Search GitHub for Delphi and you will find a modest pile of repositories next to the millions filed under JavaScript or Python. Check Stack Overflow's annual developer survey and Object Pascal barely registers a percentage point. By the metrics most of the industry uses to gauge a language's health, Delphi looks like it is running on fumes. And yet Idera keeps shipping RAD Studio releases, TIOBE keeps finding it climbing back into its top ten, and a real, working developer population keeps turning up in places those metrics never look. The interesting question about Delphi in 2026 is not really whether it is popular. It is why the standard tools for measuring popularity are so bad at finding it.
That distinction matters for a practical reason, not just a rhetorical one. If you are the person who just inherited a Delphi codebase, or the manager deciding whether to fund keeping one running, the popularity charts will tell you almost nothing useful about how much company you have, how deep the tooling actually goes, or how likely you are to find help. The places that would actually tell you are quieter, older, and mostly not indexed by the tools built to summarize a language's health in one chart.
Two different economies of visibility
Most language-popularity metrics are, underneath the marketing, measurements of open-source visibility: public repositories, public questions, public stars, public job postings that name the language outright. That is a reasonable proxy for a language whose community habit is to build in the open. It is a much worse proxy for a language whose community habit, going back three decades, has been to build proprietary, closed-source, often regulated commercial software and never publish a line of it. Delphi did not choose obscurity. It chose a category of software to be good at, and that category rarely ships with a public GitHub link attached, a Stack Overflow tag someone remembers to use, or a conference talk anyone outside the company gets to see.
Built to be invisible from day one
This is not a side effect that showed up later, it is close to the original product design. Delphi's founding pitch, a drag-and-drop visual designer paired with a compiler producing a genuine native Windows executable, was aimed squarely at small in-house teams building serious internal software fast, not at shops planning to publish their source or write about their architecture. A two- or three-person team at a manufacturer, a regional bank, a hospital IT department, or a logistics company could ship a working inventory system, a claims processor, or a scheduling tool without hiring a dedicated ops function to manage a separate runtime, and without needing a public engineering presence at all. RAD, in the original 1990s sense of the term, was explicitly a bet on small teams and fast internal delivery, and small internal teams building software nobody outside the company will ever see is exactly the population that a popularity index built around public activity is structurally unable to count. The obscurity and the productivity come from the same design decision. That is worth sitting with for a moment, because it means the usual assumption, that a quiet language must be a declining one, does not actually hold here. Quiet was close to the point from the start.
Software you've already used without knowing it
Some of what Delphi built is more visible than its reputation suggests, if you know where to look. Total Commander, the long-running Windows file manager built by Christian Ghisler, has been developed in Delphi since it launched as Windows Commander in the 1990s; current Windows builds still use Delphi for the 32-bit version and Free Pascal's Lazarus for the 64-bit one, according to the project's own documented history. Skype's original client, before later corporate rewrites, had its interface built in Delphi as well, a detail that keeps surfacing in developer retrospectives on the app's early architecture. Neither fact is exactly a secret. Both are the kind of thing you only learn if you go looking, because neither product ever advertised its implementation language to end users, and there was never a reason to. That is the whole pattern in miniature: the software works, millions of people used it daily without a second thought, and the language underneath it stayed invisible by design, not by decline.
Where the real fleet actually lives
“Delphi and Freepascal are predominantly used to write high-quality, commercial, closed source business applications.”
Jon Lennart Aasenden, former Embarcadero developer relations lead, on his personal blog
Aasenden, who worked inside Embarcadero's developer relations organization before moving on to build his own Object Pascal tooling, has made this argument more directly than most people are willing to. His point is that public metrics undercount Delphi because its actual working population is writing software that legally and practically cannot be public. He points specifically to medical systems running hospital networks, ATM software, and point-of-sale terminals, describing the standard those systems have to meet as running "24/7 with zero downtime," the kind of infrastructure that gets audited by regulators rather than starred by strangers on the internet. None of that shows up on a trending-repositories page. All of it depends, every single day, on a compiler most industry surveys have quietly stopped asking about. It is also, not coincidentally, exactly the kind of deployment where compiling to one native executable with no separate runtime to install, patch, or certify is a genuine operational advantage rather than a curiosity, the same underlying property that gets Delphi an unflattering mention in malware research for close to the opposite reason.
