The latest move in the hardware world: Intel’s newest laptop silicon is no longer a slide at a trade show. Systems built around the Core Ultra X9, the flagship chip in Intel’s Panther Lake generation, started reaching buyers this month, with fresh product listings and updated spec sheets dated as recently as September 14-15, 2026. The timing puts Intel squarely in the middle of a three-way fight for the AI PC market, pitted against AMD’s Ryzen AI 400 series and Qualcomm’s Snapdragon X2, just as Mercury Research clocked ARM-based laptops at a record 15.3% of the client CPU market in the second quarter of 2026.
None of the three chipmakers is standing still. Intel needs Panther Lake to stop the bleeding after Mercury Research put its share of x86 client CPUs at 69.7% in Q2 2026, a multi-decade low. AMD is pushing Ryzen AI 400 laptops with a 60 TOPS neural processor as its answer. Qualcomm, meanwhile, is leaning on its Hexagon NPU’s 80 TOPS rating to argue that Windows on ARM has finally closed the software gap. The numbers below come from Intel’s own product listings, AMD and Qualcomm press materials, and market-share data multiple outlets attribute to Mercury Research.
What Shipped This Week: Core Ultra X9 Hits Retail
Panther Lake, Intel’s brand name for the Core Ultra Series 3 family, was announced at CES in January 2026, with broad retail availability following by late January. What changed in mid-September is that Intel’s own product page for the Core Ultra X9 388H carried an update timestamped September 14, 2026, and OEM listings began showing up with that exact silicon inside. One example circulating in German retail channels this month is a limited-run 2-in-1 convertible built around the Core Ultra X9 378H, paired with 32GB of soldered LPDDR5X memory, a 1TB PCIe 4.0 SSD, and Intel’s integrated Arc B390 graphics, priced at €2,499 with deliveries beginning in the first week of September.
That single laptop is not the story by itself. It matters because it shows Panther Lake’s top-tier X9 silicon moving from a spec sheet into shipping hardware nine months after its CES debut, a gap that is fairly typical for Intel’s mobile launches but one competitors are trying to exploit while it lasts.
Inside the Core Ultra X9: Cores, Clocks, and NPU TOPS
The Core Ultra X9 388H is Intel’s current halo mobile part. It packs 16 cores and 20 threads in a hybrid layout, boosts to 5.1GHz, and carries 18MB of Intel Smart Cache. The built-in Arc B390 graphics unit handles the GPU workload, while a dedicated NPU rated at 50 TOPS of INT8 performance runs the AI-specific tasks Windows increasingly expects a modern laptop to handle locally. Intel’s marketing also cites a combined CPU-plus-GPU-plus-NPU platform figure as high as 180 TOPS on the top SKUs, though that number blends three separate processors rather than describing the NPU alone.
Lower down the X9 and X7 stack, TOPS ratings hold remarkably steady. Every H-series Panther Lake part Intel has listed, from the 388H down to the Core Ultra 5 325, carries an NPU rated between 47 and 50 TOPS. That consistency is deliberate. Intel needed every current-generation Panther Lake chip to clear a specific bar, one Microsoft set for the entire Windows ecosystem.
Why 40 TOPS Became the Line in the Sand
Microsoft’s Copilot+ PC program is the reason NPU TOPS ratings suddenly matter to ordinary laptop buyers. Microsoft’s own Windows 11 specifications page sets the bar plainly: a Copilot+ PC needs a processor with an NPU capable of more than 40 TOPS, alongside baseline RAM and storage requirements. Microsoft’s developer documentation repeats the same 40 TOPS floor and ties several on-device Windows AI features directly to it, including local Recall indexing, offline Copilot voice features, and live captioning.
That 40 TOPS number is why Intel, AMD, and Qualcomm all cleared it with room to spare in their current generations, and why laptops that fall short of it draw unflattering headlines. Acer’s Swift Air 16, for instance, shipped with an NPU rated around 17 TOPS against the 40 TOPS bar, and it was marketed without the Copilot+ label as a result. Every chip discussed in this piece clears the line by a wide margin, which is exactly why the real competition has moved to how far past 40 TOPS each vendor can push.
AMD’s Counter-Punch: Ryzen AI 400 and XDNA 2
AMD’s answer to Panther Lake is the Ryzen AI 400 series, codenamed Gorgon Point, built on Zen 5 CPU cores paired with an XDNA 2 neural processor. The flagship mobile part, the Ryzen AI 9 HX 475, runs 12 cores and 24 threads, boosts to 5.2GHz, and carries 36MB of combined cache. Its NPU is rated at 60 TOPS, the highest standalone figure of any x86 chip currently shipping, and its integrated Radeon 890M graphics unit adds 16 RDNA 3.5 compute units clocked up to 3.1GHz.
