A second well-known leaker has now backed up the claim that rattled PC hardware forums this week: Nvidia’s next flagship GPU generation, the Rubin-based GeForce RTX 6000 series, may not reach store shelves until 2028. TechSpot reported the update on September 21, 2026, noting that leaker kopite7kimi — who previously pointed to the second half of 2027 as a launch window — has pushed his own estimate back a full year. A separate leaker known as Kepler told the same outlet that both Nvidia’s RTX 6000 cards and AMD’s RDNA 5-based Radeon 10000 series are now tracking for 2028, not 2027.
GamesRadar+ picked up the story the same day, framing it in blunter terms: if the timeline holds, it would mark the longest gap between GPU generations in 20 years. Coverage has since spread across the hardware press, including Wccftech, Club386, tbreak.com, and OC3D, turning a single-source rumor into something closer to an industry consensus. None of it is official. Nvidia has never confirmed an RTX 60-series launch window, and the company has stayed silent since the reports broke.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What the new leak actually claims
The specifics, according to TechSpot’s reporting, are narrower than the headlines suggest. Kopite7kimi’s original call had Rubin-based RTX 6000 desktop cards arriving in the second half of 2027, roughly two and a half years after the RTX 50 series (Blackwell) launched in January 2025. That estimate has now slipped to 2028, with no month or quarter attached. Kepler’s claim goes further, tying AMD’s next architecture into the same delayed window. If accurate, that would mean both major GPU vendors sit out consumer flagship launches for all of 2027, an unusual alignment that neither company has acknowledged.
TechSpot describes kopite7kimi as one of the most closely watched leakers in the graphics card space, a reputation built over several previous Nvidia architecture cycles. That track record is the main reason the report is being treated as more than internet noise, even though it originates from anonymous social media accounts rather than an Nvidia roadmap slide or an SEC filing. Readers should treat “2028” as the current best guess among leakers, not a locked date.
Why two leakers agreeing matters
Single-source leaks about unreleased silicon are common and often wrong in the details, even when a leaker’s general direction turns out correct. What changed this week is corroboration. Kopite7kimi and Kepler run independent networks of supply-chain contacts, and their estimates converging on the same year raises the odds that Rubin’s schedule really has moved, even if the underlying reason (yield issues, packaging capacity, demand from AI data centers eating fab time, or something else entirely) remains unconfirmed. GamesRadar+ was explicit that Nvidia has never put an official date on RTX 60-series hardware, so everything downstream of that is inference built on inference.
It’s also worth separating two different questions that are getting merged in some of the coverage: when will Rubin-based data-center GPUs ship, and when will consumer RTX 6000 desktop cards ship. Nvidia’s AI accelerator roadmap and its GeForce roadmap don’t move in lockstep, and a slip in one doesn’t automatically confirm a slip in the other. The leaks specifically concern the consumer GeForce RTX 6000 line, which is where the “2028” framing applies.
How this compares to previous GPU generation gaps
Nvidia’s release cadence has already been stretching. Turing arrived in 2018, Ampere in 2020, and Ada Lovelace (the RTX 40 series) in late 2022, each roughly two years apart. Blackwell, the RTX 50 series, broke that pattern by launching in January 2025, over two years after Ada. A 2028 arrival for Rubin-based RTX 6000 cards would stretch the gap to roughly three years, which lines up with the “longest gap in 20 years” framing GamesRadar+ used in its coverage. The table below lays out that cadence using publicly known launch dates alongside the currently rumored Rubin timeline.
| Architecture | Flagship series | Launch | Gap from prior gen |
|---|---|---|---|
| Turing | RTX 20 series | 2018 | ~2 years |
| Ampere | RTX 30 series | 2020 | ~2 years |
| Ada Lovelace | RTX 40 series | Late 2022 | ~2 years |
| Blackwell | RTX 50 series | January 2025 | ~2.3 years |
| Rubin (rumored) | RTX 60 / RTX 6000 series | 2028 (per kopite7kimi, Kepler) | ~3 years (unconfirmed) |
Tech Insider covered the initial version of this story in detail when it first surfaced, including how it intersected with RTX 5090 pricing hitting $4,329 and the broader question of whether RTX 60 series skips 2027 entirely. This week’s development is the corroboration angle: a second leaker, plus a wave of outlet pickups, has hardened what was previously a single rumor into something closer to consensus.
