Dan

chillin and grillin

[Dan @dpa685]

@dsabs11 @bitcoinbutcher1 This is the datacenter level unlevered unit economics. The hope is to multiply these economics by a massively enlarged asset base to amortize/spread the corporate G&A load against. That’ll take time as building this out is not quick.

[Dan @dpa685]

@bitcoinbutcher1 Prior bear argument of low return on capital business , below the WACC, has been debunked at current GPU rates

[Dan @dpa685]

@jiahanjimliu @conceptualnomad I am bullish on a 3 year view and the capital flywheel from GPU prepayments but I'm highly skeptical how much DC financing you can get against your cloud business with 3-5 year GPU contracts, some IG / some not, versus the actual cost to build the DC. It's a very low LTV.

[Dan @dpa685]

@bitcoinbutcher1 And on cue …

[Bloomberg @business]

The flood of debt financing for AI capital expenditures is causing "indigestion" in fixed-income markets, Pimco says https://t.co/rzyiQZSZJP

[Dan @dpa685]

I think also just funding $300-350 B across 5 GWs is such a staggering number this is going to extend much longer than people think which in of itself will help prevent a bubble and keep industry supply constrained and compute pricing elevated.

Theres only so much aggregate dollars Blue Owl and Pimcos of the world can put into the entire sector. Industry supply delivery cadence is much slower than the market realizes.

[Dan @dpa685]

@InvestNorthwise 20-25 million / MW + 75% gross margin + 7 million /MW depreciation is ~16% unlevered pretax ROIC. 80-90% leverage on that at under 10%. It's not bad !

[Dan @dpa685]

@InvestNorthwise 1 GW costs 55-60B, what would one expect otherwise? They have >5 GW to build. The blended WACC is the more relevant piece

[Dan @dpa685]

@BullBearBalls @RealJimChanos that's a pretty bizarre comment for Jim to use

[Dan @dpa685]

@bitcoinbutcher1 @cantonmeow do yourself a favor: buy $CIFR, stick it in a drawer for 2 years to allow this to play out, turn off your laptop and save yourself the mental energy of the day to day deluge of news flow in the sector and go live your life

[Dan @dpa685]

@JonahLupton @yianisz Given a range of future GPU procurement costs/cap raising, GPU rental rates and EV / EBIT multiples, what are the bull/bear scenarios a few years out?

[Dan @dpa685]

@kevinxu ding ding ding

cc @mikealfred

[Dan @dpa685]

@Lazarus_Capital @OnodaCapital I was off on my math. $100 to build, $15 NOI (monthly) 3% escalators. $50 terminal value and discount it back at 6.2% (AMZN 10 yr WAL credit is yielding 6.2%) I am getting equivalent to a 7.67% cap rate. Anything assigned cap rate valuation tighter would imply a higher TV

[Dan @dpa685]

@Lazarus_Capital @OnodaCapital The datacenter world is going to change so much over 15 yrs it’s not reasonable to have conviction on terminal value economics. For me just provide a basic conservative TV level that is something more than zero and don’t try to over think it

[Dan @dpa685]

@Lazarus_Capital @OnodaCapital 15 yr lease with 3% escalators should be 7.0-7.5% cap. I think 7.25% cap translates to terminal value of 50% of construction cost

[Dan @dpa685]

@kevin__mak The Magnitar founder calls it "strong convictions loosely held".

The beauty of public market investing (non pods) is having both duration tolerance and the optionality to change your mind whenever you want

[Dan @dpa685]

@mikealfred *and on cue today......

[Dan @dpa685]

@mikealfred 3/ Will be timing drag of build out over multiple years, etc, but also if Vera Rubin 40-50MM/MW not crazy to hit 25MM/MW blended across Sweetwater. ERCOT Batch Zero delays only help them. Surprised it is not ripping more

[Dan @dpa685]

@mikealfred *EBT

[Dan @dpa685]

@mikealfred 2/ Add in 80% leverage at 7% (GPU financing, customer prepays, etc) is 60% levered equity returns. On 1.4 GW and 400 million shares (TBD) this is $25/share EBIT and the stock is $44.