📉 benchmark underperformer 🍀 rarely lucky 🧠⚠️ dumber than you think!! 💸 no financial advice 🧑🏻💻 learning how to computer 🤡 unrealistic expectations
Joined February 2018
- Tweets4.4K
- Following897
- Followers217
- Likes189K
⚠️This is me shitposting ⚠️
We've come so far in the past few years from GPT-4o to present day Qwen 27B and people are still underestimating how quickly things can scale.
Many know about the Cerebras Chip, but how many have thought of the Cerebras Slab, or even the Cube?
🤖 Made with AI
$CBRS might be the most interesting AI hardware stock to buy now and come back to in 3 years.
UBS has Cerebras going from $890M in revenue this year to $3.3B in 2027 and ~$22.6B by 2030, with operating margins swinging from −18% to 46% along the way.
At $146, the stock trades at about 20x UBS’s 2028 earnings and 10x 2029, and UBS has EPS compounding from $7.21 to $24.36 over those two years.
It’s also ~26% below the $196.74 share price UBS built the model on just two weeks ago. Same numbers, cheaper stock.
Remember I keep saying this: The growth story is inference.
Cerebras keeps the memory on the wafer itself, which is exactly what you want when agents fire off dozens of model calls per task and every millisecond of latency adds up.
Execution is everything here. But if they get even halfway there, $146 won’t look expensive for long ⚡
Stonk: AUROORA.HE
@BerkelKip @PenttiJokinen @JariValtanen
saw the potential when I didn't. 🥲
(Caveat: image is from my clanker)
Qwen Code Desktop now has that handy recap (box above the input text field) that shows up after you've left it alone for a bit, to remind you what you were last up to.
I've only seen this in Grok Code previously, very nice QoL feature.
