Google Made Big Announcements. Now Check What's Happening Inside.

On Tuesday, May 19, Alphabet CEO Sundar Pichai stood in Mountain View and told the world Google was back on offense. New models. Lower prices. AI agents baked into Search and YouTube. Pichai even told journalists that enterprise customers switching to Google's AI could save more than $1 billion a year.

Pichai didn't mention what was happening inside the company.

The Announcements, By the Numbers

Google launched Gemini 3.5 Flash at I/O 2026, a model targeting coding and automated tasks, according to Reuters via Global Banking & Finance Review. Gemini 3.5 Pro is coming next month.

Pricing moved aggressively. The AI Ultra subscription — Google's top-tier plan — dropped from $250 to $200 per month. A new $100-per-month tier was announced specifically for developers and work-related users.

Google also rolled out autonomous AI agents embedded directly in Search — capable of making purchases, monitoring ticket availability, and managing schedules in real time.

"When people use our AI-powered features in Search, they use Search more," Pichai said. The business model depends on it.

DeepMind CEO Demis Hassabis went further. "When we look back at this time, I think we will realize that we were standing in the foothills of the singularity," he said. Shipping coherent coding products has proven harder than singularity talk.

The Part Google's PR Team Hoped You'd Miss

While Pichai was on stage talking about billion-dollar savings, the Los Angeles Times published a detailed April 22 report laying out exactly how Google got caught flat-footed in AI coding — the fastest-growing and most profitable segment of the enterprise AI market right now.

The problem isn't capability. Google's models benchmark well. The problem is organizational dysfunction.

Google's coding features are scattered across half a dozen different products with different branding, according to current and former Google employees and executives cited by the LA Times. No unified experience. No clear entry point for businesses.

Some Google engineers don't even use Google's own tools. They prefer Anthropic's Claude Code instead, according to the LA Times sources. Internal decisions like that signal deeper problems.

Some Google engineers are also struggling to adopt AI coding tools altogether.

What Google Is Actually Doing to Fix It

Chief Research Officer Koray Kavukcuoglu is now working with Google's main engineering team to consolidate internal AI coding tools under Antigravity, a platform released last year, according to a Google spokesperson cited by the LA Times. The work is expected to conclude in the coming weeks.

DeepMind is forming a new coding-focused team led by research engineer Sebastian Borgeaud, per the LA Times, which credits the Information for first reporting it.

John Jumper — who shared the 2024 Nobel Prize with Demis Hassabis — is also being put to work on AI coding, according to a source familiar with the matter cited by the LA Times.

Putting Nobel laureates on a product reorganization signals Google knows it has a real gap to close.

What Mainstream Coverage Is Getting Wrong

Most tech coverage of Google I/O 2026 treated it as a triumphant comeback story. Google's stock has surged. Alphabet recently came within striking distance of Nvidia as the world's most valuable company, according to Reuters.

But the coverage is running two separate stories in isolation — the I/O announcements on one track, the internal chaos reporting on another. The I/O announcements and the internal dysfunction are connected.

Google is cutting prices because it has to. Anthropic and OpenAI have locked up enterprise coding customers while Google was busy running competing internal teams that couldn't agree on a unified product. The $50 price cut on Ultra subscriptions is a defensive move.

Pichai's line about companies "blowing through their annual token budgets" by May is real — enterprise AI spending is exploding. The question is who captures that spending. Based on the LA Times reporting and Google's own internal behavior, Anthropic and OpenAI are currently in the lead.

The Competitive Reality

Anthropic sold its coding tools by focusing. One product. Clear value proposition. Businesses understood it.

OpenAI did the same. Both companies turned AI coding into a repeatable revenue engine while Google's equivalent effort was being managed by committees across multiple divisions with overlapping mandates.

Gemini 3.5 Flash may genuinely close that gap — the benchmarks from late 2025's Gemini 3 release were legitimately strong, according to the LA Times. But benchmark wins don't automatically translate to enterprise contracts when your go-to-market is a mess.

What This Means for Regular People and Businesses

If you're a developer or a company shopping for AI coding tools right now, Google just made itself cheaper and announced a real consolidation effort.

But "coming in the next few weeks" is not a product. It's a promise. Anthropic and OpenAI have paying customers and no reason to slow down.

Google has the infrastructure, the talent, and now apparently the urgency. Whether that translates into execution remains to be seen. The I/O stage show was polished. The internal product still isn't.