cd ..

The Browser Wars Playbook Is Running Again — This Time With LLMs

|10 min read

It's 1995. A 24-year-old named Marc Andreessen has just taken Netscape public. The stock doubled on day one. Everyone's losing their minds over this thing called "the web browser." A piece of software that lets you see the internet.

Thirty years later, swap "browser" with "large language model" and the script reads almost the same.

The parallels between the 90s browser wars and today's LLM race aren't cute analogies. They're the same structural forces playing out again. If you understand the first one, you can see the second one more clearly.

Let's break it down.

The Pattern: A New Layer Becomes The Gateway

Every tech war starts the same way. A new technology creates a gateway — a chokepoint between people and something they want.

In the 90s, the gateway was the browser. The internet existed before Netscape, but it was ugly, technical, and inaccessible. Netscape Navigator turned it into something your mom could use. Suddenly, the browser wasn't just software. It was the door to the internet. Whoever controlled the door controlled what people saw, which search engine they used, which bookmarks came pre-installed, which companies got traffic.

In 2026, the gateway is the LLM. Information exists everywhere — documents, databases, APIs, the entire web. But an LLM is what makes it usable. You talk to it in plain language. It talks back. It writes your code, drafts your emails, explains your medical reports. The LLM is the new door.

The underlying force: when a new layer becomes the default way people interact with something massive, controlling that layer becomes existentially important.

That's why Microsoft went to war over browsers. That's why OpenAI, Google, Anthropic, and Meta are going to war over LLMs.

Act 1: The Scrappy Pioneer

Netscape Navigator launched in 1994. Within months, it owned around 80% of the browser market. Andreessen was on the cover of magazines. The product was genuinely good — fast, visual, and it just worked. Netscape was the company that made the internet real for most people.

ChatGPT launched in November 2022. Within two months, it had 100 million users — the fastest-growing consumer app in history. OpenAI was the company that made AI real for most people. Sam Altman became the face of the movement, just like Andreessen had been.

In both cases, the pioneer didn't invent the underlying technology. Browsers existed before Netscape (Mosaic, Lynx). LLMs existed before ChatGPT (GPT-3, BERT, academic research going back years). But the pioneer was the one who packaged it in a way that normal humans could use and love.

The lesson: the company that makes a complex technology accessible to the masses gets the first-mover advantage — even if they didn't invent it.

Act 2: The Sleeping Giant Wakes Up

Microsoft initially thought the future was interactive cable TV, not web browsers. By the time Bill Gates recognized the threat, Netscape already owned the market. Gates held a now-famous press conference on December 7, 1995 — the anniversary of Pearl Harbor, which the press had a field day with — declaring Microsoft's internet strategy. The message was clear: we're coming.

Google had been working on AI for years (it literally invented the Transformer architecture in 2017), but it was caught flat-footed when ChatGPT went viral. Internally, it was a "code red." Sundar Pichai scrambled. Google rushed out Bard, which fumbled its public demo by getting a factual answer wrong. It was embarrassing. But then Google regrouped, rebranded to Gemini, and started leveraging what it does best: scale, infrastructure, and distribution.

Microsoft came late to browsers but had Windows on 95%+ of desktops — the ultimate distribution channel. Google came late to the LLM consumer race but has Search on billions of devices, Gmail, Android, Chrome, and YouTube. Both sleeping giants had the same weapon: distribution that the pioneer couldn't match.

This keeps repeating: in a platform war, the company with the strongest existing distribution channel has a structural advantage — even if their product is initially worse.

Act 3: The Distribution War

This is where the browser wars got ugly, and it's where the LLM wars are heading.

Microsoft didn't beat Netscape by building a better browser (at least not at first). IE 1.0 was mediocre. IE 2.0 was passable. What Microsoft did was:

  • Bundle IE with Windows for free. If your product comes pre-installed on every computer sold, you don't need a marketing budget.
  • Strike deals with OEMs (Dell, Compaq, IBM) to pre-install IE and make it hard to ship Netscape.
  • Pay AOL — then the biggest internet gateway in the US — to use IE as their default browser.
  • Make IE free while Netscape charged businesses for licenses.

The strategy was blunt: make IE the path of least resistance. Don't give people a reason to seek out alternatives.

Now look at the LLM space in 2025-2026:

  • Google is baking Gemini into everything. Search, Gmail, Docs, Android, Chrome. You don't need to go find an AI — it's already there, in tools you already use. Gemini's monthly active users crossed 650 million partly because people stumble into it.
  • Microsoft invested $13 billion in OpenAI and embedded GPT into Bing, Office 365 (Copilot), Windows, and GitHub. If you use Microsoft at work (and most of the world does), you're already using OpenAI's models whether you realize it or not.
  • Apple rolled Apple Intelligence into every iPhone and Mac, partnering with multiple LLM providers to make AI a system-level feature.

The LLM isn't a destination you navigate to. It's being embedded into the tools you already live inside. Just like IE wasn't a download — it was already on your desktop.

The pattern holds: in a war over a gateway technology, distribution eventually beats product quality. Making the product "just there" matters more than making it the best.

