# $40 BILLION AI Company JUST went INSOLVENT | The AI Bubble.

Source: https://www.youtube.com/watch?v=gkB9ROq87Vk
Recap page: https://rapidrecap.app/video/gkB9ROq87Vk
Generated: 2026-01-07T00:36:11.174+00:00

---
## Quick Overview

The AI Capex cycle is currently nearing its peak, potentially entering the late cycle/infrastructure crash phase much earlier than historical precedents suggest, primarily due to the $20 billion in HBM chip costs and the insolvency risk faced by companies like CoreWeave if they cannot secure immediate bailouts or rapidly increase revenue generation from their existing infrastructure.

**Key Points:**
- The AI Capex cycle is currently assessed to be nearing the 'Sold Out HBM Peak' (Innings 7-9, Late Cycle) stage, with aggressive investment curves suggesting an imminent infrastructure crash.
- CoreWeave's financial situation is precarious; they are burning through cash ($500M/month from operations, $4.75B/9mo on CapEx) while facing $7.9B in current debt due by March 2026, suggesting insolvency risk without immediate bailouts.
- Elon Musk's xAI recently raised $15 billion (though he called the report 'false'), which illustrates the high valuation/risk environment, but the speaker notes that Nvidia's FSD-like innovations are still ahead of the curve, forcing competitors to keep spending to catch up.
- The video highlights that HBM memory costs are a critical factor, noting that Rubin chips (Rubin -10x) will be 10 times cheaper per token than current chips, potentially leading to a rapid collapse in HBM pricing once supply catches up.
- The speaker projects that the overall AI Capex cycle may crash sooner than expected, leading to an 'Infrastructure Crash' and potential insolvency for highly leveraged players like CoreWeave, while only the most fundamentally sound companies ('Survivors') will remain profitable.
- The current environment is compared to the dot-com bubble, with analysts warning that the AI bubble might be 17 times bigger than the dot-com bust, driven by massive capital expenditure on unutilized data center capacity.

![Screenshot at 00:54: The Coreweave Cycle diagram illustrates the four stages of AI Capex \(GPT Moment, Infrastructure Build, Application Era, Late Cycle/Game Over\) showing how high capital expenditure \(red line\) and risk \(blue line\) eventually lead to an infrastructure crash and collapse, with only 'Survivors' remaining profitable.](https://ss.rapidrecap.app/screens/gkB9ROq87Vk/00-00-54.jpg)

**Context:** The video analyzes the current state of the Artificial Intelligence Capital Expenditure (Capex) cycle using a baseball analogy (Innings 1-9) to map growth and risk. Key concepts involve the massive spending on NVIDIA chips and HBM memory needed to build out AI infrastructure, contrasting the high-risk environment for heavily indebted AI infrastructure providers like CoreWeave against major tech companies like Google, Amazon, Meta, Apple, Microsoft, and Elon Musk's xAI.

## Detailed Analysis

The speaker argues that the AI Capex cycle is entering its late stage, characterized by an 'Infrastructure Crash' and significant risk, evidenced by the steep decline shown in the AI Capex Cycle Progress & Risk chart (00:54). This cycle is driven by massive investment in NVIDIA chips and HBM memory, which leads to data center construction and eventual revenue generation, but this process is currently causing severe financial strain for certain players. The speaker points to CoreWeave's financials, noting that they are burning $500 million per month from operations and spending $4.75 billion over nine months on CapEx, while holding $7.9 billion in current debt due by March 2026. The speaker suggests CoreWeave might be insolvent without external help, pointing to the fact that nearly 30% of their assets are idle data center capacity. Furthermore, the speaker highlights the cost differential between current chips and upcoming technology like the NVIDIA Rubin GPU (which is projected to be 10x cheaper per token) and the potential for HBM memory prices to collapse, which would severely impact the economics of the current cycle. The speaker concludes that investors should be cautious because while the AI boom is real, the current market environment is heavily inflated, similar to the dot-com bubble, and many companies are overly reliant on continuous fundraising, which is already showing signs of slowing down (as indicated by the money movement chart for xAI showing recent dips in funding inflow). The ultimate outcome predicted by the cycle chart is a massive infrastructure crash, followed by a small number of 'Survivors' who managed their capex and innovation effectively.

### AI Capex Cycle Stages

- GPT Moment (Innings 1-3)
- Infrastructure Build (Innings 4-6)
- Late Cycle/Peak (Innings 7-9)
- Cycle End (Game Over)
- Infrastructure Crash
- Survivors

### CoreWeave Financial Risk

- Burning $500M/month from operations; $7.5B in debt due by March 2026 (some at 15% interest); 30% of assets are idle data center capacity.

### NVIDIA's Innovation Impact

- Upcoming Rubin chips are expected to be 10x cheaper per token than current chips, suggesting HBM memory costs/prices will soon collapse.

### Elon Musk's xAI Fundraising

- Reportedly raising $15B, though Musk called the report 'false', showing capital is still flowing to top AI players.

### Financial Analysis of Sandisk (as an analogy)

- Showed $438M FCF in 3 months, but high debt servicing costs ($310M/quarter in interest) and significant cash outflow for CapEx (over $6.2B in 9 months) illustrate potential instability.

### Cycle Warning Signs

- The current position is near the 'Sold Out HBM Peak' on the red (Capex Scale) curve, indicating high risk before the crash.

### Investment Thesis

- The speaker is bullish on the long-term application layer of AI but bearish on the current hardware/infrastructure build phase due to high valuations and unsustainable capex.

![Screenshot at 00:01: Headline graphic stating 'AI Drives 90% of US GDP Growth in 2025, Raising Recession Risks'.](https://ss.rapidrecap.app/screens/gkB9ROq87Vk/00-00-01.jpg)
![Screenshot at 00:37: The speaker introduces a filing from CoreWeave showing amendments to their credit agreement, including reducing minimum liquidity requirements.](https://ss.rapidrecap.app/screens/gkB9ROq87Vk/00-00-37.jpg)
![Screenshot at 00:54: The core graphic illustrating the four phases of the AI Capex Cycle: Early Cycle, Infrastructure Build, Application Era, Late Cycle/Game Over.](https://ss.rapidrecap.app/screens/gkB9ROq87Vk/00-00-54.jpg)
![Screenshot at 03:48: A section of the CoreWeave credit agreement highlighting covenant adjustments and equity cure provisions.](https://ss.rapidrecap.app/screens/gkB9ROq87Vk/00-03-48.jpg)
![Screenshot at 15:27: A screenshot of Sandisk's cash flow statement showing significant cash burn from investing activities \($6.249B in 9 months\) against operating cash flow \($1.499B in 9 months\).](https://ss.rapidrecap.app/screens/gkB9ROq87Vk/00-15-27.jpg)
