Nvidia spent Computex 2026 talking about a chip most buyers still cannot get their hands on, then confirmed two more that do not exist yet. At a briefing in Taipei this week, CEO Jensen Huang told reporters that Nvidia’s RTX Spark platform, the company’s first dedicated consumer PC chip line, already has two successors in active development. They are tracked internally as N2X and N3X. Months later, as of September 2026, that status has not moved: Nvidia has still not released any N2X or N3X hardware, and no specs, pricing, or release dates have been attached to either chip. Huang also cracked open a door hardware makers have been knocking on since RTX Spark first leaked: whether Nvidia would ever build a chip for a gaming handheld.
The comments, first reported by Tom’s Guide after a sit-down interview with Huang, landed days after Intel unveiled its own dedicated handheld chips, the Arc G3 and Arc G3 Extreme, built specifically to challenge AMD’s Ryzen Z2 Extreme in the fast-growing handheld gaming PC category. Nvidia has no shipping handheld chip and no announced plan to build one. But confirming a two-generation roadmap before the first RTX Spark devices have fully reached store shelves is a strong signal from a company positioning itself for a fight it has not formally entered, one PCWorld’s Computex 2026 wrap-up ranked among the show’s most consequential hardware stories.
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 Nvidia Confirmed About N2X and N3X at Computex 2026
RTX Spark is Nvidia’s first Arm-based system-on-chip built for laptops, mini PCs and desktop-class “AI PC” systems. It is not designed for phones, and until this week, Nvidia had said nothing official about what comes after it. Internally, the chip carries the codename N1X. During the Computex interview, Huang confirmed that N2X and N3X are already planned as the next two generations of the same architecture, well before N1X has finished its retail rollout.
Huang framed the early confirmation as a statement of intent rather than a product announcement. He told Tom’s Guide that Nvidia intends to “expand our family” of consumer Arm chips for years, comparing the commitment to a long-running product line rather than a one-off experiment. No specifications, release windows or pricing were shared for either chip. Nvidia’s public messaging so far has focused entirely on N1X, which began reaching partner devices earlier this year.
Decoding Nvidia’s Chip Lineup: N1, N1X, N2X and N3X
The naming convention has confused plenty of buyers, so it helps to break it into what ships today and what exists only as a roadmap commitment.
The Chips Already Shipping
N1X is the chip sold under the RTX Spark name. Per specs reported by Trusted Reviews and corroborated by Computex coverage, it packs up to 20 Arm CPU cores alongside a Blackwell-generation GPU with roughly 6,144 CUDA cores, paired with as much as 128GB of unified LPDDR5X memory running at around 300GB/s of bandwidth, which Nvidia rates as 1 petaflop of AI compute. Nvidia has claimed the chip can hit 100 frames per second at 1440p resolution in current games, though that figure comes from the company’s own marketing rather than independent lab testing. A smaller companion variant called N1 also exists, aimed at lower-cost or more compact machines, and as of June 2026, Yellow.com reports the first-generation Spark lineup actually scales across a much wider band than the flagship spec sheet suggests, from 16GB of memory in entry configurations up to that 128GB ceiling, with fully loaded builds pricing out around $3,000. Microsoft’s Surface Laptop Ultra and several Asus systems already ship with RTX Spark silicon inside, and Nvidia says laptops and compact desktops from ASUS, Dell, HP, Lenovo, Microsoft Surface, MSI, Acer, and GIGABYTE are due to arrive this fall, expanding RTX Spark from a two-OEM launch into a mainstream, multi-vendor AI PC push.
The Chips Still on the Roadmap
N2X and N3X sit on the other side of the line. Nvidia itself has confirmed only that they exist as planned successors, with no core count, no GPU configuration and no official target device category attached to either name. Outside reporting has started to fill in pieces of that gap, though: EveTech reported in May 2026 that early N2X notebook designs are targeting more than 18 hours of mixed-use battery life, up from the 14 to 15 hours typical of current N1X laptops, at a premium of roughly R3,500 to R5,000 over the R22,000-to-R35,000 range N1X machines already occupy. Separate coverage from Gigazine and Tom’s Guide around Computex 2026 corroborated that N2X and N3X are being treated internally as the second- and third-generation RTX Spark SoCs for Windows AI PCs, while press reports citing Nvidia’s own roadmap, via ComputerBase and Hardwareluxx in January 2026, pointed to a notebook-class N2X SoC targeted for roughly Q4 2027. None of that is Nvidia-confirmed, and Huang has been careful not to attach a date publicly, describing RTX Spark buyers as people who will keep these systems “in their homes for 5 to 10 years,” comparing them to home theater equipment bought once and used for a decade, which suggests Nvidia is building a slow-moving platform rather than rushing annual refreshes. As of September 2026, Nvidia itself still classifies N2X and N3X as undisclosed roadmap items, with no pricing, specs, or handheld implementation confirmed, and outside analysts continue to say a handheld built on either chip is unlikely before 2027 at the earliest.
