![Who Makes GPUs Besides Nvidia, AMD? 90% Share Gap [2026]](https://tech-insider.org/wp-content/uploads/2026/10/tmp5ce1kemt.webp)
Nvidia and AMD dominate every headline about graphics chips, but neither company is the only name in the GPU business. A report published by BGR on October 1, 2026 revisited a question that keeps resurfacing in search trends and forum threads: who actually makes GPUs besides Nvidia and AMD? The answer, it turns out, spans a bigger and stranger map than most buyers realize, from Intel’s long-delayed Arc cards to a cluster of Chinese chipmakers racing around U.S. export bans.
For consumers shopping for a new graphics card, the practical list is short: Nvidia, AMD, and Intel. But widen the lens to include mobile chips, licensed GPU designs, and the AI accelerators reshaping data centers, and the list of companies building graphics silicon grows to more than a dozen. This piece breaks down who they are, what they actually ship today, and why most of them still can’t dent Nvidia’s grip on the market.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Why this question is trending again in October 2026
BGR’s analysis landed at a moment when Nvidia’s share of the discrete graphics card market is still climbing in some quarters and sliding slightly in others, depending on which data point you grab. According to Jon Peddie Research figures reported by Club386, Nvidia held roughly 90% of desktop add-in-board GPU shipments in the first half of 2026, down from a peak near 94% in the fourth quarter of 2025 as AMD’s RDNA 4 cards started showing up in retail numbers. AMD sat around 8%, and Intel’s Arc lineup held roughly 1% to 2%.
Those numbers explain why “who else makes GPUs” keeps resurfacing as a search query. Enthusiasts priced out of the latest RTX cards, IT buyers hedging against supply constraints, and investors looking for the next AI-chip story are all asking the same thing: is there a credible alternative to the Nvidia-AMD duopoly, and if so, who is it? Tech-insider.org has tracked Nvidia’s shipment dominance closely, including a report on Nvidia’s 90% GPU share as shipments hit 12.5 million units earlier this year.
Intel Arc: the closest thing to a third option
Intel is the only company besides Nvidia and AMD that sells a complete, boxed, retail discrete GPU lineup for PCs. Its current Battlemage generation includes the Arc B580, which launched at $249 with 20 Xe2 cores and 12GB of GDDR6 memory, and the Arc B570, which followed at $219 with 18 Xe2 cores and 10GB of GDDR6, according to Tom’s Hardware. Intel also folds Arc technology into integrated graphics on its Core processors, branded Arc 130 and Arc 140 in current mobile chips, per BGR’s reporting.
Intel’s GPU ambitions did not start smoothly. Its first attempt at a standalone graphics chip, the Larrabee project, was scrapped after years of delays and performance problems, BGR noted. Arc itself needed more than a year of driver updates after its 2022 launch before it became a credible budget option. That history matters because it shows how hard the GPU business is to enter, even for a company with Intel’s manufacturing scale and decades of integrated-graphics experience.
The relationship between Intel and Nvidia has also gotten more complicated. The two companies announced in September 2025 that Nvidia would invest $5 billion in Intel common stock, with Intel agreeing to build x86 PC system-on-chips that incorporate Nvidia RTX GPU chiplets. That deal effectively makes Nvidia both a competitor and an investor in Intel’s chip business, blurring the line between rival and partner in a way the GPU market hasn’t seen before. For a closer look at how Arc stacks up against mainstream cards, see tech-insider’s Arc B580 vs RTX 5060 vs RX 9060 XT comparison.
Moore Threads, Biren, and MetaX: China’s answer to export bans
The most geopolitically charged corner of the GPU market belongs to a handful of Chinese chip designers trying to build domestic alternatives after Washington cut off access to Nvidia’s most advanced accelerators. Moore Threads, founded in 2020, released its MTT S80 consumer graphics card in 2022. The card lacked DirectX 11 support until 2023 and only gained DirectX 12 compatibility through a beta driver in 2024, according to BGR, underscoring how far behind Western GPU software stacks the company still is on the consumer side.
