Respo 🔺 (@Blockchain_Hulk) header
Respo 🔺 (@Blockchain_Hulk) profile picture
For now, Planet Earth
Joined July 2020
JUST IN: Goldman Sachs reiterated Coreweave $CRWV buy rating & $106 price target after meetings with CEO, CFO & VP of IR
38
53
8
918
96,742
Respo 🔺 retweeted
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen
61,331
121,778
24,521
1,175,176
59,336,640
Respo 🔺 retweeted
Black Monday Black Tuesday Black Wednesday Black Thursday Black Friday Black Saturday Black Sunday
19
9
187
5,525
Respo 🔺 retweeted
I want to introduce $IREN at just 0.5x net asset value. Arguably, the best asymmetric upside opportunity in the entire AI sector.
$IREN As I see it, net asset value today is ~$33/share. That's assigning nothing to the operating business. I've seen estimates from others (who shall remain nameless) ranging from $28 to $40. Even with a pretty penalizing 25% liquidation haircut, we're trading below liquidation value. And that's assigning $0 to their 35 GW pipeline.
5
19
1
180
6,935
At @Cerebras, we’re aiming to double AI speed every year. To achieve this, we build on key innovations including Nexus - our new rack-scale architecture. The CS-4 is the first system built on it. The CS-5 and the CS-6 will use it too. Delivering fast AI starts with fast compute tightly coupled to SRAM. Nothing else gets you there. But you also need I/O and power. In past systems, those were treated as an intertwined collection of parts. Improve one and you had to re-engineer the whole system. Nexus is different. Power, compute, and I/O are decoupled. Each is a self-contained module that can be developed independently. Power sits at the front of the rack as modular power supplies. When the next generation of supplies comes out, plug them in. Compute sits at the back as three pluggable backpacks, each housing one Wafer-Scale Engine. The next generation of compute comes out, upgrades are dead simple. I/O is its own silicon, connecting at the edge of the wafer. Because each part is built as a separate module, an advance in one no longer requires a redesign of the others. Faster I/O can be plugged in as the industry advances. Innovation in power, compute, or I/O can reach customers as soon as they are ready. Nexus was built for scaling AI over generations and as the industry rapidly evolves. It will support multiple generations of the fastest inference in the industry. All on the same architecture.
19
37
9
315
14,911
Respo 🔺 retweeted
In all this Nvidia noise, the Cerebras quarterly earnings got ignored. Their core revenue is increasing 70% YoY to $231M and margins are getting better too (core gross margin to 41%). Backlog is at massive $26.1B.
12
8
2
105
10,459
Respo 🔺 retweeted
JUST IN: HUANG DISCUSSES RETURN AS NVIDIA CEO WITH ALLIES: FT
86
112
57
2,245
395,767
NVIDIA board credentials. No wonder Jensen got fired.
1,363
2,423
486
29,852
6,938,662
Respo 🔺 retweeted
$CBRS 👀 name dropping Muse… sure looks like $META will be next hyperscaler. Relationship goes back to Llama.
“What’s blocking the launch?” Every team asks a question like this. The answer is usually spread across Slack threads and Linear tickets. So we gave it to an agent running Meta Muse Spark on @cerebras. It searched both and came back with the blockers. The whole thing took 36 seconds. And it ran at up to 3,000 tokens per second. This is what agents can do at real-time speed.
1
3
72
6,264
Respo 🔺 retweeted
$CBRS | Wedbush 𝗿𝗲𝗶𝘁𝗲𝗿𝗮𝘁𝗲𝘀 𝗢𝘂𝘁𝗽𝗲𝗿𝗳𝗼𝗿𝗺 on 𝗖𝗲𝗿𝗲𝗯𝗿𝗮𝘀 𝗦𝘆𝘀𝘁𝗲𝗺𝘀, raises PT to $𝟮𝟬𝟱 from $𝟭𝟵𝟬 Analyst sees strong Q4 expectations and greater room for upside in 2027, supported by pricing, margins, and cloud capacity.
