--- title: 'The Irony Loop and the Myth of Open AI: Embrace, Extend, and Extinguish V2' description: >- A historical look at Microsoft's transition from 'Linux is a cancer' to championing 'open weights' AI, and why EEE V2 threatens independent developer sovereignty. summary: >- Reading Microsoft's modern stance on open weights feels surreal when you remember the 2001 WhiteHouse.gov Linux migration and the 2006 Novell deal. Here is why the modern 'open weights' pivot is Embrace, Extend, and Extinguish V2. date: 2026-07-25T00:00:00.000Z heroImage: src: /_astro/hero.B3c5PS9c.webp width: 1360 height: 759 format: webp tags: - AI - OpenSource - Linux - Microsoft - TechStrategy - DigitalSovereignty categories: - Technology Bites published: true featured: true draft: false author: Jad href: >- /posts/the-irony-loop-and-the-myth-of-open-ai-embrace-extend-and-extinguish-v2-welcome-to-the-new-era/ slug: >- the-irony-loop-and-the-myth-of-open-ai-embrace-extend-and-extinguish-v2-welcome-to-the-new-era --- In the 1980s, early open-source software pioneers challenged the prevailing belief that software would advance only if companies kept tight control over their code. This movement pushed for a transparent ecosystem where developers around the world could study, modify, and improve software. Software developed by the open-source community now supports most of the internet and underlies systems used by the world’s largest technology companies, as well as the U.S. military and federal agencies conducting scientific research, cybersecurity, and other critical missions. Open source did more than lower the cost of software; it created a shared foundation of knowledge on which generations of engineers and entrepreneurs built their institutional sovereignty. — *Source: [Microsoft Corporate Responsibility, "Open Weights"](https://www.microsoft.com/en-us/corporate-responsibility/topics/open-weight/?ref=jadmadi.net)* Reading this statement on Microsoft’s website today feels entirely surreal. For those of us who grew up in the tech landscape of the early 2000s, this is a staggering rewrite of corporate history. I vividly remember two major security crises where critical infrastructure had to route around Microsoft’s own software just to stay online. The first was July 2001, when the Code Red worm’s payload was a scheduled DDoS attack on WhiteHouse.gov; the White House’s real-time defence was pulling one of its two DNS servers offline, and within days its hosting was [reported to have shifted onto a Linux platform entirely](https://abcnews.com/Technology/story?id=98394&page=1). The second was August 2003, when Microsoft’s own public website had to route its traffic through Akamai’s network to [survive the Blaster worm’s DDoS payload](https://faculty.cc.gatech.edu/~mbailey/publications/IEEE_Security_Privacy_Blaster_Final.pdf). Right around the dawn of these crises, former Microsoft CEO [Steve Ballmer](https://www.google.com/search?q=steve+ballmer&kgmid=/m/07bkv) famously declared in a public interview that Linux was *"a cancer that attaches itself in an intellectual property sense to everything it touches."* Long before that public outburst, the [leaked Halloween Documents of 1998](https://en.wikipedia.org/wiki/Halloween_documents) had already revealed the underlying truth: Microsoft [secretly recognised open-source software and Linux as existential threats](https://medium.com/techloop/the-halloween-documents-when-microsoft-was-scared-of-open-source-f5cad1309330) to its desktop dominance. When Microsoft realised it couldn’t eradicate Linux from the enterprise server room, its strategy shifted from destruction to containment. In 2006, [Steve Ballmer](https://www.google.com/search?q=steve+ballmer&kgmid=/m/07bkv) signed the infamous interoperability agreement with Novell. In my opinion, this deal was never a genuine embrace of open source. It was a cold, calculated virtualisation play—an attempt to ensure Windows remained the dominant manager of mixed data centres using the Hyper-V blueprint. By utilising cross-licensing clauses, Microsoft weaponised its intellectual property portfolio. Corporate manoeuvres like the Novell deal deeply fractured the community. Developers and the Free Software Foundation felt Novell had sold out. Novell became a corporate Trojan horse used to threaten its own customers and rival distributions with the spectre of patent litigation. Fast forward to today. The technical battleground is no longer the Linux kernel or device drivers; it is artificial intelligence. Seeing Microsoft position itself as a "leader" by hosting an open letter signed by a coalition of 35 companies and institutions—including Meta, OpenAI, NVIDIA, Andreessen Horowitz, IBM, Hugging Face, Mistral, Palantir, Cohere, and the Linux Foundation—is a sounding alarm. The irony of OpenAI signing on to champion "openness" alongside Microsoft and Meta is inescapable; it underscores how the definition of open source has been reframed to serve corporate convenience. While slick visibility and marketing campaigns can pass unnoticed by a casual audience, seasoned computing veterans recognise these structural shifts for what they are. We are not after corporate perfection, but clarity—and we