AMD’s stack scales down from there. The Ryzen AI 9 HX 470 drops to 55 TOPS, and every SKU down to the Ryzen AI 5 430 still carries a 50 TOPS NPU, matching Intel’s floor almost exactly. AMD, whose latest press materials are collected on its newsroom page, has also pushed the same XDNA 2 silicon into desktop APUs like the Ryzen AI 7 450G, arguing that on-device AI inference should not be a laptop-only feature. OEM partners including Acer, Asus, Dell, HP, Lenovo, and MSI began shipping Ryzen AI 400 laptops in the first quarter of 2026, giving AMD roughly the same head start as Intel’s Panther Lake line.
Qualcomm’s Play: An 80 TOPS NPU and ARM’s Slow Climb
Qualcomm is not competing on core count parity with x86. Its argument is raw NPU throughput. The Snapdragon X2 Elite Extreme runs 18 Oryon cores (12 Prime plus 6 Performance), boosts to 5.0GHz, and pairs that with a Hexagon NPU Qualcomm, whose latest announcements run through its newsroom, rates at 80 TOPS of INT8 performance, a full 20 TOPS ahead of AMD’s fastest mobile chip and 30 TOPS ahead of Intel’s. Every tier in the X2 lineup, down to the entry X2E-80-100, shares that same 80 TOPS NPU rating, which Qualcomm has made a centerpiece of its marketing since the line launched in September 2025 with hardware following through the first half of 2026.
Qualcomm’s software compatibility layer for x86 apps has closed much of the gap that held Windows-on-ARM back for years, and Mercury Research’s Q2 2026 numbers back that up: ARM-based PC clients, a category that includes both Qualcomm Snapdragon laptops and Apple’s M-series Macs, hit 15.3% of the total client CPU market, up from 14.4% in Q1 and 13.9% in the fourth quarter of 2025. That is the highest share Mercury Research has recorded for ARM in the PC client market, though it still leaves x86 controlling roughly six out of every seven laptops and desktops sold.
Three-Way Spec Comparison
Laid side by side, the flagship mobile parts from each vendor show three different bets on where AI PC performance should come from. Intel splits the load across CPU, GPU, and NPU. AMD chases the highest x86 NPU figure available. Qualcomm leans on core count and NPU throughput to justify the ARM transition.
| Chip (flagship mobile SKU) | Cores / Threads | Max Boost | NPU TOPS | Integrated GPU | Launch Window |
|---|---|---|---|---|---|
| Intel Core Ultra X9 388H | 16C / 20T | 5.1GHz | 50 TOPS | Arc B390 | Q1 2026 (retail Sept 2026) |
| AMD Ryzen AI 9 HX 475 | 12C / 24T | 5.2GHz | 60 TOPS | Radeon 890M | Q1 2026 |
| Qualcomm Snapdragon X2 Elite Extreme | 18C / 18T | 5.0GHz | 80 TOPS | Adreno X2-90 | H1 2026 |
Every chip in that table clears Microsoft’s 40 TOPS Copilot+ threshold by at least 10 TOPS, so the practical difference for most buyers comes down to app compatibility, battery life, and price rather than whether Recall or Copilot Voice will run at all.
Market Share: Mercury Research’s Q2 2026 Numbers
Sales figures tell a different story than spec sheets. Mercury Research data cited by Tom’s Hardware, PCWorld, and The Register puts Intel’s share of the x86-only client CPU market at 69.7% in Q2 2026, down from prior quarters and described by multiple outlets as among the lowest levels in the company’s history. AMD’s x86 client share sits at 30.3% for the same quarter, with an even stronger 28.9% to 71.1% split in AMD’s favor specifically within notebooks compared to the desktop segment.
| Period | ARM Share of PC Client CPUs | Intel Share of x86 Client | AMD Share of x86 Client |
|---|---|---|---|
| Q4 2025 | 13.9% | Not separately reported | Not separately reported |
| Q1 2026 | 14.4% | Not separately reported | Not separately reported |
| Q2 2026 | 15.3% | 69.7% | 30.3% |
Source: Mercury Research data as reported by Tom’s Hardware, PCWorld, and The Register. Figures reflect PC client CPUs, desktop plus notebook combined. ARM share includes Qualcomm Snapdragon and Apple M-series silicon.
Vendor-level shipment data from Counterpoint Research adds context on the OEM side. Lenovo shipped 16.5 million PCs in Q1 2026, up 9% year over year, for a 26% share Counterpoint described as a first-quarter record for the company. Counterpoint also noted the top five PC vendors combined for close to 78% of global shipments in the same period, a concentration that matters because it means Intel, AMD, and Qualcomm are really fighting for design wins at a handful of companies rather than a fragmented field of buyers.