Nvidia’s RTX Spark Roadmap at a Glance
The table below lays out what is confirmed, what is rumored, and what remains a total unknown across Nvidia’s current and future Arm-based consumer lineup.
| Codename | Status (September 2026) | Target Device | Key Spec Highlight |
|---|---|---|---|
| N1 | Shipping | Compact, lower-cost mini PCs | Reduced core and memory config vs. N1X |
| N1X (RTX Spark) | Shipping | Laptops, mini PCs, AI PCs | Up to 20 Arm cores, ~6,144 CUDA cores, up to 128GB LPDDR5X, 1 petaflop AI compute |
| N2X | Confirmed, no specs | Not yet disclosed | Not disclosed |
| N3X | Confirmed, no specs | Not yet disclosed | Not disclosed |
| Handheld variant | Unconfirmed, rumored only | Gaming handheld, if built | Nvidia says it “would work with” an OEM partner, with no internal commitment made |
Jensen Huang’s “R2-D2” Pitch for RTX Spark
Huang’s most memorable line from the interview was not about specs at all. He described RTX Spark as closer to “R2-D2” than a laptop CPU, an image meant to convey a small, semi-autonomous, always-on companion device rather than a component buried inside a case. That framing matters for how Nvidia wants the platform understood. RTX Spark is being pitched less as a gaming chip and more as a persistent local AI system, one that happens to also run games well.
That distinction explains why Nvidia has been cautious about handheld gaming specifically. A gaming handheld demands sustained thermal performance in a tiny chassis, tight power budgets, and software compatibility with a games library built almost entirely around Windows and x86 hardware. An always-on AI companion chip does not face the same constraints. Bridging the two is a harder engineering problem than the roadmap slide suggests.
Why an RTX Gaming Handheld Suddenly Looks Possible
Until this week, Nvidia’s involvement in handheld gaming ran through Nintendo. Custom Tegra chips power the Switch and Switch 2, but those are closed, first-party designs Nvidia does not sell to outside PC makers. A Windows-compatible RTX Spark derivative would be a genuinely new product category for the company, competing directly for the same OEM design wins that Intel and AMD currently split between them.
Asked directly about a handheld version of RTX Spark, Huang stopped short of committing Nvidia to building one, but he did not dodge the question either. He told Tom’s Guide in June 2026 that he is “willing to work on” a Spark-based gaming handheld if a hardware partner wanted to build one around the chip, pointing to the Spark lineage’s role inside the current Nintendo Switch and Switch 2 as evidence Nvidia’s silicon already has a track record in handheld gaming form factors, even if that relationship remains exclusive to Nintendo. That framing effectively makes OEM appetite the limiting factor rather than technical readiness, and Nvidia has been explicit that the remarks are conditional rather than a product announcement: as of September 2026, no OEM partnership for a Spark-based handheld has been confirmed publicly. TechRadar’s own reporting went further, suggesting that a hypothetical Nvidia handheld chip could eventually approach RTX 5070-level graphics throughput if Nvidia chose to build one, while cautioning that pricing, not performance, would be the real obstacle to a shipping product.
The Anti-Cheat Problem Nvidia Still Has to Solve
Performance and pricing are not the only barriers. Huang and Nvidia’s own briefings flagged anti-cheat software compatibility as a real, unresolved challenge for scaling gaming on Arm-based architecture. Most competitive multiplayer titles rely on kernel-level anti-cheat systems built and certified for x86 processors. Getting major publishers to validate those systems on an entirely different instruction set is a slow, publisher-by-publisher process that Nvidia does not fully control.
Valve has already run into a version of this problem with SteamOS and its own Steam Frame and Steam Machine hardware, where certain anti-cheat-protected titles remain unsupported years after the original Steam Deck’s launch. Nvidia entering with an Arm-based chip, rather than the x86 designs AMD and Intel both use, would restart that compatibility clock from closer to zero.
