disagree w your article. disclosure: I am short $CBRS writing this so we can play this thru in the next couple quarters and see how it played just going thru points in your article 1] Semi Analysis says Jane Street makes $200m not that it pays $200m and Jane street obviously isn't going to confirm this. do you have source that Jane Street is paying $200m or has a contract for that amt? what is it> 2] Astra Ultra Fast is frontier intelligence and it runs on Nvidia -- do you think that SOL competes w astra? curious on why as no benchmark shows this being the case 3] "OpenAI buys the chips for roughly $7.5M/MW/year" OpenAI doesn't buy the chips. it's delivered through CBRS cloud on 3 yr lease. CBRS owns hardware and carries capex openAI has the *option* to buy in '27. a bit concerning you do not actually understand biz model (?) -- am I wrong here? 4] ""They're renting at roughly 7.5× annual revenue / lifetime hardware cost" per your own math $200M ÷ $7.5M/MW/yr cost = 26.7x 5] "You want to know why OpenAI is no longer offering Ultrafast on Cerebras chips for the public to use? Cerebras is literally at capacity" it never offered Ultrafast publicly on cerebras chips. was internal only for Jane Street, Podium, Basis and Rogo. it was never pulled. GPT-6 Astra Ultrafast runs on Blackwell, and SemiAnalysis says GPT-6.1 Sol Ultrafast runs on NVIDIA at low batch 6] re nvidia you said "it's not the gotcha you think" and never actually explained why. isn't that supposed to be the entire bull case tho... 7] meanwhile * nvidia paid $20b for groq tech in Dec. Groq 3 LPX as of Aug 24 is 3431 tok/ sec, Jalapeño (openAI chip) from Broadcam shipped in Q3 and guided 1.3 GW '27 and >5 GW in '28 which is about 7x the size of Cerebras entire OpenAI contract in 1 year. AMD is buying Talaas. it's a competitive knife fight, and the anchor customer is its competitor. 8] "Cerebras should rent at a huge premium... the scarce capacity should clear at very high prices" cool so core gross margin: 46.5% in Q1 → 40.6% in Q2 → 38–40% guided for Q3. Core operating margin guided to -23% to -25%. GAAP Q2 net loss $450.5M. Cerebras is renting back capacity from G42 at a ~500bp margin hit to meet its obligations do companies w scarcity margins have imploding gross margins trailing and on the guide asking for a friend 9] my favorite part "Tokens per Second Is a Dogshit Metric" your words not mine > entire article is about high TPS 10] "It makes me bearish on Anthropic because they have no way to compete" so you're bullish CBRS bc of Jane street -- which happens to be one of Anthropic's largest customers? wat meanwhile - Anthropic shipped fast mode and has cut the premium on it three times since February -- this does not suggest accelerating price curve for fast compute - Anth 3.5 GW TPU capacity from 2027 is 5x the CBRS deal - would you rather use any CBRS model right now vs Opus 5.5? the quality of ant's new model is ridiculous -- could care less if i have flash speeds on slop 11] "The moment Chinese labs release something genuinely competitive... the value of its inference infrastructure goes up massively" Open weights run on anything. hence word 'open'. On GPT-OSS 120B, Cerebras charges $0.35/$0.75 vs Groq's $0.15/$0.60. Open weights don't create lock-in; they expose Cerebras to a direct price comparison it currently is getting shrekt on so in summary: obviously not going to argue CBRS is fast for single suer decodes, speed tiers sell at a premium. but it's not a monopoly. it's in a competitive storm -- OpenAI sets price, holds termination right, $1b loan, 4.2% stake, cheapest equity in structure and is building an internal competitor to CBRS while also fine tuning a chip development LLM -- CBRS entire valuation is predicated on 2028+ billing curve when pressure is ramping now "scheduled unlocks". come on you are in crypto. the founders decided in IPO week, at ~$185–200, that they wanted $200M out and haven't bought a single share. do you think they will? when? VC cost base is $15-89, do you think they're going to hold? I don't. I think they're going to sell maybe you're right about revenue coming in 13x the street in FY2028. above my paygrade. my base case is you'll get a very good dip to buy.

Insane risk reward long Oracle short Apple imo. ROI on OpenAI and Meta AI products literally never better. Old GPUs ripping. People talking about Oracle like it’s going to default Meanwhile Zuck is Jobs 2.0 while Apple trades richest it ever has into multiple platform shifts

the importance of NEAR is that it was very heavily venture backed and absorbed hundreds of millions of dollars of unlock sales completely. the VCs are out now. this helps resolve the debate for institutional allocators, "Can anything other than BTC work"