a dynamic Intel’s own leadership addressed when discussing why relief was not expected until 2028.
- AI chip demand competing for fab capacity: Foundries producing leading-edge nodes are allocating significant capacity to AI accelerators, which can squeeze timelines for consumer GPU dies built on the same process nodes.
- Longer validation cycles for larger, more complex dies: As GPU architectures grow more complex generation over generation, engineering and validation windows tend to stretch as well.
- Cadence normalization after a compressed recent history: AMD and Nvidia have both shipped multiple architectures in relatively close succession over the past several years, and a longer gap could reflect a return to a slower, more traditional multi-year cycle.
None of these explanations are confirmed as the actual reason behind the reported delay. They are presented here as plausible industry context, not as a stated rationale from AMD or Nvidia.
Historical Context: How GPU Generation Cadence Has Shifted
To understand why a 2028 timeline would stand out, it helps to look at how AMD and Nvidia have historically spaced their major GPU architecture launches. Both companies have generally followed a roughly two-year rhythm for flagship gaming architectures over the past several product cycles, though the gap has varied.
| Company | Architecture | Flagship Launch Year | Gap From Prior Gen |
|---|---|---|---|
| AMD | RDNA (RX 5000) | 2019 | – |
| AMD | RDNA 2 (RX 6000) | 2020 | ~1 year |
| AMD | RDNA 3 (RX 7000) | 2022 | ~2 years |
| AMD | RDNA 4 (RX 9000) | 2025 | ~2-3 years |
| AMD | RDNA 5 / AT2 (reported) | 2027 (claimed, unconfirmed) | ~2 years (claimed) |
| Nvidia | Turing (RTX 20) | 2018 | – |
| Nvidia | Ampere (RTX 30) | 2020 | ~2 years |
| Nvidia | Ada Lovelace (RTX 40) | 2022 | ~2 years |
| Nvidia | Blackwell (RTX 50) | 2025 | ~2-3 years |
| Nvidia | Rubin (RTX 60, reported) | 2028 (claimed, unconfirmed) | ~3 years (claimed) |
If the leak’s 2028 timeline holds for the bulk of each company’s next-gen stack, it would extend the gap since each company’s last full architecture launch beyond the roughly two-year pattern both have followed for most of the past decade. That would not be unprecedented on its own, cadence has already been drifting longer on both sides, but it would be a further step in that direction rather than a return to a faster release rhythm.
Market Impact: What This Means for Gamers and System Builders
For anyone planning a build around a next-generation GPU in the near term, the practical impact of this leak, if it holds, is straightforward: current-generation hardware stays relevant for longer. AMD’s RDNA 4 lineup and Nvidia’s RTX 50 series would need to serve as each company’s primary consumer offering well into 2027 and, per the AMD side of the leak, potentially into 2028 for anything above the AT2 tier.
That has knock-on effects for pricing. GPU prices, especially at the high end, have already been under pressure in 2026 from AI-driven demand and memory shortages, a trend this outlet tracked when RTX 5090 prices climbed sharply earlier this year. A longer gap before next-gen replacements arrive removes one of the natural mechanisms that typically pulls prices down on existing stock, the anticipation of a new generation prompting retailers and resellers to discount outgoing inventory. If buyers know RDNA 4 and RTX 50 cards are the top options for another one to two years, there is less competitive pressure pushing current prices lower.
There is also a secondary market effect worth watching: refresh SKUs. When flagship architecture launches slip, chipmakers have historically filled the gap with mid-cycle refreshes, higher-clocked variants, or expanded memory configurations of existing dies rather than leaving the lineup static. If AT2 really is AMD’s only new die for 2027, expect pressure on AMD to backfill the rest of its stack with exactly that kind of refresh strategy.
Competitive Comparison: AMD’s AT2 Bet vs Nvidia’s Full Stack Delay
The most interesting strategic wrinkle in this leak is the asymmetry between the two companies. As reported, AMD gets a 2027 product in AT2, while Nvidia, on the same reported timeline, does not get anything until 2028. If accurate, that would hand AMD roughly a year of exclusive access to a new-architecture GPU in the market before Nvidia’s Rubin-based cards arrive, a potentially significant competitive window given how closely the two companies typically trade blows on release timing.
| Factor | AMD (per leak) | Nvidia (per leak) |
|---|---|---|
| Architecture | RDNA 5 | Rubin |
| Reported 2027 product | AT2 (only new die claimed) | None reported |
| Reported full-stack timeline | 2028 | 2028 |
| Segment of 2027 exception | Reported mid-range / RX 9070 XT successor tier | Not applicable |
| Official confirmation | None | None |
| Current shipping flagship | RDNA 4 (RX 9070 XT era) | Blackwell (RTX 5090) |
It’s worth noting that a mid-range 2027 launch is not the same competitive win as a flagship one. If AT2 lands where most of the reporting places it, as a successor to AMD’s current upper-mid-range tier rather than a true halo part, AMD would still be without a next-gen flagship to challenge Nvidia’s top end until 2028 on the same reported timeline. The practical competitive edge, if the leak is accurate, would be concentrated in the segment AMD has been trying hardest to win back: performance-per-dollar buyers rather than the enthusiast ceiling.
How the Memory Shortage Ties Into This Roadmap
This leak does not exist in isolation. It follows months of reporting on tightening memory supply across the PC and AI hardware industries in 2026. Shattered.io has covered how that shortage has already reshaped pricing on budget devices and pushed Intel’s own leadership to warn that meaningful relief was not expected before 2028, the exact same year now attached to this GPU delay leak. Whether or not memory constraints are the actual driver behind AMD’s and Nvidia’s reported roadmap shift, the overlap in timing is notable, and it fits a broader 2026 pattern where component supply, rather than pure engineering readiness, has increasingly dictated when new hardware reaches shelves.
