CRITICAL Bubble Warning: OpenAI & Oracle JUST *THREATENED* Elon & SpaceX.
Overall Thesis
The Oracle-OpenAI $300B data center deal reveals that long-term compute pricing is significantly lower than the $50B/gigawatt scarcity pricing SpaceX is receiving, suggesting analyst earnings estimates for SpaceX may be unrealistically optimistic.
Narratives
Kevin views Oracle's recent earnings as a reality check on data-center pricing, showing long-term contracts settle closer to $13-15B per gigawatt rather than the scarcity pricing implied by SpaceX's deal. He believes Oracle has cleaned up its balance sheet, is cutting costs where unprofitable, and is trading cheaply relative to its growth, though its fate is tied to OpenAI's ability to pay its contract.
Key Arguments
- Revenue grew 62% year-over-year to $11.6 billion
- Balance sheet improved with cash growing to $36.3 billion and debt being repaid
- Trading at about 20x earnings against a 30% forecast growth rate, a 0.65 PEG ratio which he calls 'dirt cheap'
- Long-term contract pricing of $13-15B per gigawatt implies a reasonable 3-4 year payback period on data centers
Predictions (1)
Kevin sees the massive Oracle-OpenAI compute deal as strongly positive for Nvidia and other hardware providers, since it signals continued heavy spending on chips and data center buildout.
Key Arguments
- The $300 billion Oracle-OpenAI deal requires 4.5 gigawatts of power, driving major hardware demand
- Called it 'mega good for the hardware providers' and 'great for the chip trade'
Predictions (1)
Kevin views Restoration Hardware as still struggling despite an earnings beat, citing collapsing operating income, rising SG&A costs, and reliance on drawing down inventory to protect margins. He believes the company hasn't yet reached peak pain and it may be too early to be bullish.
Key Arguments
- Operating income fell from $141 million to $107 million year-over-year
- SG&A costs up 20.2% while revenue only grew 2.5%
- Company is depleting cash reserves (down to $41 million) and relying on selling down inventory to offset margin pressure
- Current liabilities of $630 million far exceed $125 million in cash
Predictions (1)
Microsoft is mentioned only as a data point for compute spending comparisons (paying about $6 billion per gigawatt per year), without a distinct investment thesis or prediction.
Key Arguments
- Cited as paying roughly $6 billion per gigawatt per year for compute, used as a market rate comparison
Amazon is mentioned only as a data point for compute spending comparisons (paying about $2 billion per gigawatt per year), without a distinct investment thesis or prediction.
Key Arguments
- Cited as paying roughly $2 billion per gigawatt per year for compute, used as a market rate comparison
Hedges & Caveats
- Acknowledges the deal is positive for hardware providers like Nvidia
- Recognizes Oracle deal is a 'risky gamble' per Wall Street Journal reporting
- Notes OpenAI is a money-losing startup
- Distinguishes between short-term premium pricing deals and long-term market rate comparables
- Expresses concern about analyst underwriting assumptions rather than making definitive predictions