Moore Threads and Biren Technology, which builds the BR100-series data-center accelerators, were both added to the U.S. Commerce Department’s Entity List in October 2023 as part of a broader set of additions covering companies involved in developing advanced computing chips, per the Federal Register notice. The designation restricts their access to U.S.-origin chipmaking tools, design software, and in practice, the advanced foundry capacity at companies like TSMC that both firms had relied on. MetaX, which develops MX-series GPUs for data-center AI workloads, has not been added to the Entity List but remains exposed to the same export-control risk given its position in China’s domestic chip supply chain.
The broader export-control story goes back further. U.S. rules restricting shipments of Nvidia’s A800 and H800 chips to China, and blacklisting several Chinese GPU developers, were detailed by Tom’s Hardware as part of the same regulatory push. The practical effect has been to turn companies like Moore Threads, Biren, and MetaX into national-champion projects: not globally competitive on performance per dollar, but strategically important enough that Beijing has reportedly encouraged continued investment regardless of near-term returns.
Other Chinese players round out the picture without yet being commercially significant outside China. Innosilicon has shown off its Fantasy series GPUs aimed at desktop and compute workloads. Zhaoxin builds x86-compatible processors with integrated graphics for the Chinese PC and government market under its KaiXian brand. Loongson integrates graphics into its domestic 3A processor family, used mostly in government and secure-computing deployments. None of these three have published verified 2025-2026 shipment or revenue figures that would let them be compared directly to Nvidia, AMD, or Intel.
Qualcomm and MediaTek: the GPUs already in your pocket
Outside the discrete-card conversation, Qualcomm is arguably the second-biggest GPU designer on the planet by unit volume. Its Adreno graphics architecture ships inside every Snapdragon mobile chip and now inside the Snapdragon X Elite platform used in a new wave of Windows laptops. Adreno doesn’t compete with RTX or Radeon cards directly, but it processes more rendered pixels per day globally than either company, simply because it’s embedded in hundreds of millions of phones. Tech-insider has covered Qualcomm’s mobile graphics push in depth, including how Adreno Neural Fusion cuts mobile gaming power draw and how the Snapdragon X2 Elite compares to Intel on the laptop side.
MediaTek takes a different approach: rather than design its own GPU architecture from scratch, it licenses Arm’s Mali reference designs and tunes them for its Dimensity smartphone chips, according to BGR’s reporting. Samsung has followed a hybrid path of its own, using Arm’s Mali graphics in some Exynos chipsets and AMD’s RDNA architecture, licensed for mobile use, in others. Apple sits apart from all of them, designing fully proprietary GPUs for the iPhone, iPad, and Mac lineups rather than licensing anyone else’s architecture.
Arm and Imagination Technologies: the GPUs you never see named
Arm doesn’t sell a single physical chip. It licenses GPU architectures, Mali for mainstream mobile silicon and Immortalis for premium phones with hardware-accelerated ray tracing, to chipmakers who then build and brand their own processors around that design. It’s a licensing model, not a manufacturing one, but it means Arm’s graphics IP ends up inside an enormous share of the world’s smartphones, tablets, and embedded devices without most buyers ever seeing the Arm name on a spec sheet. Tech-insider has covered Arm’s push beyond mobile into data-center silicon in its report on Arm’s AGI CPU.
Imagination Technologies runs a similar playbook with its PowerVR GPU architecture, licensing designs to chipmakers for use in mobile, embedded, automotive, and FPGA-based products rather than selling finished chips under its own name. Both companies measure success in licensees and royalty revenue rather than discrete-GPU market share, which is part of why they rarely come up when people search for Nvidia and AMD alternatives, even though their designs are everywhere.