The RTX 50 series pricing backdrop
A delayed RTX 60 series matters most because of what it means for the cards already on shelves. Blackwell pricing has been climbing steadily through 2026, driven by memory costs rather than GPU die shortages. Tech Times reported in February 2026 that average RTX 50-series pricing had risen 19% over a three-month window across ten regional markets it tracked, with the RTX 5090 moving from roughly $1,999 at a November 2025 low to more than $3,000 for commonly available models, and some premium partner cards clearing $5,000. By late summer, research site shanethegamer.com put the RTX 5090’s typical street price in the $4,600 to $5,000-plus range, citing memory costs as the primary driver. Tech Insider’s own reporting in August found RTX 5090 listings on Amazon reaching $4,329, more than double the launch MSRP. Nvidia’s workstation-tier RTX PRO 6000 Blackwell, announced March 18, 2025 with 96GB of GDDR7 memory, tells a similar story in miniature: TechPowerUp and Tom’s Hardware both record that announcement alongside an $8,565 launch price, with Nvidia scheduling Workstation and Max-Q edition distribution for April 2025 and broader workstation-manufacturer availability following in May 2025, a figure that’s since climbed well past MSRP on secondary markets as the broader memory squeeze has taken hold.
None of that pricing pressure traces back to Rubin or the RTX 6000 delay directly. It’s a separate, ongoing story about GDDR7 and DRAM supply that Tech Insider has tracked in pieces on the HBM4 memory crunch and DRAM price jumps pushing GPU relief to 2028. But the two stories now share a punchline: whether you’re waiting on cheaper Blackwell cards or a next-generation replacement, 2028 keeps showing up as the earliest realistic relief point. Even bulk buyers haven’t escaped the squeeze: distributor Connection was listing RTX PRO 6000 Blackwell units at $8,435 apiece in March 2025, according to pricing tracker GPU Poet, barely under TechPowerUp’s $8,565 single-card launch figure for the same Max-Q variant, a card Nvidia specs at a 300W maximum power draw as of its July 2026 listing, a sign that volume discounts aren’t absorbing the memory-driven cost increases either.
| GPU | Launch MSRP | Reported street price (Aug–Sept 2026) | Source |
|---|---|---|---|
| RTX 5090 | $1,999 | ~$4,600–$5,000+ | shanethegamer.com research, Sept 2026 |
| RTX 5090 (Amazon listings) | $1,999 | $4,329 | Tech Insider, Aug 2026 |
| RTX 5080 | $999 | ~$1,500 | Market pricing survey, Aug 2026 |
| RTX 5070 Ti | $749 | ~$1,100 | Market pricing survey, Aug 2026 |
| RTX 5070 | $549 | ~$900 | Market pricing survey, Aug 2026 |
| RTX PRO 6000 Blackwell (workstation) | N/A | $16,000, listed out of stock | Nvidia US marketplace, mid-Aug 2026 |
Figures vary by tracker and time window because retail availability has been inconsistent all year, so treat the ranges above as directional rather than a single fixed price. What’s consistent across every source is the direction: up, sharply, since the January 2025 launch MSRPs.