The Handheld Chip War Nvidia Would Be Entering
Nvidia would not be walking into an empty market. AMD’s Ryzen Z2 Extreme currently powers most premium handhelds on shelves today, built around eight cores split between three full Zen 5 cores and five efficiency-focused Zen 5c cores, paired with a 16-compute-unit Radeon 890M-class integrated GPU. Intel spent Computex 2026 attacking that position directly, launching the Panther Lake-based Arc G3 and Arc G3 Extreme chips, purpose-built for handhelds rather than adapted from laptop silicon. Intel’s flagship Arc G3 Extreme pairs 14 CPU cores with a 12-core Arc B390 GPU, and outlets including The FPS Review have reported early tests showing it beating the Z2 Extreme by a wide margin in gaming workloads at matched power levels.
That leaves two established, x86-based competitors already shipping real hardware into a category Nvidia has only just started hinting at. PC Gamer’s live Computex coverage captured just how crowded the show floor was with competing handheld silicon, from AMD’s incumbent chips to Intel’s new lineup. Any Nvidia handheld chip would need to justify itself against two rivals with a head start, existing OEM relationships, and full Windows driver maturity.
Handheld Gaming Chip Specs: AMD vs Intel vs Nvidia
Here is how the current and previous-generation handheld silicon lines up against Nvidia’s still-hypothetical entry.
| Chip | Maker | CPU Cores | GPU | Status |
|---|---|---|---|---|
| Ryzen Z1 Extreme | AMD | 8 cores / 16 threads (Zen 4) | Radeon 780M, 12 CU | Previous gen, powers original ROG Ally |
| Ryzen Z2 Extreme | AMD | 8 cores / 16 threads (3 Zen 5 + 5 Zen 5c) | Radeon 890M-class, 16 CU | Shipping, powers ROG Ally X and Legion Go family |
| Arc G3 | Intel | 14 cores (2P + 8E + 4 LPE) | Arc B370, 10 Xe3 cores | Shipping June 2026 |
| Arc G3 Extreme | Intel | 14 cores (2P + 8E + 4 LPE) | Arc B390, 12 Xe3 cores | Shipping June 2026, MSI Claw 8 EX AI+, Acer Predator Atlas 8 |
| RTX Spark (N1X) | Nvidia | Up to 20 Arm cores | Blackwell, ~6,144 CUDA cores, 1 petaflop AI compute | Shipping for laptops and mini PCs only, no handheld config exists |
Market Impact: What This Means for AMD and Intel
Search interest tells part of the story. Query volume for “Nvidia RTX Spark” climbed from under 900 searches in May to roughly 27,000 in June 2026, according to keyword-tracking data, a spike almost entirely driven by Computex coverage and the N2X and N3X roadmap news specifically. That kind of attention has real value for a company trying to establish a new hardware category from scratch.
For AMD, the immediate risk is limited. Nvidia has not shipped a competing product, and Ryzen chips remain the default choice across most handheld gaming PCs sold today. The bigger risk is longer-term and psychological: OEMs currently locked into AMD or Intel handheld designs now have a third option to evaluate before their next product cycle, even if that option is years away from shipping. For Intel specifically, the timing stings more. The company just spent Computex 2026 trying to convince the same OEMs, MSI, Acer and others, that Arc G3 Extreme deserves design wins away from AMD. A credible Nvidia rumor complicates that pitch before Intel’s own handheld chips have even finished their first retail cycle.
RTX Spark’s Real Rivals: Apple, Qualcomm and Nvidia’s Paused Gaming GPUs
Remember that RTX Spark was never designed purely as a gaming chip. Nvidia has positioned it against Apple Silicon and Qualcomm’s Snapdragon X line in the broader AI PC category, where the pitch is local AI processing and battery efficiency rather than frame rates. Handheld gaming, if it happens at all, would be a secondary use case built on top of a chip designed for an entirely different fight.
That context matters more given Nvidia’s unusual 2026 on the desktop side. The company skipped a new generation of discrete gaming GPUs entirely this year, breaking a three-decade product cadence. Reading the N2X and N3X confirmation alongside that pause suggests a company reallocating engineering and manufacturing capacity away from traditional desktop graphics cards and toward the Arm-based AI PC platform and its next-generation datacenter chips instead. A handheld chip would fit naturally into that shift, since it reuses the same architecture rather than requiring a new discrete GPU line.
