Anthropic *SUDDENLY* Delays IPO & Nscale IPO Looks BAD
Overall Thesis
Anthropic's repeated IPO delays and Nscale's concerning S1 filing suggest potential financial or competitive weaknesses in the AI startup sector.
Narratives
The host believes Anthropic's repeated IPO delays (September to October to November) signal that insiders and prospective investors are uneasy with the company's financials after reviewing pre-IPO documents. He argues Anthropic is losing market-leadership on model benchmarks, is deeply unprofitable relative to its compute costs, and is timing its IPO to catch a bull market rather than for fundamental strength.
Key Arguments
- Anthropic has delayed its IPO multiple times, most recently pushing from October to November
- Some advisors reportedly want to wait for Q3 financials to show a 'stronger competitive position,' which the host finds suspicious given Anthropic's fading leaderboard position versus GPT-5.1/Astra
- Anthropic is targeting ~$100B ARR by year end (up from ~$9B) while needing 5 gigawatts of compute, implying ~$20B revenue per gigawatt versus Elon Musk's reported $40B/gigawatt cost structure — 'these numbers are massively upside down'
- 100 VC investors met with Anthropic and reportedly still suggested delaying the IPO, which the host says 'doesn't sound bullish'
- Anthropic's growth projections rely partly on compute supply from Nscale, whose own S-1 shows financing gaps, making Anthropic's plan 'building on a sand castle'
The host describes Nscale's S-1 filing as containing multiple red flags: disclosed material weaknesses in internal controls, a $44.6 billion Anthropic services deal for which 'we have not obtained binding commitments' for financing, heavy customer concentration, and a huge gap between active GPUs (25,000) and committed GPU purchases (461,000). He compares the setup to 19th-century bear-pelt speculation, arguing Nscale is selling compute it cannot yet deliver or finance.
Key Arguments
- Nscale disclosed 'material weaknesses' in internal control over financial reporting in its PCAOB audit
- The company entered a $44.6 billion services agreement with Anthropic but states it has 'not obtained binding commitments for any of the financing required' to fund it
- Largest customer (implied Microsoft/Anthropic) accounted for 52% of revenue in H1 2026, a concentration risk
- Balance sheet shows $1.44B cash versus $2.8B long-term debt and $24B in outstanding purchase commitments not yet delivered
- Only 25,000 active GPUs versus commitments for 461,000 GPUs — a ~18x gap the host compares to running a marathon and then committing to run 20x further
- CEO stands to collect roughly $350 million in an IPO award tied to 2.5% of shares if the company goes public
The host calls CoreWeave 'another disaster,' noting the company does not have enough cash to cover near-term bills despite raising additional capital. He implies the AI-infrastructure financing model across these companies is unsustainable without continuous new fundraising.
Key Arguments
- CoreWeave has about $8 billion in cash and receivables against $18 billion in bills to pay
- CoreWeave raised a $3 billion convertible note upsized to $3.7 billion, but the host says even adding that amount 'is still not enough to pay their bills over the next 12 months'
The host views Nebius's balance sheet as comparatively healthier than CoreWeave's or Nscale's, citing sizable cash and plant/property/equipment relative to debt, though he does not hold a position and offers no forward price view.
Key Arguments
- Nebius has roughly $8 billion in cash and $13 billion in plant, property and equipment against $8.5 billion in long-term debt
- The host states 'their balance sheet is actually okay' compared to peers
The host frames Nvidia as a net winner regardless of how the AI infrastructure financing plays out, since it collects cash upfront from GPU sales to companies like Nscale, and is also providing convertible-note financing to help customers afford its chips.
Key Arguments
- Nvidia is lending Nscale roughly a billion dollars via a convertible note to help fund GPU purchases
- The host states Nvidia 'is in a great spot because they collect the cash right away,' regardless of downstream financing risk at customers
Hedges & Caveats
- No specific price targets or directional predictions made
- Analysis based on speculation about insider knowledge and document review
- Acknowledges uncertainty about reasons for delays ('what are the insiders seeing that we're not seeing')
- Commentary on competitive positioning (Claude vs. GPT-4o/Astra) is observational, not predictive