Is CoreWeave a Buy Before Earnings? $99 Billion of Backlog Against $1.27 of Losses

TL;DR

• No before the print, and the July tape is the reason. The stock fell from the low $90s to near $60 on July 29 and is back at $90.67 twelve days later. This market has no settled opinion of CRWV, and unsettled opinions plus earnings equals violence.
• The bull's exhibit: a contracted backlog near $99.4 billion, fattened by Meta's expanded commitments. The bear's exhibit: an expected loss of about $1.27 per share and $31-35 billion of capital spending to deliver that backlog.
• The Street leans bull: Deutsche Bank just raised its target to $150 from $135, and the average target sits near $138 against a $90 stock.
• Consensus wants $2.56 billion of revenue, up about 112%.

Is CoreWeave a Buy Before Tuesday Night?

No. CoreWeave is the purest available bet on one question: whether contracted AI compute demand converts into profitable delivery before the funding costs eat the contract value. A $99.4 billion backlog against roughly $5 billion of annualised revenue means the market is not buying a business, it is buying a construction schedule with customers attached.

I find the backlog genuinely impressive, and we have covered how it got there. But the July round-trip, $90 to $60 to $90 inside five weeks on no fundamental change, is the market confessing it cannot price this thing. When the holders themselves do not agree on a valuation within 50%, an earnings print is not a catalyst, it is a lottery drawing. The 52-week range, $60.55 to $153.20, says the same thing louder.

What Tuesday Has to Show

The revenue number will not decide anything; it is guided and contracted. Three disclosures would:

• Backlog conversion pace. How much of the $99.4 billion turns into revenue in the next four quarters, not the next five years. A backlog that mostly converts post-2028 deserves a bigger discount than the bulls apply.
• The funding bridge. $31-35 billion of capex against current cash flow means more debt or more equity. Which one, at what cost, is the whole margin story.
• Customer concentration. Meta got bigger this quarter. Whether anyone else did decides if this is a diversified utility or a two-customer construction contract.

Clean answers on those three and I get constructive on weakness, because the bull case would finally have arithmetic instead of a number that merely sounds large. Until then this stays a watch, and Super Micro reporting the same evening makes Tuesday night the AI-infrastructure verdict either way.

The One-Line Read

CoreWeave into Tuesday is a $99 billion promise wearing a $1.27 loss, on a stock the market repriced by 50% twice in five weeks: no before the print, and the backlog-conversion number on the call, not the revenue line, decides whether the answer changes.