Linux Gains Ground: Kernel 7.2, Market Share, and

A lot tends to happen in the Linux world in any given week, but every so often the news doesn't just accumulate — it compounds. This is one of those weeks.

Linux 7.2 landed. Desktop market share crossed 10% in North America by at least one major measurement. The Arch User Repository got a meaningful security overhaul after a rough summer of malware incidents. KDE revived its long-term support program with actual money behind it. Framework started shipping a laptop with Fedora pre-loaded. And three major gaming companies — Nvidia, GOG, and Epic — are each making moves toward Linux, at varying degrees of seriousness. Oh, and FFmpeg 9.0 shipped, quietly powering roughly everything as usual.

None of these stories is unconnected. Together they sketch something that's been building for years and is now difficult to ignore.

What 7.2 Actually Does

Linux 7.2 is a technically dense release, but two changes deserve particular attention because they reflect how the kernel is adapting to modern hardware realities.

The first is cache-aware scheduling. On processors with multiple core clusters sharing different last-level caches — increasingly the norm — moving tasks between those clusters means dragging their data around too. The 7.2 scheduler tries to keep related tasks within the same cache domain, reducing the cache misses and cache thrashing that quietly degrade performance. The gains are workload-dependent, but the direction is right.

The second is a fairer GPU scheduler. The existing GPU job scheduler operates on a first-in, first-out basis, which sounds reasonable until you have an interactive application trying to share the GPU with a heavy background workload. Linux 7.2 replaces this with an approach modeled on the CFS task scheduler — the same basic fairness philosophy that governs CPU scheduling — which should meaningfully improve responsiveness for mixed GPU workloads.

There's also a cleanup of the MGLRU memory reclaim loop that reportedly yields up to 30% improvement in certain workloads, and USB4 streaming support that lets applications push raw data over Thunderbolt without needing socket abstractions. Btrfs improvements round things out.

When this reaches you depends entirely on your distribution's release cadence. Rolling release users will see it soon. Everyone else: check your distro's changelog schedule.

The 10% Number, and Why It's Complicated

StatCounter's July 2026 data puts Linux at 10.65% of desktop web traffic in North America and over 7% globally. Those numbers have spread through the Linux community with predictable enthusiasm. They deserve a more careful reading.

StatCounter measures browser-reported OS across more than a million websites and billions of page views. It is not an install census. And the jump itself is suspicious: in June, Linux sat at 5.5% in North America while 9.2% of traffic was classified as "unknown." In July, that unknown category essentially evaporated while Linux jumped to 10.65%. As Michael Tunnell notes in This Week in Linux, "Linux probably didn't suddenly double its user base in a month. Some of this increase is probably just StatCounter getting better at identifying traffic it previously couldn't classify."

That's a fair read. But the corroborating signals matter. Cloudflare Radar — drawing from its own HTTP request dataset, entirely separate methodology — put Linux at 9.7% of North American desktop traffic in July. Valve's optional Steam hardware survey, measuring a completely different population (gamers who opt in), hit 4.01% for Linux in the same month, with macOS at 2.3%.

Three independent measurement systems, all pointing in the same direction at the same time. That's not a statistical artifact. It's a trend.

What's driving it is a familiar list: Proton and SteamOS improving gaming, steadily improving hardware support, a desktop experience that genuinely keeps getting better, and — let's be honest — growing frustration with the direction of Windows. The YouTube Linux content wave of the past year has probably moved some people too.

The measurement problem itself is worth sitting with. The Linux community's allergy to telemetry — earned through years of watching big companies abuse user data — means the ecosystem can't cleanly answer "how many users do you have?" Tunnell floats a proposal for a voluntary, anonymized machine-ID ping system that would just count active installs without attaching any identifying information, and he's not wrong that the absence of reliable numbers costs Linux credibility when negotiating with software vendors: "That 'Linux is not big enough' nonsense would vanish." Whether the community would accept even the most privacy-respecting version of that system is a genuinely open question. The distrust isn't irrational.

The AUR's Summer of Malware

The Arch User Repository had a rough few months. Multiple waves of malicious package adoptions — attackers taking over orphaned packages and injecting malware — forced the Arch team into a reactive posture that culminated in temporarily disabling package pushes entirely in early August.

The core problem was structural: previously, if a package was orphaned, any user could adopt it and immediately become its maintainer. No friction, no review. That's a feature in a trust-based community and an attack vector in one experiencing coordinated malicious activity.

The fix, deployed with AUR web 6.5 on August 11th, adds a human approval step: adoption now creates a request rather than an immediate transfer of control, and an existing maintainer must approve before ownership changes hands. Earlier safeguards — disposable email rejection, mandatory email verification, 24-hour token expiry — addressed account creation abuse but left the adoption pathway open.