The memory shortage sitting underneath all of this
GDDR7 and HBM supply constraints are the backdrop the RTX 6000 delay rumors are landing on top of. Tech Times reported in July 2026 that a 3GB GDDR7 module cost between $60 and $70, with 2GB modules running roughly $20 — component costs that ripple directly into finished card pricing across Nvidia’s entire lineup, not just the flagship RTX 5090. AI data-center demand for HBM has been pulling fab capacity away from consumer GDDR production for much of 2026, a dynamic Tech Insider examined in its coverage of the Nvidia B200’s premium pricing and the GPU loan boom it triggered on the data-center side. That memory intensity is baked into the hardware itself: Puget Systems’ benchmarking logged the RTX PRO 6000 Blackwell carrying a full 96GB of memory and 125 TFLOPS of compute back in June 2025, specs that sit alongside Nvidia’s own listing of 4,000 AI TOPS and a 600W maximum power draw for the Workstation Edition, current as of October 2026, a spec sheet that leaves the card especially exposed to exactly the GDDR7 and HBM cost swings described above.
If Rubin-based consumer cards are genuinely tracking for 2028, memory supply is one of the more plausible explanations floating around hardware forums, though it’s worth repeating that Nvidia hasn’t confirmed a cause any more than it’s confirmed a date. Fabs building HBM4 for AI accelerators and GDDR7 for GeForce cards draw on overlapping capacity at TSMC and memory partners like Samsung and SK Hynix, and a squeeze on one side can delay the other.
AMD’s parallel delay changes the competitive picture
The detail that separates this week’s report from the original rumor is Kepler’s claim about AMD. If RDNA 5 and the Radeon 10000 series are also tracking for 2028, this stops being a Nvidia-specific story about a single vendor slipping a roadmap and becomes a broader statement about the health of the leading-edge GPU supply chain. Tech Insider has separately reported on AMD overtaking Nvidia in unit GPU sales earlier in 2026, largely on the back of more available and better-priced Radeon cards while Blackwell supply stayed tight. A synchronized 2028 slip for both companies’ next architectures would extend that dynamic rather than resolve it, since neither vendor would be forced to compete on a fresh flagship until then.
Intel’s Arc roadmap hasn’t featured in any of the reporting around this story, and the company has not issued a public statement about its next-generation discrete GPU timeline in response to the Nvidia and AMD rumors. That silence leaves Intel as the wildcard in an otherwise two-horse delay narrative.
What Nvidia’s silence signals
Nvidia has a consistent pattern of not commenting on unreleased product timelines, so the absence of a denial or confirmation this week isn’t unusual on its own. What is notable is how long the company has now gone without laying out any public RTX 60-series roadmap at all, even in the vague terms it sometimes uses at GTC or CES keynotes. GamesRadar+ flagged this explicitly, noting Nvidia has never officially confirmed an RTX 60-series launch window in the first place, which means every date currently circulating, 2027 or 2028, originates entirely from the leaker community. By October 2026 Nvidia still had not officially announced a GeForce RTX 6000, while PC Games Hardware reported a possible RTX 6060 launch in 2029, a rumor that pushes the lower end of the stack even further out.
That leaves Nvidia in a comfortable position commercially, even if it’s an awkward one for messaging. Every month without a confirmed successor is another month Blackwell inventory, however expensive, remains the top consumer option, and another month Nvidia’s roughly 90% share of discrete GPU shipments stays largely uncontested at the high end. That pricing power extends to the professional tier, too: VideoCardz tracked the Nvidia Marketplace listing for the RTX PRO 6000 climbing to $13,250 in June 2026, then Tom’s Hardware clocked it jumping again to $16,000 by August 2026 — nearly double its original launch price in barely over a year.
What this means if you’re deciding whether to buy now
For anyone weighing a Blackwell purchase against waiting for the next generation, a 2028 timeline changes the calculation meaningfully compared to a 2027 one. Waiting an extra year for a still-unconfirmed launch date is a harder sell when current RTX 50-series cards, despite the price increases documented above, remain the fastest consumer GPUs available and are unlikely to be meaningfully undercut before Rubin actually ships. Buyers hunting for a workaround have also turned to gray-market and modified cards. Tech Insider reported on modified RTX 5090 cards with 96GB of VRAM selling through Alibaba for roughly $3,888, a sign of how far some buyers are willing to go rather than wait out the memory shortage or a next-gen launch that keeps slipping.
The practical read: nothing about this week’s report changes what’s available to buy today, only the outer bound of how long today’s flagship cards may need to remain the flagship cards.
