ChatGPT Is Becoming the Next Yahoo

In 2000, Yahoo was worth $125 billion and synonymous with the internet itself. Sixteen years later, Verizon bought what was left for $4.48 billion. The collapse wasn't sudden. It was a slow erosion masked by huge user numbers, fueled by a strategy of "do everything for everyone" while focused competitors quietly carved out the parts that actually mattered.

That's the playbook ChatGPT is currently running. The data is starting to show it.

The Numbers Nobody Wants to Talk About

ChatGPT still posts headline metrics that look untouchable. 900 million weekly active users. $25 billion annualized revenue as of February 2026. The most downloaded app of 2025 with 770 million installs.

Now look at the trend lines.

Mobile app market share (Apptopia):

• January 2025: 69.1%
• January 2026: 45.3%
• March 2026: below 40%

Web traffic share (Similarweb):

• January 2025: 87.2%
• January 2026: 60–65%, depending on source

Competitor growth in the same window:

• Gemini: 5.4% to 21.5% web traffic share — a 370% YoY surge
• Grok: 1.6% to 15.2% mobile share at peak, currently around 13.5%
• Claude: under 2% to 10% mobile share in three months

That's a 19 to 22 point market share collapse in twelve months. According to Similarweb, this is the steepest decline of any dominant tech platform in recent memory. ChatGPT is not losing because it stopped growing. It's losing because everyone else is growing faster, in some cases by an order of magnitude.

The Yahoo Parallel Isn't Cute, It's Structural

Yahoo's death wasn't caused by Google having a better search bar. It was caused by Yahoo's leadership confusing scale with strategy. Four specific failures killed them — and ChatGPT is repeating three of them already.

1. Portal Sprawl Over Product Focus

Yahoo's fatal pivot was the portal strategy: cram every feature onto one page so users never leave. Mail, news, sports, finance, search, chat, video, shopping. It looked like dominance. It was actually a confession that the company didn't know what it was.

OpenAI is doing the exact same thing in 2026. In the last twelve months they've shipped Sora (since shut down), the Atlas browser, ChatGPT-as-shopping-assistant, ChatGPT-as-presentation-editor, advertising rollouts on the free tier, a humanoid robotics research arm, and a hardware device with Jony Ive. Sam Altman called an internal "code red" in December 2025 specifically because the team had lost focus on the core product.

When your CEO has to issue a code red to refocus on the thing that made you famous, you are not winning. You are spreading.

2. Distribution Beats First-Mover Advantage

Google didn't kill Yahoo Search by being cleverer. It killed Yahoo by owning the surfaces. Chrome, Android, the browser address bar, default search agreements. Distribution ate the directory.

The same thing is happening to ChatGPT right now, just compressed into eighteen months. Gemini's 370% growth wasn't driven by a better model. It was driven by being baked into Google Search, Workspace, Android, and Chrome. 73% of Gemini enterprise accounts use it through Workspace. That isn't a chatbot win. That's distribution doing what distribution always does.

ChatGPT is a destination. Gemini, Copilot, and increasingly Claude (via Anthropic's enterprise integrations and Claude Code) are features inside surfaces people already live in. Yahoo learned this lesson the hard way. OpenAI is learning it right now.

3. Burning Cash to Stay Relevant

Yahoo spent its decline years acquiring Tumblr for $1.1 billion (later written down to $260 million), buying Broadcast.com, and chasing initiatives that never returned capital. The spending masked the fact that the core was eroding.

OpenAI's numbers make Yahoo's spending look conservative:

• Projected $14 billion in losses for 2026 alone
• Projected $74 billion in operating losses by 2028
• Cumulative cash burn projected at $115 billion through 2029
• Burn rate holding at 57% of revenue through 2027
• $1.4 trillion in compute commitments signed over the next eight years
• Spending $1.69 for every $1 of revenue

Compare that to Anthropic, which forecasts dropping its burn to roughly one-third of revenue in 2026 and 9% by 2027, with break-even targeted for 2028. Anthropic is also explicitly avoiding the costly forays into image and video generation that are bleeding OpenAI dry. One company is building a business. The other is building a moonshot and hoping the dominance bet pays off before the cash runs out.

OpenAI is also leaning on tactics that should look familiar to anyone who watched late-stage Yahoo: introducing ads on the free tier (Altman's own "last resort"), scrambling for sovereign wealth fund money in the UAE, and renegotiating with Microsoft to defer revenue share payments to later years. These are not the moves of a company in command of its market.

4. The Brand Becoming a Generic Verb

This is the one that should scare OpenAI most. "Yahoo it" never caught on as a verb because Yahoo was always one of several. "Google it" did, and that linguistic monopoly was worth billions.

ChatGPT briefly looked like it would own the verb. People said "ask ChatGPT" the way they said "Google it." That window is closing. Apptopia found that one in five AI users now uses multiple apps. Power users on Claude are spending an average of 139 minutes per day in the app, up from 98 minutes in February 2026. When users diversify their tools, the generic verb dies. When the verb dies, the moat dies with it.

What's Different (And Why It Might Not Save Them)

To be fair, ChatGPT has two things Yahoo never had.

First, real revenue. $25 billion annualized is not a vanity number. 92% of Fortune 500 companies have ChatGPT seats. The enterprise contracts are sticky in ways Yahoo's banner ads never were.

Second, genuine technical leadership in some domains. ChatGPT still has the deepest brand recognition in AI, and OpenAI's research org continues to ship.

But here's the hard part. Yahoo also had real revenue at its peak. Yahoo also had the deepest brand recognition. Neither saved them once the structural conditions shifted: distribution moved to competitors, focus dissolved into sprawl, and the burn rate outpaced the strategy. OpenAI is currently three for three on those conditions.

What This Means If You're Building

For senior engineers and founders watching this play out, three takeaways are worth internalizing.

Don't bet your stack on a single model provider. The capability gap that justified ChatGPT-only architectures in 2023 is gone. Gemini 3, Claude Opus 4.7, and Grok are at functional parity for most production use cases. Build with provider abstraction in your code. The companies that hardcoded openai.ChatCompletion.create everywhere are about to spend Q3 ripping it out.

Distribution is the moat, not the model. If you're building a SaaS product, the lesson here is identical to the Yahoo lesson. Whoever owns the surface where the user already works wins. Embed your AI into existing workflows (Slack, email, calendar, IDE, the browser) instead of asking users to come to you.

Watch the burn ratio, not the headline revenue. OpenAI's $25B revenue is impressive until you remember they're spending $1.69 for every dollar earned. Yahoo had real revenue too. So did Pets.com. Cash flow positivity in 2030 is a long time to wait for a company whose share is dropping 22 points a year.

The Question Worth Asking

ChatGPT isn't dead. It probably won't die. Yahoo isn't dead either; it's just irrelevant, which for a company that was once worth $125 billion amounts to the same thing.

The real question is whether OpenAI's leadership recognizes the pattern in time to fix it, or whether they keep doing what every late-stage Yahoo CEO did: announcing bold new initiatives, acquiring side projects, defending the headline user count, and missing the actual signal hidden in the share data.

What would you do differently if you were running OpenAI right now — and at what point does the answer become "you can't"?