1
9
49
6,218
Respo 🔺 retweeted
There's no close competitor to Cerebras' inference speed in the industry. A very important detail as mentioned by @beniduboss in replies to my post below: - Kimi K3.1 (2.8T params, largest open model) on Moonshot’s own API: ~32 tok/s - on Cerebras: up to 2,000 tok/s. that is whopping 60x faster 🤯 - This means 10k-tokens agent step goes from ~5 min to ~5 sec. Cerebras is quietly building a monopoly on fast inference.
robinhood:0x5c90450bbb4273d7b2f17cf6917aeb237a569679 has been one of the best buy opportunities in the last month or so (been accumulating): 1. It's been trading at almost half the IPO price (big discount) 2. The *contracted* backlog itself is $25B compared to a $23B market cap. 3. Q4 guidance (revised upwards!) is already implying ~40% QoQ growth. 4. Signed a deal with Microsoft and another (unnamed) hyper-scaler in October. 5. The chip breakdown within 18 months claims by the short seller firm (Culper Research) were already rebutted. 6. They might benefit if there are short to medium term turmoil at Nvidia. I really don't get the hate and negative sentiment towards them. (Ultra) fast inference is here to stay and the stock is available at over 50% off. I am bullish.
2
4
14
2,085
Respo 🔺 retweeted
$CBRS 3 Hyperscaler Wins, a 3x Revenue Outlook for 2027, and a Bull Thesis Getting Stronger (earnings call rundown) Today’s earnings call delivered substantial support for the arguments I’ve been making about Cerebras: demand for fast inference, OpenAI’s capacity constraints, large-model support, hybrid inference, the hardware roadmap, and the economics of owning the underlying technology. Microsoft $MSFT joins the customer list, production disaggregation is live, margins are improving, and management continues to expect revenue to triple in 2027. There is a lot here for Culper, SemiAnalysis, and the broader FUD circuit to digest. I wonder has anyone else on X been saying this same thing? @andrewdfeldman 😉Andrew Feldman, on today’s earnings call: -“There are also a set of persistent falsehoods proselytized to the financial community, namely that our systems can't support large models, obviously false, as they currently do, and that our manufacturing yields are low. Both are false.” Here are the developments that stood out across the call, earnings release, and 10-Q. Financial measures labeled “core” use Cerebras’ non-GAAP definitions. RECORD REVENUE AND IMPROVING PROFITABILITY - Record Q3 core revenue of approximately $231M, up 70% year over year and 10% sequentially, beating the guidance midpoint by approximately $16M. - Record core cloud and other services revenue of $145M, up 272% year over year and 13% sequentially. Core hardware revenue reached $86M, up 5% sequentially. - The release also reported record GAAP revenue of $210.8M, up 55% year over year, with GAAP cloud and other services revenue growing 265%. - Core gross margin reached 41%, beating guidance by 2 percentage points. Management says the margin trough occurred in Q2, 1 quarter earlier than expected. - Core cloud gross margin improved to 46.9%, up 5.1 percentage points sequentially. CFO Bob Komin explained: “The primary driver of the increase was new business delivered at higher pricing and margin.” - Core operating margin improved to -14.3%, beating the guidance midpoint by nearly 10 percentage points. Core operating loss held approximately flat at $33M as revenue increased. - Core net income turned positive at approximately $30M, and year-to-date adjusted core EBITDA reached a positive $2M. - Bob Komin, CFO: “We also delivered positive non-GAAP net income for the first time as a public company and ahead of schedule.” GUIDANCE RAISED, WITH A SUBSTANTIAL Q4 STEP UP - Q4 core revenue guidance increased to $310M–$330M. The midpoint implies approximately 39% sequential growth. Management expects approximately 85% year-over-year growth. - Q4 core gross margin guidance increased to 47%–49%. The 48% midpoint represents a 7-percentage-point improvement from Q3. - Q4 core operating margin guidance is -1% to +1%, putting the business around core operating breakeven. - Full-year core revenue guidance increased to $940M–$960M, with 42%–44% core gross margin and -10% to -8% core operating margin. - Management continues to expect revenue to triple in 2027, with expanding margins. Komin expects core gross margin to exceed 50% in the near term, with progress toward the longer-term 60% gross margin and 40% operating margin targets. - Bob Komin: “In closing, Q3 was another record quarter for Cerebras, and Q4 will be too.” MICROSOFT, AWS, ORACLE, AND A MUCH LARGER OPPORTUNITY PIPELINE - Microsoft $MSFT signed a cloud deal with