Huawei: the GPU maker hiding inside an AI chip strategy
Huawei occupies an unusual spot on this list because it isn’t trying to sell a conventional graphics card at all. Its Kirin smartphone processors include Huawei-developed integrated graphics, while its separate Ascend line targets data-center AI training and inference, directly in the lane Nvidia dominates with its data-center GPUs. Tech-insider previously reported on the Ascend 950PR chip’s attempt to challenge Nvidia in China’s domestic AI accelerator market. Huawei has publicly framed combining its Kirin and Ascend silicon as a push toward self-reliant, on-device computing, independent of U.S. chip supply chains.
Like Moore Threads and Biren, Huawei’s accelerator business sits inside the broader U.S.-China chip conflict. Export controls on advanced manufacturing equipment and design tools continue to constrain how far ahead Huawei, or any Chinese chipmaker, can push its silicon without access to the most advanced foundry nodes available to Nvidia and AMD.
GPU market share: the numbers behind the duopoly
However many names show up on a list of GPU makers, the shipment data makes clear that only three companies currently matter in the retail discrete-card market, and one of those three barely registers.
| Company | Approx. discrete GPU shipment share (H1 2026) | Primary market | Flagship current product |
|---|---|---|---|
| Nvidia | ~90% | Gaming, workstation, data-center AI | GeForce RTX 50-series / data-center Blackwell GPUs |
| AMD | ~8% | Gaming, data-center AI | Radeon RX 9000-series |
| Intel | ~1-2% | Budget gaming, integrated graphics | Arc B580 / B570 (Battlemage) |
| Moore Threads, Biren, MetaX (combined) | Not independently reported for global retail shipments | China domestic gaming and data-center AI | MTT S80 / BR100-series / MX-series |
Source: Jon Peddie Research figures as reported by Club386; company product data via Tom’s Hardware and BGR.
That 90% figure is itself a moving target. It climbed as high as 94% in the fourth quarter of 2025 and eased back toward 90% in early 2026 as AMD’s newer Radeon cards gained some shelf presence, illustrating how even AMD’s modest gains can shift the headline number by several points without threatening Nvidia’s overall control of the category.
A second map: who makes GPUs by segment, not just market share
Market share figures only tell part of the story because they cover retail discrete cards, not the much larger universe of integrated and licensed graphics silicon. The table below sorts GPU designers by the segment they actually compete in, which is a more useful way to understand where each company has real leverage.
| Segment | Leading or notable companies | Business model |
|---|---|---|
| Discrete gaming/workstation GPUs | Nvidia, AMD, Intel | Design and sell complete GPU chips to board partners |
| Data-center AI accelerators | Nvidia, AMD, Huawei (Ascend), Biren, MetaX | Sell or deploy accelerator chips for AI training/inference |
| Mobile integrated GPUs | Qualcomm (Adreno), Apple, MediaTek, Samsung, Huawei (Kirin) | Design proprietary or licensed GPUs embedded in SoCs |
| Licensed GPU IP | Arm (Mali/Immortalis), Imagination Technologies (PowerVR) | License architecture designs to chipmakers for royalties |
| China domestic consumer/enterprise GPUs | Moore Threads, Innosilicon, Zhaoxin, Loongson | Build GPUs or integrated graphics for China’s domestic market |
Board partners: the names on the box aren’t always the GPU maker
Part of the confusion around “who makes GPUs” comes from retail packaging. Companies like ASRock, Gigabyte, Sparkle, and formerly EVGA don’t design GPU chips themselves; they license reference designs from Nvidia, AMD, or Intel and build the finished card around that silicon, adding their own cooling, power delivery, and warranty support. EVGA’s exit from the graphics card business, which tech-insider covered in a report citing a former employee’s account of Nvidia’s dealings with board partners, is a reminder of how much leverage Nvidia holds over the companies that actually put cards on store shelves, even though none of them are GPU designers in the traditional sense.