must not play along with fake praise for corporate "openness." To understand Microsoft’s sudden pivot to "openness," we have to look past the marketing and expose the four underlying corporate playbooks driving the 2026 AI ecosystem. ## 1. The "Commoditise Your Complement" Strategy In the 2000s, IBM didn’t back Linux out of pure altruism; they spent billions supporting it to destroy Microsoft’s operating system margins. By making the operating system a free commodity, IBM forced the industry to compete on what complemented it: enterprise hardware and consulting services, where IBM ruled. Today, Microsoft and Meta are running the exact same play by backing "open weights" models like Llama. By flooding the market with highly capable, free-to-use weights, they effectively commoditise the AI model layer itself. If high-quality foundational models are free, proprietary-only rivals lose their pricing power. Meta wins by ensuring no single gatekeeper can lock them out of the AI ecosystem, and Microsoft wins because developers cannot host these massive "free" models on thin air—they must run them on expensive cloud infrastructure. ## 2. The Illusion of "Openness" Without Compute In the early days of open source, if you possessed the Linux source code, you possessed the entire means of production. You could compile it and run it on a cheap, off-the-shelf Pentium computer in your basement. The open-weights AI landscape offers no such democratisation. Owning the weights of a 400-billion-parameter model is functionally useless to an independent developer. You cannot run or fine-tune it effectively without renting massive, centralised GPU clusters—ironically hosted by hyperscalers like Microsoft Azure. The only real counterweight to this compute monopoly is the rise of highly specialised micro-LLMs. These compact, hyper-targeted models are small enough to run locally on consumer hardware and are trained to excel at specific semantic tasks or contextual synthesis. But for anything larger, the traditional open-source ethos of decentralised power has been severed; open-weights AI remains entirely tethered to centralised corporate capital. ## 3. The Data Scraping Paradox The foundational pillars of early open source were public collaboration and transparent data sharing—principles that built platforms like GitHub. Yet several of the tech giants signing this open weights letter are still defending long-running copyright litigation over how aggressively they scraped the open internet to build their proprietary empires. The hypocrisy is stark, even if it hasn’t been fully proven in court. Microsoft leveraged its $7.5 billion acquisition of GitHub to train Copilot on open-source code, and it’s still fighting [*Doe v. GitHub*](https://githubcopilotlitigation.com/), a [class action now on appeal at the Ninth Circuit](https://www.bakerlaw.com/the-copilot-litigation/), where developers allege Copilot’s suggestions strip out the attribution and licence terms — GPL, MIT, Apache — meant to protect their work. A district court already dismissed most of the original claims, and Microsoft disputes the rest, but the core question — whether "open" only ever flowed in one direction — is still alive on appeal. Meanwhile, rivals deploy strict legal frameworks, restrictive terms of service, and aggressive anti-bot technical barriers to block anyone else from scraping their web properties. As we saw in the [open-source AI rebellion](/posts/open-source-ai-rebellion), by lobbying Washington to focus purely on the freedom of deploying weights, these companies are pulling off a magic trick: they appear to champion democratisation while their own data pipelines stay locked down, and how that training data was harvested in the first place stays conveniently out of the conversation. ## 4. Embrace, Extend, and Extinguish (EEE) V2 For decades, Microsoft’s defining strategy was EEE: Embrace a public standard (like HTML or Java), Extend it with proprietary features that only function inside the Windows ecosystem, and ultimately Extinguish competitors who rely purely on the open public standard. We are witnessing the rollout of EEE V2. By hosting this open letter, Microsoft > is embracing the open-weights movement to position itself as the patron saint of AI developers. Next comes the extend phase: weaving these open models into proprietary Azure frameworks, custom developer tools, and localised Copilot extensions. Once the development pipeline becomes dependent enough on Microsoft’s infrastructure to operate efficiently, the pure, independent open-weights ecosystem risks being extinguished, leaving Azure as the permanent, inescapable landlord of the AI era. Just like the Novell deal twenty years ago, "openness" is being deployed by Microsoft not as a philosophy, but as a calculated business shield—this time, to stave off strict government regulation while cementing their cloud infrastructure dominance. The vocabulary has changed, but the playbook, in my view, remains exactly the same. Can true digital sovereignty exist when the weights are open, but the compute is locked behind corporate tollbooths? Or will the next decade belong to those who build sovereign local architectures from the edge up?