Cerebras today. CEO Andrew Feldman expects revenue from that agreement to begin in H1 2027, with further details disclosed during that period. - That brings Cerebras to 3 hyperscaler wins in 6 months: Amazon’s AWS $AMZN, Oracle $ORCL, and Microsoft $MSFT. This strongly supports my argument that sophisticated customers see meaningful commercial value in Cerebras’ architecture. - Oracle $ORCL is purchasing Cerebras hardware for its U.S. data centers. The Cerebras-powered OCI inference service is expected to launch in summer 2027. - Andrew Feldman, on Oracle: “It is the start of a large partnership in which our equipment will be deployed in Oracle data centers and part of the OCI cloud.” - New hardware orders begin shipping in Q4 and continue through at least H1 2027. Komin expects hardware to represent closer to 50% of revenue in Q4 and Q1. - Cerebras is expanding distribution through neocloud relationships, including Fireworks. Its cloud customer list includes Cognition, Figma $FIG, Block $XYZ, and Bilt. - Bob Komin: “Over the last few months, our pipeline has grown by much more than 10x, driven by a diverse set of opportunities from both new prospects and existing customers, including hyperscalers, neoclouds, sovereigns, enterprises, and AI natives.” That is a rapidly expanding opportunity pipeline across several customer categories. OPENAI: DIRECT SUPPORT FOR THE CAPACITY ARGUMENT I HAVE BEEN MAKING - OpenAI business grew sequentially in Q3. The 10-Q reports $72.4M of Q3 revenue under the OpenAI arrangement. - Andrew Feldman: “OpenAI grew in Q3 as compared to Q2, and we expect more growth in the future.” - Feldman described the Ultrafast GPT-5.6 Sol preview as an overwhelming success and explicitly attributed the limited broader rollout, including Astra, to insufficient Cerebras capacity. - Andrew Feldman: “Our CS-3, CS-4, and CS-5 are more than capable of supporting Astra-scale models and larger models as well.” - Asked about the cause of delivery delays, Feldman answered: “Our delays right now are exclusively land, power, shell.” He added: “We have capacity. We have all materials. We are waiting on the completion of land, power, shell.” - Feldman confirmed that Cerebras had received no cancellations for delayed capacity. OpenAI also has the contractual option to purchase hardware. - RPO increased to $26.1B, already incorporating management’s estimated delay credits. It excludes the Oracle deal signed after September 30, as well as other subsequent agreements. RPO includes pass-through revenue. - The initial OpenAI commitment remains 750 MW. Its additional 1.25 GW option remains unexercised and outside committed RPO. - Management expects to regain the delivery schedule in H2 2027. Capacity potentially eligible for cancellation during 2026 represents less than 5% of committed megawatts, according to Komin. -“Given the importance of this [OpenAI] partnership to our business, we understand the close scrutiny it gets from investors. -“The details of the agreement or news about it are unlikely to come from any source other than Cerebras or OpenAI. For example, there were rumors that Cerebras couldn't support Astra for technical reasons. These are false. Our CS-3, CS-4, and CS-5 are more than capable of supporting Astra-scale models and larger models as well. -“At the same time, I expect, and you should expect, OpenAI to build capacity using a range of hardware solutions. OpenAI will retire older models, deploy new ones across those compute options for internal and external use. Their compute needs are vast, with plenty of opportunities for different solutions to play a role as they scale.” LARGE MODELS AND THE WAFER ARCHITECTURE - Cerebras was a day-zero launch partner for Kimi K3.1, a 2.8-trillion-parameter model, running at up to 2,000 tokens per second, according to management. - Feldman said new-model bring-up time has fallen to just a few days, alongside improving software maturity and cloud uptime. - The combination of existing frontier-model deployments, day-zero support, and the explicit Astra statement strengthens my argument that Cerebras can scale to increasingly capable models. - Feldman directly rejected claims of low manufacturing yields. His explanation covered both fab yield and parametric yield, including spare tiles, routing around defects, broader operating tolerances, and independently adjustable voltage across 84 wafer sections. - Andrew Feldman: “The result is exceptionally high yield at both the fab and the factory.” - Feldman attributed recent hardware-margin pressure to the manufacturing ramp and legacy customer pricing. New hardware orders carry higher prices, with management expecting improved