How we got here: a short history of the GPU duopoly
Nvidia’s dominance isn’t new. The company has held the majority of graphics card shipments since 2006, the year it introduced its CUDA parallel-computing platform, according to BGR. That decision to open CUDA to general-purpose computing a decade before AI training became a trillion-dollar industry is widely credited as the single biggest reason Nvidia has the software ecosystem advantage it holds today. AMD didn’t launch its open-source ROCm computing platform, built to compete directly with CUDA, until 2016, a full ten-year head start Nvidia has never fully given back.
Nvidia’s path hasn’t been free of controversy either. The U.S. Securities and Exchange Commission fined the company in 2018 for misrepresenting how much of its revenue growth came from cryptocurrency mining demand rather than gaming demand, per BGR’s reporting, an early sign of how volatile and hard to forecast crypto-driven GPU demand could be for the whole industry, a pattern that repeated itself during later mining booms.
Competitive comparison: can anyone actually catch Nvidia?
Intel has the clearest shot at meaningfully growing its discrete-GPU share, not because Arc is winning head-to-head benchmarks, but because Intel already has the fabrication relationships, driver engineering teams, and OEM laptop partnerships to keep iterating. Its $5 billion tie-up with Nvidia on x86 chiplets is a hedge, not a surrender: Intel still says it’s continuing to execute on its own GPU roadmap alongside the Nvidia collaboration.
AMD remains the only company with a complete, competitive high-end gaming GPU lineup besides Nvidia, but its retail market share has stayed in the single digits for years despite competent hardware, a gap analysts generally attribute to Nvidia’s software ecosystem (DLSS, CUDA, and ray-tracing tooling) rather than raw silicon performance.
China’s domestic GPU makers face a different kind of ceiling. Export controls have cut Moore Threads and Biren off from TSMC’s most advanced nodes, forcing reliance on domestic foundries like SMIC that trail several process generations behind. That means even well-funded companies like these are, for now, building GPUs that solve a supply-chain problem for Chinese buyers rather than competing on raw performance with Nvidia’s or AMD’s latest chips.
Market impact: what this means for buyers and investors
For PC buyers, the practical takeaway hasn’t changed much: Intel’s Arc cards remain the only real third option if you want a boxed discrete GPU outside Nvidia or AMD, and they’re competitive mainly at the budget end of the market below $250. For enterprise and AI infrastructure buyers, the calculus is different. Nvidia’s data-center GPU dominance has pushed hyperscalers and governments alike to fund alternatives, which is part of why Huawei’s Ascend line and China’s “GPU dragons” keep attracting capital despite lagging performance, since the bigger goal is supply-chain independence rather than immediate competitiveness.
Investors watching the space should note that Nvidia’s near-monopoly cuts both ways: it supports extraordinary margins, but it also means any crack in that dominance, whether from Intel’s roadmap, AMD’s next architecture, or a surprise jump in Chinese domestic chip performance, could move the stock sharply given how much of Nvidia’s valuation already assumes continued dominance.
Predictions: where the alternative-GPU market goes from here
Based on current trajectories reported across the sources above, a few outcomes look likely over the next one to two years:
- Intel will keep Arc alive as a budget-tier option rather than abandon the business, leaning on its Nvidia chiplet partnership to offset R&D costs while it continues its own GPU roadmap.
- AMD’s retail share will likely stay in the single digits unless it closes the software-ecosystem gap with Nvidia, regardless of how competitive its raw hardware becomes.
- China’s domestic GPU makers (Moore Threads, Biren, MetaX) will keep narrowing the performance gap on paper, but will remain boxed out of the most advanced manufacturing nodes as long as current export controls hold.
- Qualcomm and MediaTek’s integrated GPU volumes will keep dwarfing every discrete-card maker combined, even though neither company ever shows up in “best GPU” conversations.
- Huawei’s Ascend line will keep absorbing domestic Chinese AI-chip demand that can no longer legally buy Nvidia’s top accelerators, deepening the split between a U.S.-aligned and a China-aligned AI hardware stack.