What Buyers Actually Pay
HP’s OmniBook Ultra 14 (G2) is one of the few laptops sold in nearly identical chassis with both Intel and Qualcomm silicon inside, which makes it a useful price comparison. Windows Central’s coverage of the line put the Intel Core Ultra 7 365H configuration at $1,699.99, with a higher-end Core Ultra 9 386H version around $2,339.99. The Snapdragon X2 Elite version of the same chassis lists at $2,049.99 through Best Buy, landing between the two Intel configurations.
Microsoft’s own Surface line prices Snapdragon X2 hardware lower: the Surface Pro starts at $1,499 and the Surface Laptop with Snapdragon X2 Plus starts at $1,599, according to reporting on Microsoft’s June 2026 refresh. Qualcomm, which has also been raising prices on its own chip lineup elsewhere in its AI PC push, has described its target price bands as roughly $1,000-plus for Elite-tier laptops, around $800 for Plus-tier systems, and $600 and up for entry X-tier machines, positioning the platform to undercut premium Intel and AMD configurations at the low end while still competing at the top.
Historical Context: From an x86 Duopoly to a Three-Horse Race
For roughly two decades, the laptop CPU story was Intel versus AMD, full stop. Apple’s 2020 shift to its own M-series silicon was the first real crack in that arrangement, but Apple never sold chips to other PC makers, so the rest of the Windows market stayed x86-only. Qualcomm’s original Snapdragon X Elite, launched in 2024, was the first serious attempt to bring ARM to mainstream Windows laptops beyond Apple’s walled garden, and it shipped with a 45 TOPS NPU that looked impressive at the time.
Snapdragon X2’s 80 TOPS rating, announced in September 2025, nearly doubled that figure in a single generation, which is part of why Windows Central and other outlets framed it as a bigger leap than the jump from Snapdragon X Elite to X Plus had been. Intel and AMD’s responses, Panther Lake and Ryzen AI 400 respectively, both launched within weeks of each other at CES 2026, effectively turning what used to be a slow-moving two-company market into an annual three-way product cycle.
Why Local LLM Inference Is Driving the NPU Arms Race
The TOPS numbers climbing so fast are not just marketing theater. Windows features like on-device Copilot processing, Recall’s local semantic indexing, and live translation all run inference on the NPU specifically so they can work offline and without sending data to the cloud. AMD has leaned hardest into this argument, positioning its Ryzen AI Max PRO 400 variant, a separate high-memory line from the mainstream Ryzen AI 400 covered here, as capable of running language models with roughly 300 billion parameters locally when paired with up to 192GB of unified memory.
That framing matters for how buyers should read TOPS figures. A 50, 60, or 80 TOPS NPU rating measures peak throughput for AI-specific instructions, not general CPU or GPU performance, and it says nothing on its own about how much memory a system has available to actually hold a large model. Intel, AMD, and Qualcomm all clear Microsoft’s Copilot+ floor comfortably, so for most buyers the memory configuration and storage speed of a specific laptop will matter more day-to-day than which vendor’s NPU is rated a few TOPS higher.
Competitive Comparison: Who Wins Which Buyer
Intel’s Core Ultra X9 still holds the strongest software compatibility story, since every Windows application runs natively without an emulation layer, and Intel’s Arc B390 graphics give it a real edge over Qualcomm’s Adreno GPU for creative and light gaming workloads. AMD’s Ryzen AI 9 HX 475 currently posts the best combination of raw CPU multi-thread performance and NPU throughput within the x86 world, making it the pick for buyers who want to stay on x86 but still want AMD’s 60 TOPS ceiling. Qualcomm’s Snapdragon X2 wins on battery life and NPU throughput outright, and its case has gotten measurably stronger now that app compatibility has improved since the original Snapdragon X Elite launch, but buyers running specialized x86-only software still have reason for caution.
Market Impact: OEMs, Investors, and the Windows Ecosystem
For OEMs, having three viable silicon vendors clearing the Copilot+ bar means product lineups can differentiate on chip choice the way they used to differentiate mainly on screen quality or chassis material. HP’s decision to sell nearly identical OmniBook Ultra 14 chassis in both Intel and Qualcomm configurations is a preview of what more vendors are likely to do through the rest of 2026, letting buyers effectively vote on architecture with their wallet inside the same physical design.
For Intel specifically, which is also navigating a memory supply crunch CEO Lip-Bu Tan has flagged as a multi-year problem, the stakes are higher than a single product cycle. A 69.7% share of x86 client CPUs, against AMD’s steadily climbing 30.3%, still leaves Intel dominant, but the trend line across the past several quarters has consistently moved in AMD’s favor. Panther Lake’s job is to slow or reverse that slide while simultaneously fending off Qualcomm’s ARM push, a two-front fight Intel has not had to wage at this intensity before.