Tunnell is appropriately measured about what this achieves: "This doesn't make the AUR bulletproof to attackers because, you know, losers are losers and can still cause a mess in other ways, but they won't be able to do it in mass like before." That's the right framing. Adding friction to mass attacks is meaningful even when determined individual attacks remain possible. The AUR's community-maintained model is both its strength and its vulnerability — and the Arch team is threading that needle carefully.

KDE Puts Money on the Table

Here's the story I find most interesting from a sustainability standpoint.

KDE killed its old Plasma LTS program a few years ago, and with good reason: the previous LTS only covered Plasma itself, leaving KDE Frameworks and KDE Gear applications on a different support cycle. You could end up with an "LTS" desktop depending on unsupported components underneath it. Kubuntu was essentially the only distribution leaning on it. The program was LTS in name more than practice.

The new "Bullet-Proof KDE" initiative is different in kind, not just degree. Kubuntu Focus is committing over $100,000 across three years to fund development and CI infrastructure for Plasma 6.6 LTS — and crucially, this LTS covers the entire KDE software stack: Plasma 6.6, KDE Frameworks 6.24, and KDE Gear 25.12. Bug fixes and security patches get backported across all three layers, tested against Kubuntu 26.04 LTS specifically. Work reportedly began in May, and Kubuntu 26.04 users have already received several fixes — desktop theme rendering, media widget scaling, remote desktop reliability.

The governance angle matters as much as the technical one. This is a collaboration between Kubuntu Focus, Tech Paladin Software, KDev, and the Kubuntu team — not a single vendor making promises. And because it's open source, any distribution can build on the work. Tunnell puts it simply: this demonstrates "that the phrase 'put your money where your mouth is' can be used for so much good in the open-source world."

It can. But it also raises the question every OSS sustainability story raises: what happens when the funding runs out?

Framework Ships Linux as a Default

The Framework Laptop 12's refresh — Intel Core Series 3, Thunderbolt 4, Wi-Fi 7, fingerprint reader, backlit keyboard with open-source ZMK firmware — would be noteworthy hardware news on its own. The lead story is that Framework is now offering Fedora 44 with KDE Plasma as a pre-installed option, shipping in October.

Tunnell discloses he works for Framework's marketing team, so take his enthusiasm with that context. The underlying data point he shares is striking regardless: nearly 80% of Framework Laptop 12 owners were already running some Linux distribution, with Fedora leading in user surveys. This isn't a company gambling on Linux adoption — it's a company responding to what its customers were already doing. Framework has also been seeding hardware to projects like Bazzite and CachyOS ahead of launch, so community support arrives when the hardware ships rather than months later.

There's a broader signal here. When a hardware company sees 80% Linux adoption among its users and responds by shipping Linux as a first-class pre-installed option — coordinating with Fedora and KDE on the out-of-box experience — that's the flywheel turning. More hardware compatibility means more users means more developer attention means better software means more users.

Gaming's Cautious Opening

The gaming landscape shift is worth tracking with calibrated expectations.

Nvidia's GeForce Now native Linux app graduated from beta to stable release, with an official Flatpak repository for clean installation and updates. That's real and usable now.

GOG's situation is less concrete: a hired specialist, active exploration of the best approach for Linux support in GOG Galaxy, and — critically — no specific plans, results, or timeline to announce. "Investing actual time and resources" is a statement of intent, not a commitment.

Epic is the wildcard. During a Discord AMA, an Epic developer answered a question about a Linux launcher with "soon," while clarifying it won't be in the next preview release because Linux support requires more than a native build. Tim Sweeney has publicly downplayed Linux multiple times over the years, which makes this shift notable even if "soon" covers a lot of territory.

What's easy to overlook is that community tools — Heroic Launcher, Lutris — already made GOG and Epic libraries practically accessible to Linux gamers. Official clients don't make those projects obsolete. What first-party support adds is QA, better account feature integration, and most importantly, a signal to game developers that Linux is a platform worth caring about. That signaling effect may matter more than the launchers themselves.

Steam's Linux ecosystem remains by far the most developed and integrated. That's unlikely to change soon. But for years the Linux gaming story was essentially just Valve's story. The fact that it no longer is — whatever form that takes — reflects how much the underlying numbers have shifted.

The open question is whether "soon" and "actively exploring" translate into shipping software within a timeframe that matters, or whether Linux gaming's new corporate attention is primarily a hedge against a platform that suddenly looks too large to ignore.

— Dev Kapoor, Open Source & Developer Communities Correspondent, Buzzrag