Market impact for Nvidia, AMD, and GPU retailers
A confirmed multi-year gap between GPU generations would reshape near-term retail and secondary-market dynamics more than it would move Nvidia’s stock price on any single trading day. Extended product cycles tend to support used and gray-market pricing for current-generation cards, since there’s no imminent successor to trigger a resale value drop. That dynamic has already shown up in the data-center segment: Tech Insider reported the Nvidia B200’s resale value climbing to 158% of its launch price, a sign that extended relevance windows push up, not down, what older-but-still-current hardware commands on secondary markets. If Blackwell consumer cards follow a similar pattern through 2027 and into 2028, expect used RTX 5090 and RTX 5080 pricing to stay firmer than historical GPU depreciation curves would predict. The same scarcity premium shows up in the cloud and in the data center: Spheron’s marketplace was pricing RTX PRO 6000 instances at $1.29 per hour in spot pricing as of March 2026, the same month Nvidia listed the RTX PRO 6000 Blackwell Server Edition at 24,064 CUDA cores and 120 TFLOPS of FP32 compute, while CloudRift clocked roughly 8,400 tokens per second running a 30B-parameter AWQ model on that same hardware, per Spheron’s reporting. Nvidia has also been scaling rack-level supply to meet that demand, announcing RTX PRO server systems configurable with 2, 4, or 8 GPUs in August 2025 and confirming that 8-GPU, 4U RTX PRO 6000 server builds were available that same month. Akamai took the same hardware to the edge, announcing thousands of RTX PRO 6000 Server Edition GPUs across 4,400 edge locations in March 2026 and describing it as the first global AI Grid implementation, then launching 8-card RTX PRO 6000 GPU instances in beta on July 1, 2026. Those throughput and deployment figures help explain why renting, not just buying, has become a live option for buyers priced out of outright ownership.
For AMD, a synchronized delay removes the urgency to rush RDNA 5 to market ahead of schedule just to beat Nvidia to a new architecture, since both companies would be competing on refined versions of current silicon (Blackwell refreshes, RDNA 4 successors) rather than clean-sheet designs for at least another year beyond what was previously expected.
Predictions: what happens between now and 2028
- Expect Nvidia to stay silent on any official RTX 60-series date through at least its next two major keynotes, consistent with its pattern of not confirming unreleased consumer GPU timelines.
- Watch for a Blackwell refresh or Super-tier refresh cards in the RTX 50 lineup as a stopgap, a common Nvidia move when a full architecture change is pushed out further than usual.
- Expect continued price volatility on RTX 5090 and RTX 5080 cards tied to GDDR7 and broader DRAM supply, independent of whatever happens with the Rubin timeline.
- If AMD’s RDNA 5 delay is confirmed through official channels, expect AMD to lean harder into pricing and availability as its main competitive argument against Nvidia through 2027.
- Expect more leaker crosstalk and partial corroboration over the coming months as supply-chain contacts continue feeding both kopite7kimi and Kepler-style accounts, gradually sharpening (or contradicting) the current 2028 estimate.
Historical context: why generational gaps keep growing
The two-year GPU generation cadence that defined the Turing, Ampere, and Ada eras was itself already unusual by older industry standards, and it depended on steady node improvements at TSMC alongside predictable memory supply. Blackwell’s slip past that two-year mark in January 2025 was the first sign the pattern was breaking down, driven partly by Nvidia prioritizing AI data-center silicon over consumer parts amid the broader AI infrastructure buildout. A further stretch to 2028 for Rubin-based consumer cards would extend that trend rather than reverse it, suggesting the two-year cadence PC gamers got used to across the 2018-2022 window may not return anytime soon, if leaker estimates hold.