Nvidia’s Rubin Platform Shows Where the Engineering Capacity Is Going
The N2X and N3X roadmap confirmation reads differently once it sits next to Nvidia’s datacenter timeline. As of August 2026, Nvidia says its next-generation Rubin architecture, the successor to Blackwell, is now in full production, with Rubin-based systems shipping from partners in the second half of 2026. Japanese outlet Gigazine’s June 2026 Computex coverage went further on the consumer side, reporting a longer-range roadmap in which a Rubin-generation Spark chip follows the current Blackwell-based N1X in 2027, with a Feynman-based Spark generation pencilled in for 2029. That is the same window Nvidia has attached to its gaming GPU pause and to the still spec-free N2X and N3X commitments, and it is unlikely to be a coincidence. Rubin is where Nvidia’s fabrication capacity, packaging allocation and engineering attention are concentrated this year, well ahead of anything on the consumer roadmap slide.
The scale gap is the clearest evidence of that priority. Nvidia says its Vera Rubin NVL144 rack-scale system, also shipping in the second half of 2026, delivers 3.6 exaflops of FP4 compute and 1.2 exaflops of FP8 compute per rack, aimed at hyperscalers and cloud providers training and serving frontier AI models. RTX Spark, by comparison, is rated at roughly 1 petaflop of AI compute in a chip meant for a laptop or a desk, not a data center rack. Exaflop-scale systems and petaflop-scale consumer silicon are not fighting over the same silicon budget in any direct sense, but they are competing for the same fabrication capacity and the same slice of Nvidia’s roadmap attention, and Rubin is clearly winning that allocation fight in 2026.
That context helps explain a pattern that otherwise looks strange: Nvidia confirmed N2X and N3X exist, in public, with zero specifications, release windows or pricing attached. A company with spare roadmap bandwidth to spend on consumer hardware would likely have shared more. Instead, Huang’s “expand our family” language reads as a placeholder announcement, enough to keep OEMs and reviewers engaged with the RTX Spark line while Rubin absorbs the bulk of Nvidia’s actual production capacity. It also helps explain why Nvidia declined to commit to a handheld chip during the same interview. Building a third silicon variant for a new device category would draw further against the same finite capacity Rubin is currently consuming, and Nvidia has given no indication it is ready to make that trade before Rubin’s second-half-2026 rollout is complete.
For anyone tracking the RTX Spark line specifically, the practical read is that N2X and N3X are unlikely to arrive with meaningful detail until Rubin’s initial production run is further along. Nvidia has not said so directly, but a company running Rubin at full production while still withholding basic specs for its next consumer Arm chip has already shown which platform gets built first.
Historical Context: Nvidia’s Uneven Handheld Track Record
Nvidia has tried handheld gaming hardware before, with mixed results. The Shield Portable launched in 2013 as an Android-based handheld with a fold-out screen and physical controls, and it never found a mass audience before Nvidia discontinued the line. Tegra chips found far more success indirectly, powering Nintendo’s Switch since 2017 and the Switch 2 more recently, but that relationship is exclusive to Nintendo and does not translate into open sales to other PC OEMs.
RTX Spark represents a different approach entirely. Rather than building a first-party handheld again, Nvidia is building a chip platform and letting the roadmap tease do the marketing work, watching to see which OEM moves first. It is a lower-risk strategy than another Shield-style device, but it also means Nvidia is not guaranteed any specific handheld will ever ship, regardless of how much interest the N2X and N3X news generates.
Competitive Comparison: Can Nvidia Actually Win Handhelds?
Nvidia’s biggest advantage is brand pull. The RTX name carries more consumer recognition in gaming than any AMD or Intel chip label, and Nvidia’s DLSS upscaling technology has historically outperformed AMD’s FSR and largely matched Intel’s newer XeSS 3 suite. If Nvidia brings that same upscaling advantage to a handheld chip, it arrives with a real, immediately obvious selling point rather than needing years to build a reputation.
Its biggest disadvantage is everything else. AMD and Intel both ship x86 chips with mature Windows drivers and full anti-cheat certification built over years of iteration. Nvidia’s Arm architecture would need publishers to re-certify anti-cheat systems game by game, a process that could take years even with strong OEM demand. Pricing is the other open question. Nvidia’s existing GPUs already sit at a premium versus AMD equivalents, and nothing in the RTX Spark launch suggests Nvidia plans to compete on price rather than performance.
5 Predictions for Nvidia’s Next Move
- N2X ships first as a laptop and mini PC refresh, not a handheld chip, likely in 2027, continuing the AI PC focus rather than pivoting to gaming hardware.
- Nvidia partners with an existing OEM, most plausibly Asus or MSI, rather than building a first-party handheld itself, mirroring the “we’ll work with them” language from Huang’s interview.
- Any RTX-branded handheld that does emerge launches at a higher price than current Z2 Extreme or Arc G3 Extreme devices, trading affordability for brand and upscaling quality.