Predictions: Where the AI PC Chip Race Goes Next
- ARM’s PC client share likely keeps climbing gradually rather than spiking, following the 13.9% to 14.4% to 15.3% pace seen across the last three reported quarters, putting it in range of 17-18% by mid-2027 if the trend holds.
- Expect Intel to accelerate its next Panther Lake refresh or a successor architecture reveal in early 2027 rather than waiting for a full CES cycle, given the pressure from both AMD and Qualcomm simultaneously.
- NPU TOPS figures across all three vendors will likely plateau in the 50-90 TOPS range for mainstream chips through 2027, with vendors instead competing more on memory bandwidth and on-device model support than on TOPS alone.
- More OEMs will follow HP’s lead in offering the same laptop chassis across Intel, AMD, and Qualcomm configurations, making chip choice a checkbox rather than a reason to pick an entirely different model.
- Backed by a recently expanded total addressable market forecast from AMD’s own leadership, its client x86 share is likely to keep narrowing the gap with Intel gradually, though a full crossover to AMD majority share in notebooks specifically remains unlikely before 2028 given Intel’s OEM relationships and manufacturing scale.
What This Means for Laptop Buyers Right Now
Anyone shopping for a new laptop in the back half of 2026 can treat the Copilot+ 40 TOPS floor as table stakes rather than a differentiator, since Intel, AMD, and Qualcomm all clear it comfortably in their current lineups. The more useful questions are the ones TOPS ratings do not answer: whether the software you run daily has any compatibility issues on Windows on ARM, how much RAM a specific configuration ships with, and how a laptop’s real battery life holds up outside of vendor-supplied benchmarks. For buyers who need guaranteed native compatibility with older Windows software, Intel or AMD configurations remain the safer default. For buyers prioritizing battery life and who mostly run modern, actively maintained applications, Qualcomm’s Snapdragon X2 has closed enough of the compatibility gap to be a realistic option for the first time.
Frequently Asked Questions
What is the Intel Core Ultra X9, and is it the same chip as Panther Lake?
Core Ultra X9 is the product name for Intel’s flagship mobile chip within the Core Ultra Series 3 family, which Intel branded internally as Panther Lake. The X9 388H is the top-end SKU, with 16 cores, 20 threads, and a 50 TOPS NPU.
How many TOPS does the Intel Core Ultra X9 NPU deliver?
Intel rates the Core Ultra X9 388H’s NPU at 50 TOPS of INT8 performance, per Intel’s own product listing. Intel also cites a combined CPU-plus-GPU-plus-NPU platform figure up to 180 TOPS on the highest SKUs, though that number covers all three processing units together, not the NPU alone.
What is the minimum NPU requirement for a Microsoft Copilot+ PC?
Microsoft requires an NPU capable of more than 40 TOPS for Copilot+ PC certification, according to Microsoft’s own Windows 11 specifications page. All current Intel Core Ultra, AMD Ryzen AI 400, and Qualcomm Snapdragon X2 chips exceed that floor.
How does AMD’s Ryzen AI 400 compare to Intel’s Core Ultra X9 on paper?
AMD’s flagship Ryzen AI 9 HX 475 carries a 60 TOPS NPU, ahead of Intel’s 50 TOPS, but runs fewer CPU cores (12 versus 16). Both chips clear Microsoft’s Copilot+ threshold with room to spare.
Why does Qualcomm’s Snapdragon X2 have a higher TOPS rating than Intel and AMD?
Qualcomm designed its Hexagon NPU specifically to lead on AI throughput as its main selling point against x86 rivals, since it cannot compete on native x86 software compatibility. The Snapdragon X2 Elite Extreme’s 80 TOPS rating is roughly 60% higher than Intel’s current flagship and about a third higher than AMD’s.
Is ARM actually catching up to x86 in the laptop market?
Slowly. Mercury Research data shows ARM-based PC clients, including Qualcomm and Apple silicon, rising from 13.9% in Q4 2025 to 15.3% in Q2 2026. That is a real, sustained increase, but x86 still accounts for roughly 85% of the PC client market.
Which chip should I buy for an AI PC in late 2026?
All three current-generation chips clear the Copilot+ 40 TOPS bar comfortably, so the choice comes down to software compatibility and price rather than raw NPU specs. Intel and AMD remain the safer picks for guaranteed native Windows software compatibility. Qualcomm’s Snapdragon X2 is a realistic option for buyers prioritizing battery life who mostly run modern, actively updated apps.
When did Intel’s Core Ultra Series 3 laptops start reaching real buyers?
Intel announced Core Ultra Series 3 (Panther Lake) at CES in January 2026, with broad retail availability following by late January. Updated product listings and new OEM laptops carrying the Core Ultra X9 continued appearing as recently as mid-September 2026.