margins beginning in Q4 and a longer-term 50% hardware gross-margin target. CULPER’S WAFER-LIFE NARRATIVE HAS SPECIFIC COMPANY DATA TO ANSWER - In his November 5 response, Feldman reported average CS-3 wafer junction temperatures below 65°C, a maximum of 85°C, and qualification by TSMC $TSM for 10 years of continuous operation at 105°C. He also reported that fewer than 0.5% of deployed CS-3 wafers had been decommissioned since the generation’s 2024 introduction. Those are company disclosures directly relevant to the assumptions behind Culper’s approximately 2-year lifespan claim. - Today’s 10-Q adds another concrete datapoint: warranty utilization for the first 9 months fell approximately 50% year over year, from $16.3M to $8.2M. Additions to warranty reserves fell approximately 25%, from $14.3M to $10.8M. HYBRID INFERENCE IS NOW SERVING CUSTOMERS - This directly supports one of my central arguments about Cerebras’ role in heterogeneous infrastructure: AMD $AMD GPUs handle prefill, and Cerebras $CBRS handles decode. - Andrew Feldman: “Today, our cloud serves customers using AMD GPUs for pre-fill and Cerebras CS-3 systems for decode.” - The software platform supports AMD and other GPUs, with plans to deploy it across Cerebras’ cloud and at selected customer data centers. - Management says disaggregation can increase the tokens delivered by a data center by up to 5x, improving revenue potential per megawatt. - Andrew Feldman: “Disaggregated inference delivers substantially more high-speed tokens for the same power.” - Andrew Feldman: “We believe disagg will become one of the default offerings in major cloud deployments.” This is the specialized decode opportunity I have been talking about, now supported by production deployments. FAST TOKENS HAVE COMMERCIAL VALUE - Feldman said cloud customers report exceptional margins on fast tokens and describe Cerebras tokens as their most profitable. - Andrew Feldman: “That is why fast tokens command a premium, and our cloud customers tell us that fast tokens generate exceptional gross margins and that Cerebras fast tokens are the most profitable tokens they sell.” - On low-batch GPU competition, Feldman estimated that Cerebras becomes more economical around 250–300 tokens per second, with a larger advantage at higher speeds. - Andrew Feldman: “By the time you get to 500, 800, 1,000 tokens, we are vastly more economical.” - These are management’s competitive assessments, with economics dependent on the workload and configuration. The improving reported cloud margins add financial evidence to the commercial argument. - Bilt supports my personal-agent thesis. The earnings release describes a platform serving 8M consumers and 50,000 businesses. Faster interactions, useful assistants, and broader consumer distribution create a meaningful market for this technology. The people mocking the personal-assistant example have an increasingly substantial business opportunity to explain away. CS-4 IS SHIPPING, AND CS-5 IS AHEAD OF SCHEDULE - CS-4 shipments have begun, with full production approaching. Management describes approximately 2x performance and throughput, with the CS-3/CS-4 crossover expected in Q1 2027. - WSE-5 taped out on June 15, returned from the fab ahead of schedule, and is already running models in CS-5 system testing. - Andrew Feldman: “The chip is back, has been brought up, and is already running models and is undergoing in-chassis testing.” - CS-5 production is expected in early summer 2027. Feldman says it will more than double CS-4’s speed. - The earnings release reiterates a CS-5 design target of up to 10,000 tokens per second per user. - WSE-5 uses TSMC’s $TSM 5 nm process, which Feldman says offers lower wafer costs and less competition for wafer capacity. - Andrew Feldman: “We believe our roadmap will enable us to maintain that advantage for years to come.” The roadmap I have been emphasizing now has additional shipping, silicon, and system-testing milestones behind it. AI-ASSISTED DEVELOPMENT COULD FAVOR HARDWARE DIFFERENTIATION - This came up directly during the earnings-call Q&A. Barclays $BCS analyst Tom O’Malley asked about frontier labs developing hardware quickly, citing Jalapeño, and accelerating model deployment across different computing architectures. He asked how Cerebras benefits as infrastructure becomes more heterogeneous and models become easier to move between hardware platforms. - Andrew Feldman pointed to Anthropic’s reported ability to bring up AMD $AMD GPUs in approximately 3 days. He then explained why he considers this important for the industry. - Andrew Feldman: “This is a very interesting and important development. Particularly those