AI Compute Prices RISE: How SpaceX And Nebius Stocks Benefit
Overall Thesis
Rising GPU compute rental prices indicate strong AI demand and data center buildout, which should benefit Nvidia, Nebius, and SpaceX stocks.
Narratives
The host argues that spot rental prices for Nvidia GPUs (B200, H100, H200) are rising sharply even as global compute supply expands, which he says proves AI demand is outstripping supply. He believes this trend confirms the broader AI trade, including Nvidia, remains intact and will keep climbing as adoption accelerates over the next 6-9 months.
Key Arguments
- B200 rental price rose ~80% in 8 months and 21% in the last month despite massive new supply coming online
- Even six-year-old H100 chips are seeing rising prices, indicating demand is outstripping supply across all chip generations
- Only ~3-5% of businesses currently use AI meaningfully, leaving enormous room for demand growth once a clear breakthrough product emerges
Predictions (1)
The host frames Nebius as a leading 'Neocloud' that is a direct beneficiary of rising GPU rental prices, since Neoclouds are laser-focused on AI compute versus diversified hyperscalers. He personally holds the stock (bought around $175-180 during a dip) and believes it and other Neoclouds will keep rising as long as compute prices stay elevated.
Key Arguments
- Neoclouds are solely focused on GPU/AI compute, giving them speed and cost advantages over hyperscalers like AWS
- Nebius stock is up 147% year-to-date and 127% over one year
- Neocloud stocks should stay long-term correlated with rising GPU rental prices, which the host says are the 'secret' indicator of compute demand
The host is Tesla's largest core position and believes robotaxi/Cybercab scaling is not priced into the stock, arguing that a million robotaxis could generate massive earnings that would justify a much higher share price. He also states plainly that he expects Tesla stock to rise more than SpaceX going forward.
Key Arguments
- 100,000 Cybercabs/robotaxis roughly double Tesla's free cash flow; a million robotaxis could generate $30-50 billion in earnings
- At a million robotaxis, Tesla would trade at a P/E of only ~20 while growing 100% a year, which the host calls 'absolutely not priced in'
- Tesla is described as the host's 'total focus' core stock alongside SpaceX
The host describes SpaceX as the ultimate long-term winner of the AI compute buildout due to its Starship-enabled ability to build compute infrastructure in space, though he notes SpaceX is a privately traded company and expresses uncertainty about near-term stock price ceiling versus its current ~$1.95-2 trillion valuation.
Key Arguments
- SpaceX's reusable Starship could make space-based compute buildout cheaper than on Earth via AI satellites
- The host views recent private share unlock prices (~$135-150) as a trading range, calling $135 a robust floor but uncertain about upside given the already-high ~$2 trillion valuation
- He expects Tesla stock to outperform SpaceX going forward
Hedges & Caveats
- Analysis based on emerging GPU rental marketplace data with limited historical precedent
- Acknowledges possibility of overbuilding and supply-demand imbalance as risk factors
- Speculative connection between compute pricing and stock performance not yet proven