Historical context: this isn’t the first “third GPU maker” attempt
Intel’s current Arc effort is actually its second attempt at building a discrete GPU business. The earlier Larrabee project, which BGR’s retrospective describes as plagued by delays, performance shortfalls, and eventual cancellation, was Intel’s first swing at this market more than a decade before Arc shipped. That history is worth remembering whenever a new “Nvidia killer” narrative appears, because the GPU market has quietly buried several well-funded challengers before. S3 Graphics, Matrox, and 3dfx all once held meaningful discrete-GPU share before consolidating or exiting the category entirely as the software and driver demands of modern graphics outpaced what smaller players could sustain.
What buyers should actually do with this information
If you’re shopping for a graphics card today, the alternative-maker conversation is mostly academic. Intel’s Arc B580 and B570 are worth considering only if your budget caps out around $250 and you don’t need the most demanding ray-tracing performance. Everyone else on this list, from Moore Threads to Imagination Technologies, either doesn’t sell a product you can buy at a U.S. retailer or doesn’t compete in the category you’re shopping in at all. The real value in understanding this map is strategic: knowing who’s building what helps explain supply constraints, pricing pressure, and where the next wave of GPU investment is actually headed.
FAQ: GPU makers besides Nvidia and AMD
Who makes GPUs besides Nvidia and AMD?
Intel is the only company selling complete, retail discrete GPUs for PCs beyond Nvidia and AMD, with its Arc B580 and B570 cards. Beyond discrete cards, Qualcomm, Apple, MediaTek, Samsung, and Huawei design integrated GPUs for mobile devices, while Arm and Imagination Technologies license GPU architectures rather than sell finished chips. China’s Moore Threads, Biren, and MetaX build domestic GPUs and AI accelerators primarily for the Chinese market.
Does Intel actually compete with Nvidia and AMD on GPUs?
Intel’s Arc lineup holds roughly 1% to 2% of discrete GPU shipments, according to Jon Peddie Research figures reported by Club386. It competes mainly in the budget segment under $250 and has not yet released a high-end card to rival Nvidia’s or AMD’s top models.
What is Moore Threads and why is it in the news?
Moore Threads is a Chinese GPU designer founded in 2020 that released its MTT S80 consumer card in 2022. It was added to the U.S. Entity List in October 2023, restricting its access to U.S. chipmaking tools and advanced foundry capacity, which has slowed its ability to close the performance gap with Nvidia and AMD.
Is Qualcomm’s Adreno a real GPU, or just a mobile chip feature?
Adreno is a full GPU architecture that Qualcomm designs and ships inside Snapdragon mobile chips and the Snapdragon X Elite platform used in Windows laptops. It doesn’t compete with discrete gaming cards, but by unit volume it’s one of the most widely used GPU designs in the world.
Why don’t Arm and Imagination Technologies sell their own GPUs?
Both companies operate licensing businesses: they design GPU architectures (Mali and Immortalis for Arm, PowerVR for Imagination) and license them to chipmakers, who then manufacture and brand their own processors. This generates royalty revenue rather than direct hardware sales.
How much of the GPU market does Nvidia actually control?
Nvidia held roughly 90% of desktop discrete add-in-board GPU shipments in the first half of 2026, according to Jon Peddie Research data reported by Club386, after peaking near 94% in late 2025. AMD held about 8%, and Intel held roughly 1% to 2%.
Are Chinese GPU makers a real threat to Nvidia and AMD?
Not yet on a global basis. Export controls have cut companies like Moore Threads and Biren off from the most advanced manufacturing nodes at TSMC, forcing reliance on domestic foundries that trail several process generations behind. Their near-term role is supplying China’s domestic market rather than competing globally.
What does Huawei’s Ascend chip line have to do with GPUs?
Ascend is Huawei’s data-center AI accelerator line, built to compete in the same AI training and inference workloads Nvidia’s data-center GPUs dominate. It isn’t marketed as a traditional graphics card but performs a similar computational role for AI infrastructure customers inside China.