who have a long-term focus, I don't think people have really grokked what this means to our ecosystem. Compilers and low-level software have historically been a fundamental impediment to new architectures. They have favored the incumbent. They were the software moat that you heard so much about.” - “If Anthropic can in fact do what they say they do, it means that the value of this low-level software shrinks and over time disappears. It means that the moats from that low-level software also shrink and disappear. If the low-level software has little value, the thing that is left to provide value is innovative hardware.” - “It means that sort of differentiation will be more dependent on hardware differentiation. I think this is a tremendous advantage to innovative hardware companies like Cerebras and others who build pioneering hardware and new and different architectures. I think this will remove an advantage that the incumbents have long had.” My read for $CBRS: as AI-assisted development makes adapting software to different architectures faster and cheaper, customers can more readily select hardware based on speed, throughput, and economics. That could reduce the switching barriers that have benefited incumbents such as NVIDIA $NVDA and increase the commercial value of Cerebras’ architectural advantages. This connects directly to the argument I have been making about increasingly capable AI accelerating development and expanding the opportunity for differentiated inference hardware. CAPACITY AND MANUFACTURING ARE SCALING - Approximately 700 gross MW of data center capacity is contracted for delivery by the end of 2027, with negotiations progressing for gigawatts of additional capacity. - Facilities in Canada, France, and Norway are expected to come online over the next several months. Additional capacity has been contracted in Norway and Finland, with a São Paulo facility in the final contracting stages. - More than 75% of the data center capacity coming online in 2027 is expected outside the U.S. - Manufacturing is scaling with Flex $FLEX and Sanmina $SANM. Management expects to exit 2026 with more than 10x the manufacturing capacity available at the start of the year, with additional factories planned for 2027. - Andrew Feldman: “TSMC has committed sufficient capacity of wafers to us for 2027.” - Cerebras’ own wafer processors avoid HBM, providing a supply-chain advantage in a memory market Feldman expects to remain constrained through 2028. The GPUs used in heterogeneous deployments have their own memory requirements. THE BALANCE SHEET AND BUSINESS MODEL - Cerebras ended Q3 with approximately $7.8B in cash, cash equivalents, restricted cash, and marketable securities, plus an unused $850M revolving credit facility. - The OpenAI working-capital loan and ongoing pass-through payments funded most of the company’s net working-capital requirements and capital expenditures so far in 2026. - The 10-Q shows approximately $175.5M of noncash repayments against the OpenAI loan, leaving approximately $829.1M of principal outstanding at September 30. - Customer upfront payments provide another source of funding. Designing and manufacturing its own systems allows Cerebras to deploy that equipment at its own manufacturing cost. - Andrew Feldman: “We are an AI hardware company with two channels to market.” - The customer base is broadening materially. The 10-Q shows MBZUAI and G42 together represented approximately 45% of quarterly revenue, down from 84% a year earlier. Revenue billed to U.S. customers increased approximately 229% year over year. - Komin confirmed that IPO lockup releases were fully complete as of November 9. A PARTICULARLY USEFUL REMINDER FOR SEMIANALYSIS - SemiAnalysis’s own CS-4 report recognized the benefits of faster inference, attractive customer economics, improved manufacturability, and disaggregation. Today’s call provides further evidence of deployment and commercialization in those areas. Its specific long-context and concurrency concerns can be evaluated against the expanding production record. - Dylan Patel, founder and CEO of SemiAnalysis, in Cerebras’ August CS-4 announcement: “CS-4 will scale ultrafast tokens for larger models and significant user volumes.” That is a particularly useful quote to keep handy. AWS $AMZN, Oracle $ORCL, Microsoft $MSFT, growing OpenAI revenue, production AMD $AMD disaggregation, improving core margins, and next-generation silicon already running models. My conviction in $CBRS has been built around the commercial value of this technology and the pace of its improvement. Today added a substantial amount of evidence. BULLISH AND LONG. We will see $200 again in no time.