Act 4: The Feature War (And Why Users Suffer)

During the browser wars, Netscape and Microsoft shipped new versions at a breakneck pace. Every few months, a new release. Features were piled on faster than they could be stabilized. Netscape invented JavaScript. IE introduced CSS support. Netscape added the <blink> tag. IE countered with <marquee>. Each browser had proprietary extensions that only worked in their browser.

The result? Web developers had to build their sites twice — once for Netscape, once for IE. You'd see banners on websites: "Best viewed in Netscape Navigator" or "Optimized for Internet Explorer." Standards? Nobody cared. Shipping mattered more than interoperability.

The LLM space in 2025-2026 looks eerily similar:

  • Three major frontier models dropped within 12 days of each other in late 2025 (GPT-5.1, Gemini 3 Pro, Claude Opus 4.5).
  • Every lab is racing on "reasoning depth" — OpenAI calls it adaptive reasoning, Google calls it Deep Think, Anthropic calls it extended thinking. Same concept, different branding.
  • Context windows keep getting bigger. Function calling formats differ between providers. Tool use protocols vary. If you build on one model's API, switching is painful.
  • Pricing is being slashed aggressively — Anthropic cut Opus pricing by roughly 60%, Google is subsidizing Gemini to grab market share.

And just like the 90s, users and developers are caught in the middle. Your app works great on Claude? Cool, it might behave differently on GPT. Your prompts are tuned for Gemini? They'll need reworking for Claude. We don't have banners that say "Best used with ChatGPT," but we might as well.

Same story, different decade: when companies compete on speed of iteration rather than standards, fragmentation is inevitable. Users pay the cost.

Act 5: The Antitrust Question

The browser wars didn't end with a better product winning. They ended with a lawsuit.

In 1998, the US Department of Justice and 20 state attorneys general sued Microsoft for abusing its monopoly. The core allegation: Microsoft illegally tied IE to Windows and used its market dominance to crush Netscape. A Microsoft executive was quoted as saying the goal was to "cut off Netscape's air supply." The judge found Microsoft guilty of monopolization. Microsoft was nearly broken up into two companies.

The LLM wars haven't hit this stage yet, but the ingredients are all there:

  • Google controls Search (90%+ market share) and is embedding Gemini directly into it. If that isn't a bundling strategy, what is?
  • Microsoft's $13B investment in OpenAI, combined with the Office/Windows integration, creates a tight coupling between the dominant productivity suite and a specific AI provider.
  • The FTC and EU regulators are already scrutinizing these deals. The cloud partnerships (Amazon-Anthropic, Google-Anthropic, Microsoft-OpenAI) look a lot like the OEM deals of the 90s.

We're maybe five years away from the LLM antitrust equivalent. But the structural parallels are already visible.

And here's the inevitable next act: when a gateway technology gets bundled with an existing monopoly, regulators eventually show up. The question is when, not if.

What The Browser Wars Tell Us About How This Ends

Here's the thing nobody talks about: the browser wars didn't produce a lasting winner.

Microsoft won by killing Netscape. IE hit 96% market share by 2002. And then... Microsoft got complacent. IE barely updated between 2001 and 2006. It became slow, insecure, and full of pop-ups. The IE brand became a joke.

Into that vacuum walked Firefox (born from Netscape's open-sourced code — ironic, right?) and then Chrome. By 2012, Chrome was the dominant browser. Today, it owns roughly 65% of the market and nobody's close to catching it.

The company that won the browser war lost the browser market a decade later. The real winner was the open-source project that rose from the ashes of the loser.

What does this mean for LLMs?

  1. Whoever "wins" the current LLM war probably won't stay on top. Market dominance leads to complacency. It happened with IE, it'll happen with LLMs.

  2. Open-source will matter more than anyone expects. Just as Firefox and eventually Chrome emerged from outside the original duopoly, open models (Llama, DeepSeek, Qwen, Mistral) are the dark horse. DeepSeek already showed you can get near-frontier performance at a fraction of the cost. Chinese open-weight models are being downloaded millions of times.

  3. The real winner will be the one that becomes infrastructure, not a product. Chrome didn't win because it was the "best" browser (debatable). It won because it became the platform — extensions, developer tools, Chromebook, everything built on top. The LLM that wins long-term will be the one that becomes the invisible platform underneath everything, not the one with the best chatbot.

  4. Standards will eventually emerge. The fragmentation of the 90s web eventually gave way to W3C standards, proper CSS, and a shared HTML specification. The LLM space will go through the same growing pains. Prompt formats, tool calling protocols, and evaluation benchmarks will standardize. It'll just take time.

The Meta-Lesson

If there's one takeaway from all of this, it's this:

Technology wars follow gravity, not genius.

The browser wars weren't decided by who had the best engineers or the cleverest features. They were decided by distribution, bundling, pricing to zero, and regulatory intervention. The LLM wars will be decided by the same forces.

The best model doesn't win. The most available model wins — until it gets complacent, and then something leaner and hungrier replaces it.

Netscape → IE → Firefox → Chrome.

ChatGPT → ??? → ??? → ???

The cycle is already in motion. The only question is where in the loop we are.

If you've been through enough tech cycles, the patterns become hard to unsee. The names change. The logos change. The press conferences get fancier. But the underlying forces — distribution advantages, bundling strategies, feature fragmentation, open-source disruption, regulatory backlash — they just keep rhyming.