Why did the market interpret Meta's plan to rent compute capacity as negative despite no CapEx cuts?
Answered at 0:41 in Why the Markets Are Pricing AI Wrong | Gavin Baker · Invest Like The Best
Why did the market interpret Meta's plan to rent compute capacity as negative despite no CapEx cuts?
When Meta announced it would rent out compute capacity, the market interpreted this as very bearish and expected CapEx cuts. However, Meta did not cut CapEx. What actually happened was that Meta saw SpaceX sell optimized trading clusters at massive premiums to contracted rates and recognized the same opportunity. Meta's telemetry on CapEx plans showed no shift and continued getting more aggressive. Shortly after, Meta also released Llama 1.1, described as the best model in a long time, contradicting any narrative of reduced AI investment.
Full answer and detailsRead the whole answer, key facts and every other question from this episode
In short
When Meta announced it would rent out compute capacity, the market interpreted this as very bearish and expected CapEx cuts.
Gavin Baker, Investor at Invest Like The Best · Why the Markets Are Pricing AI Wrong | Gavin Baker · Invest Like The Best · answer at 0:41
Key takeaways
- When Meta announced it would rent out compute capacity, the market interpreted this as very bearish and expected CapEx cuts.
Quote to remember
“I haven't been able to find one that is like a quantitative metric. The underlying fundamentals are improving and NVIDIA is actually at its lowest forward PE of the last 10 years.”
— Gavin Baker
The full answer · about 1 min
When Meta announced it would rent out compute capacity, the market interpreted this as very bearish and expected CapEx cuts. However, Meta did not cut CapEx. What actually happened was that Meta saw SpaceX sell optimized trading clusters at massive premiums to contracted rates and recognized the same opportunity. Meta's telemetry on CapEx plans showed no shift and continued getting more aggressive. Shortly after, Meta also released Llama 1.1, described as the best model in a long time, contradicting any narrative of reduced AI investment.
Key facts
- NVIDIA at lowest forward PE of last 10 years as of recording
- GPU rental prices rising 50-60% in 6-7 months (mid-$2 to near $4 per GPU hour for Blackwell)
- Hyperscaler operating cash flow acceleration from 28% to 32% (35% adjusted for one-time items)
- 2 trillion dollars in potential hyperscaler operating cash flow if Blackwell/Rubin monetize above Ampere rates
- 700 billion dollars of credit demand potentially removed by improved operating cash flow
People, companies and topics mentioned
More answers from this episode
- 0:19Why is NVIDIA trading at its lowest forward PE in the last 10 years despite accelerating AI demand metrics?
- 0:28What is the primary factor that would reverse the current bullish AI infrastructure outlook according to Gavin Baker?
- 0:38What is driving GPU rental price increases despite expectations they would decline in 2024-2025?
- 0:44How are hyperscaler operating cash flows accelerating despite market concerns about AI infrastructure financing?
- 0:44How does open source model growth affect AI infrastructure demand and compute margins?