Today, @cerebras announced financial results for Q3 2026, delivering record revenue, and raising full year outlook. Q3 2026 highlights: • Record Revenue: GAAP cloud revenue grew 265%; Core cloud revenue grew 272% • 700 MW of data center capacity now under contract • Commercial wins with 2 new hyperscalers • Day-zero launch partner for @Kimi_Moonshot K3.1, running at up to 2,000 tps • Raised Q4 guidance and full year outlook across the board
14
7
1
74
9,767
Respo 🔺 retweeted
Replying to @wallstengine
With Jensen gone and Groq stuck inside that quagmire, the coast is clear for Cerebras to ship into all the hyperscalers. Microsoft, Oracle & Amazon have signed, Meta, SpaceX and Google will follow.
1
2
14
868
So $NBIS and $IREN both miss on 2026 ARR.. while $CRWV beats and raises its ARR guidance.
12
7
122
9,480
Respo 🔺 retweeted
CoreWeave Database is now generally available. Managed PostgreSQL now runs within your CoreWeave Kubernetes Service (CKS) cluster, so agents and RAG pipelines no longer cross clouds to access their data. Built on CloudNativePG, with pgvector included. crwv.co/uuHsu
5
9
3
84
4,823
Respo 🔺 retweeted
There are plenty of very high quality names at very attractive prices now, so I don't fault anyone for bailing on $AAOI in favor of otherstocks, but to be fair, if you're judging it based on 2026 numbers, you're missing the forest for the trees. The next few years have explosive growth potential. If they get anywhere close to the numbers they think they can do, the shares have massive upside.
People are losing faith in $AAOI, and it’s not hard to see why. Management delays and a Q4 guidance miss have shaken confidence.
3
2
1
30
3,652
$AAOI IS ROCKETING!!! 🚀🎉 I still can’t believe this stock dipped into the $40s. The market handed investors an incredible opportunity. Demand is immense, and if management executes on its projected growth ramp, I believe $200+ by the end of 2027 is definitely achievable. At the lows, $AAOI traded below 10× projected FY2027 consensus earnings. Let that sink in. Stay convicted, $AAOI holders. 🫡
18
6
148
14,746
Respo 🔺 retweeted
$CRWV CEO cancelled his 10b5-1 stock sales. CoreWeave is up 20%!
34
23
10
427
22,477
CoreWeave $CRWV expects to turn adjusted net income positive in the first half of 2028, and for the full year 2028 It expects its adjusted net loss to "reduce meaningfully" in 2027, and EBITDA minus interest to more than double "Adjusted operating income at $379 million nearly tripled from the second quarter, while interest expense rose about a third. That is the path to earnings," the CFO said
6
22
3
185
27,127
Respo 🔺 retweeted
Wait, you're kidding me, right? 😂 For a year the bears wanted us to believe GPUs are worthless after 2 or 3 years. Now $CRWV says 300,000+ GPUs come off contract in 2027 and 2028, and pricing for those GPUs rose 50% quarter over quarter. Apparently many investors are surprised... ?? What did they think would happen? That CoreWeave throws paid-off GPUs in the trash? Lost for words 🤣
$CRWV $NBIS $IREN CoreWeave said more than 300,000 GPUs will come off contract in 2027 and 2028. Pricing for thos GPUs rose 50% quarter over quarter in Q3 and continues rising. Recontracting that existing capacity could generate billions in additional annualized revenue without any additional GPU capex. Bullish.
31
17
1
582
38,832