- Anti-cheat compatibility remains the single biggest blocker cited by both Nvidia and outside reviewers through at least 2027, limiting any early Arm-based handheld to single-player and indie libraries.
- AMD and Intel respond by accelerating their own roadmaps rather than cutting prices, given that Nvidia has not shipped a competing product and has years of runway before it needs to.
What Happens Next: Timeline and Open Questions
Nothing here is scheduled. Nvidia has not attached a date to N2X, N3X, or any handheld variant of either chip, and the company’s own framing suggests that is intentional. Huang’s comparison of RTX Spark to home theater equipment kept “for 5 to 10 years” points toward slower, more deliberate product cycles than the near-annual refreshes typical of gaming GPUs. Expect incremental N2X details to surface at CES 2027 at the earliest, with any handheld-specific news likely to come from an OEM partner announcement rather than from Nvidia directly, a pattern Tom’s Hardware’s Computex retrospective noted across most of this year’s chip roadmap reveals.
The open questions that matter most: whether any OEM publicly commits to building an RTX-powered handheld, whether Nvidia solves anti-cheat compatibility before or after a hardware launch, and whether N2X targets gaming performance at all or simply extends the AI PC pitch that defined RTX Spark’s first generation.
Frequently Asked Questions
What is Nvidia RTX Spark?
RTX Spark is Nvidia’s first Arm-based system-on-chip for consumer laptops, mini PCs and AI PCs, internally codenamed N1X. It combines up to 20 Arm CPU cores with a Blackwell-generation GPU and up to 128GB of unified memory.
What are N2X and N3X?
They are the next two generations of Nvidia’s RTX Spark architecture, confirmed by CEO Jensen Huang at Computex 2026. As of September 2026, no specifications, devices, pricing or release dates have been announced for either chip, and Nvidia continues to treat them as undisclosed roadmap items.
Is Nvidia building a gaming handheld?
Not officially. Huang said Nvidia would work with an OEM partner interested in building one, but the company has not committed to designing or shipping a handheld device itself.
How does RTX Spark compare to Intel Arc G3 Extreme and AMD Ryzen Z2 Extreme?
RTX Spark is not currently configured for handhelds, so a direct gaming comparison is not yet possible. Arc G3 Extreme and Ryzen Z2 Extreme are both shipping in real handheld devices today, with Intel’s chip reported to outperform AMD’s in early gaming benchmarks at matched power levels.
When will N2X and N3X launch?
Nvidia has not given an official timeline. Based on Huang’s comments about long product life cycles, industry watchers expect the earliest N2X details around CES 2027, though press reports citing Nvidia’s roadmap, via ComputerBase and Hardwareluxx in January 2026, have pointed to a notebook-class N2X SoC arriving around Q4 2027, with a Rubin-generation Spark chip to follow in 2027 and a Feynman-based Spark generation reportedly slated for 2029; none of this is confirmed by Nvidia, so treat it as speculative.
Why did Jensen Huang compare RTX Spark to R2-D2?
Huang used the comparison to frame RTX Spark as a small, semi-autonomous AI companion device rather than a conventional laptop processor, reflecting Nvidia’s positioning of the chip around local AI tasks rather than gaming alone.
Will anti-cheat software work on Nvidia’s Arm-based chips?
Not consistently today. Most kernel-level anti-cheat systems are built for x86 processors, and Nvidia has acknowledged that publisher-by-publisher re-certification for Arm architecture remains an unresolved challenge.
What does this mean for AMD and Intel’s handheld chip business?
No immediate impact, since Nvidia has not shipped a competing product. The longer-term effect is added uncertainty for OEMs weighing their next handheld design partner, at a moment when Intel is still trying to win business away from AMD’s incumbent Ryzen Z2 Extreme.
What is Nvidia’s Rubin platform, and is it related to RTX Spark?
Rubin is Nvidia’s next-generation datacenter AI architecture, the successor to Blackwell, built for AI training and inference at hyperscale rather than for consumer devices. Nvidia says Rubin is now in full production, with systems including the Vera Rubin NVL144 rack, rated at 3.6 exaflops of FP4 compute and 1.2 exaflops of FP8 compute, shipping from partners in the second half of 2026. It does not share silicon with RTX Spark or the N2X/N3X consumer roadmap, but both platforms compete for the same Nvidia fabrication capacity and engineering attention.
For more hardware coverage, follow Tech Insider’s ongoing reporting on the 2026 handheld gaming chip race.