Wall Street has already been reacting to AMD’s and Nvidia’s shifting product and manufacturing priorities this year. Coverage of AMD’s own guidance updates showed the company leaning harder into AI silicon as a growth driver even as its consumer GPU roadmap details stayed vague, a pattern consistent with a company that may be deprioritizing a fast gaming GPU cadence in favor of data center allocation. Chip stocks have also shown they can move fast on GPU-adjacent news, as seen when Intel and AMD shares both dropped following an unrelated Nvidia PC chip announcement earlier this year.
What Outlets Covering the Leak Agree and Disagree On
Several outlets picked up Kepler_L2’s claim within the same news cycle, and while the core storyline, an exception chip named AT2 shipping in 2027 while the rest slips to 2028, is consistent, the technical details attached to AT2 are not. Some coverage frames AT2 as a roughly 64 compute unit successor to AMD’s RX 9070 XT, while other reporting describes it with a lower 40 compute unit figure and a mid-range positioning. Both cannot be precisely correct at the same time, which is itself a useful signal: when independent outlets covering the same source diverge on hard specs, it typically means the underlying leak was light on technical detail and outlets filled gaps with their own interpretation or additional unverified sourcing.
That divergence is a good reason to treat every AT2 spec beyond “RDNA 5-based” and “reportedly the only 2027 die” as unverified until AMD publishes something official. Notebookcheck and PC Gamer both maintain ongoing GPU roadmap trackers worth checking as more detail emerges.
Predictions: What Happens Next
With no official confirmation from AMD or Nvidia, anything beyond the leak itself is speculative. That said, based on how similar roadmap leaks have played out in past cycles, a few outcomes look more likely than others.
- Expect AMD and Nvidia to stay silent on specific dates for months. Neither company typically confirms or denies leaked roadmap timelines this far ahead of launch, and both are more likely to let the report sit unaddressed than issue a correction or confirmation.
- Expect additional leaks to narrow AT2’s actual specs. Given how much the current reporting disagrees on compute unit counts, further leaks refining AT2’s die size, memory bus, and target price tier are likely over the next two to three months.
- Expect current-gen GPU prices to stay firm rather than drop. Without a confirmed near-term successor at the high end, RTX 5090 and top RDNA 4 pricing has less reason to soften, particularly with AI-driven memory demand already applying upward pressure.
- Expect refresh SKUs to fill AMD’s and Nvidia’s 2027 gaps. If flagship next-gen parts really do slip to 2028, both companies have historical precedent for bridging the gap with refreshed variants of existing architectures rather than leaving their lineups static for a full year or more.
- Expect the memory shortage narrative to keep intersecting with GPU roadmap coverage. As long as DRAM and HBM supply stays tight into 2027 and 2028, expect every major chip roadmap story, GPUs included, to get read partly through that lens.
What to Watch For Confirmation
Readers looking to separate confirmed fact from ongoing leak speculation should watch for a few specific signals. AMD financial disclosures and investor calls, tracked on the company’s own investor relations site, occasionally include forward-looking language about product cadence even when specific chip names aren’t mentioned. Nvidia’s GTC and CES keynote slots have historically been where the company telegraphs its next architecture name well before launch, so the absence or presence of a Rubin-branded consumer GPU mention at the next major Nvidia keynote will be a meaningful data point. And any AMD product-page listing, driver leak, or firmware reference to “AT2” by name would move that detail from rumor to strong circumstantial evidence.
Frequently Asked Questions
Is it confirmed that AMD and Nvidia are delaying their next-gen GPUs to 2028?
No. The claim comes from leaker Kepler_L2 and was reported by Wccftech on September 16, 2026. Neither AMD nor Nvidia has confirmed the timeline.
What is AT2?
AT2 is the codename attached to an AMD GPU that reports describe as RDNA 5-based, reportedly the only new AMD gaming GPU launching in 2027 if the broader delay claim is accurate. No official AMD source has confirmed AT2’s existence, specs, or launch date.
Is AT2 a flagship GPU?
Most coverage of the leak positions AT2 as a mid-range part, with some reports comparing it to a successor for AMD’s current RX 9070 XT tier. These positioning details are unconfirmed by AMD.
What is Nvidia Rubin?
Rubin is the architecture name associated with Nvidia’s next-generation GPUs in reporting tied to this leak, expected to eventually power a future GeForce RTX 60 series. Nvidia has not published a consumer gaming roadmap confirming Rubin-based RTX 60 cards or a launch date.
Should I wait to buy a GPU if this leak is true?
That depends entirely on your timeline and needs. If the leak holds and next-gen flagships don’t arrive until 2028, current RDNA 4 and RTX 50 hardware would remain the top options for a longer stretch than usual, which for many buyers weakens the case for waiting.
Why would AMD and Nvidia delay their GPU launches?
No official reason has been given by either company. Industry-wide memory shortages, fab capacity competition from AI chip demand, and longer validation cycles for complex dies are plausible contributing factors based on broader 2026 industry trends, but none of these have been confirmed as the specific cause of this reported delay.
Who is Kepler_L2?
Kepler_L2 is a hardware leaker described by outlets covering this report as a prominent, long-time source of early GPU architecture and roadmap details. That characterization comes from media coverage, not from any official standing with AMD or Nvidia.
When will we know for sure?
There is no confirmed date. Official confirmation would most likely come through AMD or Nvidia investor communications, a keynote announcement, or a leaked product listing referencing AT2 or Rubin by name.
