By ByteBot
Ali Ghodsi, CEO of $134 billion Databricks, called out the AI startup bubble at Fortune Brainstorm AI this week, criticizing companies “worth billions of dollars with zero revenue” as “clearly a bubble” that’s “insane.” Speaking in San Francisco on December 8-10, Ghodsi predicted the situation will get “much, much, much worse” in the next 12 months before a correction. The irony? His company just raised $10 billion at that $134 billion valuation.
This isn’t another analyst warning—it’s an insider admitting the system is broken while participating in it.
The Circular Financing Scheme Inflating AI Valuations
Ghodsi’s core critique targets “circular financing,” where investors fund companies that buy from each other, creating artificial demand. Nvidia invests $100+ billion in OpenAI, tied directly to how many chips OpenAI agrees to purchase. Amazon commits $8 billion to Anthropic while Anthropic commits $30 billion in Azure compute purchases. The money flows in a circle, inflating revenue figures without real market demand.
The numbers are absurd. Cohere raised $500 million at a $5.5 billion valuation with only $22 million in revenue—a 250x revenue multiple. OpenAI’s valuation increased by $29 billion per month between October 2024 and October 2025. That’s nearly $1 billion per day. Zero-revenue startups command multi-billion dollar valuations based purely on future potential, not current business fundamentals.
This creates a house of cards. When the bubble pops, companies relying on investor-funded revenue will collapse together. Developers using AI tools from these startups need to identify which have real customers versus just investor money propping them up.
The Two-Faced VC Problem: Exhausted but Still Funding
Here’s where it gets uncomfortable for Ghodsi. Databricks itself is valued at $134 billion after raising $10 billion—putting him in the awkward position of criticizing a bubble his company benefits from. He’s not wrong, but he’s also not innocent.
VCs privately tell Ghodsi they’re “exhausted” with the AI hype cycle. Yet publicly, they keep writing checks. This explains why the bubble keeps inflating despite everyone knowing it’s unsustainable. The disconnect between private skepticism and public enthusiasm creates the perfect conditions for a crash.
12-Month Timeline: “Much, Much, Much Worse” Before Correction
Ghodsi predicts the bubble will worsen significantly before correction, giving developers a specific timeline to evaluate their AI tool dependencies. His “12 months from now, it’ll be much, much, much worse” statement points to mid-2026 for market worsening, followed by correction.
He’s not alone in this prediction. 40% of CEOs at a recent Yale summit raised significant concerns about AI exuberance and believe a correction is imminent. The IMF warned the AI bubble could burst comparable to the dot-com crash, though likely less systemic since AI companies have real revenue (unlike many dot-com startups).
For developers, this timeline matters. If you’re choosing AI platforms, tools, or considering joining an AI startup today, you have 12 months before correction separates sustainable companies from hype-driven ones. Choose accordingly.
The Nuance: AI Agents Are Real, Even If the Bubble Exists
Ghodsi isn’t anti-AI. Despite his bubble concerns, he’s bullish on actual AI usage. His key stat: 80% of databases on Databricks are now created by AI agents, not humans. He predicts this will reach 99% within a year.
This distinction matters. Not all AI is bubble. Ghodsi separates zero-revenue hype companies from real AI adoption. Databricks’ 80% AI agent statistic shows genuine value—agents autonomously spinning up infrastructure faster than human developers ever could. This is measurable adoption, not vaporware.
The takeaway: Focus on AI tools showing real usage metrics rather than funding announcements. Databricks demonstrates value through adoption stats (80% agents). Zero-revenue startups demonstrate nothing but investor confidence.
Red Flags vs Green Flags for Developers
How do you evaluate which AI startups will survive? Look for these signals:
Red Flags (avoid these): Zero revenue with multi-billion valuations (250x+ revenue multiples). Major investors who are also major customers (circular financing). Growth driven by funding rounds rather than adoption. “Vibe revenue”—marketing hype without real customer metrics.
Green Flags (favor these): Real usage metrics like Databricks’ 80% AI agent adoption. Capital efficiency generating $5-7 enterprise value per dollar raised. Strong retention and engagement rates. Customers who aren’t also investors. Open source alternatives exist, reducing vendor lock-in risk.
The mid-2026 correction Ghodsi predicts will separate these two groups. Companies with zero revenue and circular financing will struggle when investor money tightens. Companies with real adoption and efficient capital use will survive.
Related: AI Infrastructure Spending Hits $758B by 2029, But Skills Gap Grows Faster
Key Takeaways
- Databricks CEO Ali Ghodsi called out AI startups with “billions in funding, zero revenue” as an “insane” bubble at Fortune Brainstorm AI, predicting it will get “much, much, much worse” in the next 12 months before correction (mid-2026 timeline).
- Circular financing creates artificial demand—investors fund companies that buy from each other (Nvidia-OpenAI $100B+ deal, Amazon-Anthropic $8B funding tied to $30B compute purchases), inflating valuations without real market demand.
- The irony is thick: Ghodsi’s $134B company benefits from the bubble he’s criticizing, and VCs privately admit they’re “exhausted” with AI hype while publicly still funding it.
- Not all AI is bubble—80% of databases on Databricks are created by AI agents (not humans), showing real adoption versus zero-revenue hype; focus on usage metrics, not funding announcements.
- Red flags for developers: 250x+ revenue multiples, circular financing, investor-customers, “vibe revenue.” Green flags: Real adoption stats, capital efficiency ($5-7 EV per $1 raised), retention, non-investor customers, open source alternatives.
Ghodsi may be a hypocrite for criticizing a bubble his company benefits from, but that doesn’t make him wrong. Developers have 12 months to get ahead of the correction by choosing AI tools based on real usage rather than investor hype.