What outlets are and aren’t confirming
It’s worth being precise about what has and hasn’t been verified as this story spreads. TechSpot’s September 21 report and GamesRadar+’s coverage both trace directly back to kopite7kimi and Kepler as the sources, not to Nvidia or AMD. Wccftech, Club386, tbreak.com, and OC3D have all picked up and amplified the same underlying leak in their own coverage, which adds visibility but not independent verification, since they’re drawing on the same two leaker accounts rather than separate sourcing. No outlet in this cycle has published an Nvidia or AMD statement confirming or denying a 2028 timeline.
Where the RTX 6000 Rumor Stands Now
The only documented change is kopite7kimi’s timeline revision from an earlier second-half-2027 expectation to 2028 on September 21, 2026. He wrote that the GR20x gaming-product timeline had been postponed again, which substantially aligns with Kepler_L2’s earlier 2028 view. Reports describe 2028 as an unofficial estimate, not a company-confirmed schedule, so “RTX 6000 series release date: 2028” should not be read as a confirmed date.
The “GeForce RTX 6000” label is being used as a working name for a possible Rubin-generation product, not as a confirmed Nvidia product name. RTX 6060, RTX 6080, and RTX 6090 are speculative names in the reporting and should not be treated as announced models. No official Nvidia prices for RTX 6000 cards can be confirmed.
For shoppers, the GeForce RTX 5090 remains the newest GeForce desktop flagship on sale, according to Nvidia’s graphics card page. Nvidia’s current results do not establish a newer released GeForce desktop GPU than the existing RTX 50-generation products. The newest Nvidia hardware found was RTX Spark PCs, which Nvidia announced as globally available on October 16, 2026 through ASUS, Dell, HP, Lenovo, Microsoft, and MSI, starting at $2,599. Nvidia describes the RTX Spark laptops as bringing Blackwell RTX graphics and up to 128GB of unified memory, which makes them a separate product category from a future GeForce desktop generation. Until Nvidia publishes a dated announcement, treat every RTX 6000 date as a leak and plan purchases around what is on sale today.
Frequently asked questions
Has Nvidia confirmed the RTX 6000 series will launch in 2028?
No. Nvidia has never officially confirmed a launch window for the RTX 60 series. The 2028 estimate comes from leakers kopite7kimi and Kepler, as reported by TechSpot and GamesRadar+ on September 21, 2026.
What architecture will the RTX 6000 series use?
Leaker reporting ties the RTX 6000 series to Nvidia’s Rubin architecture, the successor to the current Blackwell generation used in the RTX 50 series.
Is AMD’s next GPU generation also delayed to 2028?
Leaker Kepler claims AMD’s RDNA 5-based Radeon 10000 series is also tracking for 2028, according to the same TechSpot report. AMD has not confirmed this.
Should I buy an RTX 5090 now or wait for the RTX 6000 series?
If the 2028 timeline holds, waiting means holding off for roughly two more years. Given current RTX 50-series cards remain the fastest consumer GPUs available despite elevated pricing, most buyers who need a GPU now have little reason to delay a purchase based on this rumor alone.
Why is the RTX 6000 series reportedly delayed?
No official reason has been given by Nvidia. Industry reporting points to broader GDDR7 and DRAM memory supply constraints, alongside AI data-center demand competing for the same fab and memory capacity, as plausible contributing factors, though this remains unconfirmed.
Who are kopite7kimi and Kepler?
Both are pseudonymous leakers who cover graphics card rumors. TechSpot describes kopite7kimi as one of the most closely followed leakers in the GPU space, with a track record built across several previous Nvidia architecture cycles.
How does this compare to past Nvidia GPU generation gaps?
Nvidia moved on a roughly two-year cadence from Turing (2018) through Ada Lovelace (2022). Blackwell broke that pattern by launching in January 2025, over two years after Ada. A 2028 Rubin launch would stretch the gap to around three years, which GamesRadar+ described as the longest gap between GPU generations in 20 years.
Will RTX 50-series prices come down before the RTX 6000 series launches?
That depends largely on GDDR7 and DRAM supply normalizing, which is a separate issue from the Rubin launch timeline. Current reporting shows RTX 50-series prices still well above launch MSRP through September 2026, with no clear signal of near-term relief.