A community that measures itself in forums, not stars
“We're one of the largest german-speaking Communities for Developers using Borland Delphi”
Delphi-PRAXiS's founder, introducing the forum (running since June 2002) on a web-directory listing
Delphi-PRAXiS, active since 2002, is still one of the more substantial German- and English-speaking hubs for the language, still fielding current VCL, FireMonkey, database, and Windows API questions from developers with production systems rather than hobbyists chasing a trend. Communities like it, alongside Embarcadero's own official forums and newer hubs like learndelphi.org, are where the real usage signal actually lives: not in commit graphs, but in people with running systems asking specific, current questions about them, year after year, long after a trend-chasing crowd would have moved on. It is a quieter kind of activity than a trending GitHub repository, and it never produces a headline number for a slide deck, but it is a far better indicator of a live, working population than a star count that mostly measures who is currently fashionable to attach your name to in public.
The visible edge: bridging outward, not just holding still
The part of Delphi's ecosystem that is genuinely public tells a different story from "frozen in time." FireMonkey, the cross-platform UI framework Embarcadero has shipped alongside the older, Windows-only VCL since Delphi XE2 in 2011, began life as VGScene, a vector-based rendering engine built by Eugene Kryukov's KSDev in Ulan-Ude, Russia, before Embarcadero acquired the technology and rebuilt it into FireMonkey. It now targets Windows, macOS, iOS, and Android natively, with a third-party project called FMX Linux extending it to Linux as well, and it picked up an official Python integration in 2021. That bridge to Python is not a one-off experiment either: Embarcadero maintains python4delphi, DelphiFMX4Python, and DelphiVCL4Python as open repositories on its own GitHub organization, letting Python code drive a native FireMonkey or VCL interface directly, instead of the more usual traffic pattern of Delphi developers learning Python to escape the platform. A language community that was purely nostalgic would not be building integration layers into the ecosystem currently eating everyone else's lunch. This one is, in public, on GitHub, where anyone can watch it happen.
The one public number, and what it actually proves
Put those two facts side by side and they argue against the obvious reading of either one alone. If Delphi's only public signal is a search-and-job-posting index like TIOBE, a climb back into the top ten looks almost inexplicable, a nostalgia blip nobody can quite explain. If you also know there is a 600-plus-component open library still sitting inside the current version of the IDE's own package manager, actively surfaced to developers today rather than archived, the TIOBE number looks less like a fluke and more like a lagging indicator finally catching up to infrastructure that was never actually gone. Popularity indexes measure what people are searching for and hiring for, in public, right now. They were never going to be fast at noticing a population that spends its working hours inside a hospital's IT department or a bank's back office instead of on Hacker News.
None of this is an argument that Delphi is secretly the biggest language on the planet, hiding in plain sight. It is a much narrower and more useful claim: the tools most people reach for to answer "is this language still alive" were built to measure a specific kind of activity, and Delphi's core population has spent thirty years doing a different kind of activity almost on purpose. Reading a TIOBE chart without knowing that is reading half the instrument.
Where this leaves you
- Delphi's public metrics undercount it structurally, not accidentally: its core users write closed-source, often regulated software that was never going to appear on GitHub or in a public survey.
- Genuinely famous, everyday software, Total Commander and Skype's original client among them, was quietly built in Delphi, and neither one ever advertised the fact to its users.
- The forums where real usage signal lives, Delphi-PRAXiS since 2002 and Embarcadero's own community among them, run on an older, quieter kind of activity metric than star counts, and they have kept running the whole time.
- FireMonkey's genuine cross-platform reach and Embarcadero's own open Python-bridge repositories are evidence of a community building outward, not merely maintaining what already existed.
- A TIOBE top-ten return is a real, useful signal, but it is best read as the visible tip of a population that was already there, doing regulated, unglamorous, load-bearing work, long before any index caught up to it.