In the span of a single week in April 2026, three of the biggest names in tech made remarkably generous moves. Microsoft gave every tertiary student in Singapore free access to Copilot. Anthropic open-sourced a suite of knowledge work plugins. NVIDIA released its entire Nemotron family of AI models for anyone to download and use commercially. Generosity at this scale is never charity. It is strategy. And if you don't understand what's being exchanged, you're not the customer. You're the product.

The three moves

Let's look at what actually happened. Microsoft announced a $5.5 billion investment in Singapore's AI and cloud infrastructure, to be deployed over five years. As part of the package, every one of the country's 200,000+ university and polytechnic students gets 12 months of free Microsoft 365 Premium with Copilot baked in. That's Word, Excel, PowerPoint, Outlook, and OneNote, all with an AI assistant built into the workflow. The plan normally costs about $29 a month. Anthropic released its knowledge work plugins as open-source on GitHub. These aren't toy extensions. They're domain-specific skill packs, slash commands, and MCP connectors designed to turn Claude from a general-purpose chatbot into a specialized colleague for product managers, engineers, marketers, and salespeople. All built for Claude Cowork and Claude Code. NVIDIA open-sourced its Nemotron family of models, from the 8-billion-parameter Nano that runs on a laptop GPU to the 120-billion-parameter Super designed for enterprise-scale agentic AI. The weights, training data, and recipes are all freely available on Hugging Face and GitHub under commercial-friendly licenses. They also launched the Nemotron Coalition, a group of AI labs collaborating on the next generation of open frontier models. Three different companies. Three different products. The same playbook.

The oldest trick in business

This pattern has a name: the razor-and-blades model. King Camp Gillette supposedly invented it (he didn't actually, but the myth persists). Sell the razor cheap, or give it away. Make your money on the blades. Printers and ink cartridges. Game consoles and game licenses. Mobile phones and data plans. The pattern shows up everywhere a company can separate an initial product from its recurring complement. The tech industry refined this into something more sophisticated. Google's version is the clearest example: give away the world's best search engine, email service, document suite, analytics platform, and maps application. Charge nothing. In return, you get the most detailed picture of human intent and behavior ever assembled, which powers an advertising business worth hundreds of billions. The principle is simple. In any system of complements, it pays to commoditize one side to drive demand for the other. Make the complement free, and your core product becomes more valuable. For Microsoft, the core product is Azure cloud infrastructure and the enterprise ecosystem. A generation of students trained on Copilot inside Microsoft 365 is a generation of future employees who will ask their companies to buy Microsoft licenses. For Anthropic, the core product is Claude subscriptions and API usage. Open-source plugins that only work inside Claude's ecosystem make Claude stickier. Every workflow built on these plugins is a workflow that's harder to move to a competitor. For NVIDIA, the core product is GPU hardware. Free models optimized for NVIDIA architectures, from RTX consumer cards to H100 data center chips, make their silicon more attractive compared to AMD or custom chips from cloud providers.

You're the distribution channel

When something is free, it's worth asking: what is the company getting in return? Sometimes it's straightforward. Microsoft gets 200,000 students habituated to their ecosystem. That's a pipeline of future enterprise customers, trained at government expense. Sometimes it's more subtle. Open-source AI models generate telemetry, bug reports, community contributions, and benchmark results. They create an ecosystem of fine-tuned variants and integrations that all point back to the original vendor's hardware or platform. And sometimes the exchange is your attention and cognitive habits. When an AI tool does your thinking for you, summarizing documents, drafting emails, generating analysis, you develop a dependency that's harder to break than any software contract. You stop building the skill because the tool does it for you. That's a form of lock-in that doesn't show up on any balance sheet. This isn't conspiracy thinking. It's basic economics. Companies don't spend billions on giveaways because they're feeling generous. They spend billions because the expected return exceeds the cost.

The Singapore experiment

The Microsoft move in Singapore deserves special attention because it's a case study in ecosystem capture at the national level. Singapore has roughly 200,000 tertiary students. Microsoft is giving all of them free access to AI-powered productivity tools for a year. In a small, highly connected country where government and industry are tightly aligned, this isn't just a student discount. It's an infrastructure play. A whole generation of Singaporean knowledge workers will enter the workforce fluent in Microsoft's AI tools. They'll expect Copilot in their workplace. Their managers will notice. IT departments will standardize. And Microsoft's enterprise sales team will have an entire country's worth of warm leads. The $5.5 billion infrastructure investment isn't separate from the student giveaway. It's the same strategy. Build the cloud capacity, train the users, capture the market. The "free" student licenses are the cheapest customer acquisition cost Microsoft will ever pay.

The cognitive independence problem

There's a deeper concern here that goes beyond vendor lock-in and market strategy. When AI tools handle your summarizing, drafting, analyzing, and decision-making, they extract something more valuable than money. They extract the effort that builds competence. A student who uses Copilot to write every essay doesn't just become dependent on Microsoft. They potentially miss the struggle that develops clear thinking. This isn't an argument against AI tools. I use them constantly, and they make me dramatically more productive. But there's a difference between using a tool to amplify your thinking and using a tool to replace it. The free model makes this worse, not better. When something costs money, you evaluate whether it's worth the expense. When it's free, there's no friction. No moment of pause. You just use it, and the dependency builds invisibly.

How to use free tools wisely

None of this means you should avoid free tools. That would be impractical and, honestly, a bit paranoid. The genuine benefits are real. Students who couldn't afford productivity software now have access to world-class tools. Developers can experiment with frontier AI models without spending thousands on API credits. These are good things. But you can use free tools without being used by them. A few principles I try to follow: Stay portable. If you build a workflow on a free tool, make sure your data can leave with you. Use standard formats. Export regularly. If the tool doesn't let you export, that tells you something about the business model. Avoid single-vendor dependency. Use Microsoft's tools, but also know how to work in Google's ecosystem, or with open alternatives. The moment you can only work inside one platform, you've lost leverage. Keep your own data. Cloud storage is convenient, but keep local copies of anything important. The terms of service on free tools change without warning, and your access can disappear overnight. Build the skill, not just the output. Use AI to draft faster, but understand what it's drafting. Use it to analyze data, but know how to read a spreadsheet yourself. The tool should make you more capable, not less. Understand the exchange. Every free tool has a business model. Figure out what it is. If you can't figure it out, assume you're the product.

The defining tradeoff

The tension between accessibility and lock-in is the defining tradeoff of this era in technology. Free tools genuinely democratize access. A student in Singapore can now use the same AI-powered productivity suite as a Fortune 500 executive. A solo developer can experiment with the same models that power billion-dollar startups. That matters. But every free tool comes with strings. Sometimes those strings are visible: data collection, ecosystem lock-in, usage terms. Sometimes they're invisible: cognitive dependency, skill atrophy, attention capture. The companies making these moves aren't evil. They're rational. They've calculated that giving away the complement is the most efficient way to sell the core product. Understanding that calculation is the first step toward using their generosity on your own terms. Free tools are great for builders. Just know what you're building with, and what's being built around you.

References