8/31/26 - $cbrs - Trillion dollar co in making (LONG)Cerebras Systems, Inc. Class ANASDAQ:CBRSVROCKSTAR8/31/26 :: VROCKSTAR :: CBRS Trillion dollar co in making - read this post, can't say fundamentally i disagree w/ much of it: https://x.com/jkeynesalpha/status/2094155757620986174 - a few critical assumptions (no pun to the CITL types)- so if we actually are compute constrained, MW constrained, and the real moat is speed (why? because faster tokens are worth 3-5x as much to the machine) - and we believe that the architecture for new model creation and inference (for said model) are chip-model-specific. - okay here comes the M word... some "math": - so mkt thinks $7.5B of revenue in '28. and we know OAI contract (if cbrs delivers) is worth $20B and that's in the later part of the decade first the bear case:- so 1 GW in 2029 costs $50B in content (minimally) and CBRS is only deployed in 300 MW (you see where I'm going with this... there's a "miss" cons./ which doesn't ever get revised upward of which is $14B). Could it be lower? Lol, yeah obviously. Also, probably not. now the conservative base case:- incremental compute demand about 25-30GW a year. let's say it's 10 GW. and CBRS only gets 5% of very specific model/ partners (which is also conservative given they're putting together a roster of... well... everyone lol). that's 500 MW and we're now at $25B in 2029 (vs. cons of $14B), not "amazing" per se, but multiple would de-risk as they'd bring this online over this period of time. maybe weak sauce software multiple given still growing very high double digits of 7-8x revenue or maybe 10-12x ebitda (remember mgns at this level are probably 70+%). or call it 25B * 7.5 = $190B or round up $200B in 2029. mkt still a non-believer, so need 20% per annum so $200B/1.2/1.2/1.2 = $115B at YE26... still above the $41 today. lol. bull case? you want it? - incremental compute demand of 20B is addressable to cbrs- they take 20% of that load or 4 GW in 2030 or revenue of $200B, which for comparison is about 10% below where cons. thinks AMD will be at that point and mkt thinks that co is worth $756B.- but remember, CBRS will likely still be scaling share of TAM, higher growth and not competing head to head on GPUs like NVDA... so multiple, likely higher, all else equal and you have the making of a $1T mcap company. Now today, again say 20% disc. rate (Ke) and you're at $1,000B/1.2/1.2/1.2/1.2 = ~$500B of implied valuation (given perfect information). you want to blend these for a YE30 target?- epic bear case. goes to zero. vapor nonsense. give it 10% (tho it's likely 0%... but let's assume catastrophic fail).... $0- bear of 40% scenario of $50B (20% above curr cap, deliver very little beyond what's announced)- conservative base of 30% worth $200*1.2=$240B (need to gross up a year from the '29 target we produced above)- bull case, eat cake, throw one at the haters and put one in the fridge... of $1T of $20%= 10% x $0B + 40% x $50B + 30% x $240B + 20% x $1,000B = $292B in 2030 or using 20% Ke today that's (292/1.2/1.2/1.2/1.2) = $140B value today when trading at $41B... basically >3x in my own back of the envelope math. the variant perception here is that you need to assume compute will go multi-chip-modal in the coming years. room for gpus, tpus and the larger wafer designs with all three winning. nvda will gain the most absolute dollar, but cbrs will gain the most growth/ share. and you own that growth/share taker in basically any industry, especially in one that's re-defining the world (AI) and where there isn't a "second best" (don't tell me it's Groq that Nvidia bot, yes Groq is great, but there's no pure play competitor to what cbrs is building... PS - Nvidia tried to buy cbrs before it IPO'd... :) - will be writing much more on this name- it's a 7% spot position for me- i'd add much more in the $170s, probably take it to 10%- I'd be about 15% if we saw low $160s again (mid-term, rates... macro... risk/ beta off etc etc.)- and if we ever went in the sub $150s and nothing fundamental has changed, you better believe this is a 20% position for me.- anything lower... i scale it up.- the asymmetry here is striking. and i didn't think that existed beyond ETH (Sbet is our exposure), but here we are. zigging when the rest of the world wants to zag. - will take time. and that's also part of the edge. we're spot. we're chill. and we do more work unbothered. V