I really can't understand why so many people are bullish on Cerebras $CBRS. The more I research this company, the more uneasy I feel. It's not just about whether it's worth this price; what if all the revenue data you're looking at is fake? What if it's actually just a lab experiment that's hard to scale? If this company isn't public, its actual cash flow is already at zero.
The problem is: what are you actually pricing it at?
Is the revenue real?
Out of a GAAP net profit of $237.8 million, $363.3 million comes from one-time non-cash gains (related to debt cancellation from the G42 contract restructuring). After adjusting for that, the non-GAAP net loss hits $75.7 million, a year-over-year increase of 247%.
But if you dig deeper, over 75% relies on G42 (Middle East) plus the year-end rush from OpenAI. More critically, Cerebras signed a contract with OpenAI that included a 10% equity warrant, practically a giveaway, which will offset future revenue based on delivery milestones. GAAP accounting treats these warrants as revenue, but the gross profit from existing operations is also below competitors.
OpenAI also lent Cerebras $1 billion in operating loans (at a 6% annual interest rate) to help it build data centers, run compute power, and then sell services back to OpenAI.
Left hand lending money, right hand recognizing revenue. Does that really count as income? Will other companies facing such a tight supply of compute power be willing to pay up, aside from OpenAI?
That $20 billion big order, has the money come through? No.
That shocking $20 billion order OpenAI has only actually paid about $1 billion (in loan form that needs to be repaid), not even counting the deposit
TSMC's 3nm capacity is in high demand, with a $24.6 billion order backlog, but there’s no long-term capacity guarantee agreement between Cerebras and TSMC; all wafer orders are one-time purchases. If no products are produced, where's the profit going to come from?
The problem is: what are you actually pricing it at?
Is the revenue real?
Out of a GAAP net profit of $237.8 million, $363.3 million comes from one-time non-cash gains (related to debt cancellation from the G42 contract restructuring). After adjusting for that, the non-GAAP net loss hits $75.7 million, a year-over-year increase of 247%.
But if you dig deeper, over 75% relies on G42 (Middle East) plus the year-end rush from OpenAI. More critically, Cerebras signed a contract with OpenAI that included a 10% equity warrant, practically a giveaway, which will offset future revenue based on delivery milestones. GAAP accounting treats these warrants as revenue, but the gross profit from existing operations is also below competitors.
OpenAI also lent Cerebras $1 billion in operating loans (at a 6% annual interest rate) to help it build data centers, run compute power, and then sell services back to OpenAI.
Left hand lending money, right hand recognizing revenue. Does that really count as income? Will other companies facing such a tight supply of compute power be willing to pay up, aside from OpenAI?
That $20 billion big order, has the money come through? No.
That shocking $20 billion order OpenAI has only actually paid about $1 billion (in loan form that needs to be repaid), not even counting the deposit
TSMC's 3nm capacity is in high demand, with a $24.6 billion order backlog, but there’s no long-term capacity guarantee agreement between Cerebras and TSMC; all wafer orders are one-time purchases. If no products are produced, where's the profit going to come from?