The Google Cloud Illusion: AI Hype, Anthropic, and the Enterprise Security Trap

Google Cloud reported a 63% surge in revenue, hitting $20 billion in a single quarter. The narrative is that the “enterprise AI revolution” has arrived. This is a lie.

The growth is not broad adoption. It is a hyper-concentrated bubble driven by a single customer and a massive, futile spending spree on security band-aids.

The Anthropic Dependency

The staggering $462 billion backlog—cited as proof of future demand—is a mirage.

Reports confirm that Anthropic signed a single, 5-year contract worth $200 billion.

  • The Concentration: Anthropic alone accounts for 43% of Google’s total backlog.
  • The Loop: Google is an investor in Anthropic. Google funds Anthropic, and Anthropic buys compute from Google. It is a circular flow of venture capital, not organic market demand.

If you strip away Anthropic and other frontier labs (OpenAI, xAI), the remaining backlog growth is linear and ordinary. This isn’t a boom; it is a single bet placed by a company spending its investor’s money on another company’s infrastructure.

The “Real” Revenue is Inflation

Even the current $20 billion quarterly revenue is inflated by the same bubble.

  • Startup Burn: AI labs are burning billions in VC cash on training runs. This flows directly into Google’s pockets.
  • The AI Tax: Traditional enterprises are paying 6x premiums for “AI-ready” infrastructure out of FOMO.
  • Internal Spending: A significant chunk of Google Cloud revenue comes from other Alphabet divisions buying their own chips.

The “boring enterprise utility” argument is a cover. Companies are panic-buying capacity they don’t need, fueled by a fear of being left behind.

The Profitability Split

Who is actually making money in this ecosystem? Not the frontier labs.

  • The Value Destroyers: OpenAI and Anthropic burn billions to train models, hoping to sell them for pennies. They are research projects, not businesses.
  • The Value Extractors: Companies like Baseten and Fireworks AI are profitable. They don’t train models; they run them. They sell inference infrastructure—plumbing, not science.

The cloud giants are profiting from the destroyers, while the extractors quietly build sustainable utility businesses.

The Enterprise Security Paradox

Why are “boring” companies like Glean making hundreds of millions? They are selling a liability shield.

The core promise of Enterprise AI—“Chat with your company’s data”—is a security nightmare.

  • The Conflict: AI (LLMs) is probabilistic and open. Security (RBAC) is deterministic and closed.
  • The Paradox: If you lock the AI down to strict permissions, it becomes a useless text summarizer. If you open it up to be useful, it leaks the CEO’s salary.

Glean sells a “security wrapper” to bridge this gap. It claims to map permissions in real-time across Slack, Jira, and Drive.

The Wild Goose Chase

Glean’s solution is a wild goose chase. It is XKCD 927 in action: creating a new standard to fix the problem of too many standards.

Glean doesn’t fix the “oversharing link” problem. If an employee shares a file with “anyone in the company,” Glean sees it as authorized. The leak happens anyway.

The software is 90% traditional database plumbing and 10% AI wrapper. It exists so CISOs can tell the board, “We bought the guardrails.” It is a compliance purchase, not a technical solution.

Conclusion: Renting the Goose Chase

Google Cloud’s explosion is real cash, but it is built on structural rot.

It is renting data centers to a single massive customer (Anthropic) and selling panic infrastructure to enterprises chasing a security solution that doesn’t exist. The revenue is real, but the foundation is sand. The moment the VC funding for the labs dries up, or the moment enterprises realize they bought a liability shield that doesn’t work, the bill comes due.