Bottleneck Investor (@BottleneckFindr) header
Bottleneck Investor (@BottleneckFindr) profile picture

No BS. No regurgitating other users AI slop. Long on asymmetric bottlenecks. Swing trading overreactions. DM me ur thesis! Nothing I post is financial advice!

Nasdaqistan
Joined June 2023
Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)Photo by Bottleneck Investor (@BottleneckFindr)
$NBIS The charts I prefer to look at today. They just had to expand the y-axis again as UK south 1 set a new record for b300 spot prices.
1
1
23
1,168
Bottleneck Investor retweeted
At these ridiculous prices, your $NBIS thesis should focus more than ever on Token Factory + managed inference not just GPU capacity. $NBIS Token Factory is booming volume and inference performance metrics wise across a wide variety of models on Open Router. Snapshot: 851B tokens served. DeepSeek V4.1 Flash 1017: 717B DeepSeek V4 Flash 0731: 27.9B Kimi K3.1 Pro: 25.3B Qwen3 Embedding 8B: ~12.1B Qwen3.8 27B: ~11.9B Nemotron 3.5 Super VL: 10.6B MiniMax M3.1 Flash: 9.11B GLM 5.3: 6.96B That’s a genuinely diverse inference business: heavily used DeepSeeks (The most popular and most commonly used Open Weight LLM in the market & on OpenRouter right now), the most powerful frontier open weight Kimi K 3.1 Pro, embedding models, vision language inference with the above Nemotron and even smaller 27B dense models. DeepSeek dominates the volume, but the workloads extend well beyond one hot model building a inference business that is top tier across a diverse set of models. Forget the concentrated Deepseek Jump, from my tracking, serving volumes have stepped up sharply recently, with even Frontier Kimi volumes appearing to grow 3-6× within the last few days too. Some including @jiahanjimliu still attribute the total token volume jump to the DeepSeek price cut which I've pointed out is driven by significantly more core serving performance metric improvements rather than the discount which obviously helped yet the prices with the discount are above 65% + of the inferenced provider on OR for the said model. More importantly, OpenRouter changed its default routing on October 28. It now weighs effective pricing, throughput, time to first token, reliability, rate limits and response usability while loweing the focus on raw pricing. Faster providers can charge more and still win traffic. Price matters, price alone no longer explains who wins. My hypothesis on the sudden Token Factory improvement: Eigen’s optimizations, Clarifai’s team/licensed IP and Inferize’s faster model loading has only been recently put together to work for $NBIS Token factory public endpoints. We’ve noticed more momentum there over the last couple of weeks & days with sustained uptime and performance and to me this looks like the 3 acquisitions finally being fully integrated to Token Factory. Moving on, I recently met a Baseten Forward Deployed Engineering Manager. I asked the guy where the biggest bottleneck in running an inference platform was in a software sense. He straight up said GPUs. Was more of a joke but that is the reality for most of these platform players. When revenue scales, capacity has to scale with it parallelly and they are now facing significant capacity constraints in doing so. With $NBIS owning compute and significantly improving the serving stack now while closing up the performance gap between the top tier inference players now opens up a unique opportunity: more control over placement, utilization and token economics. OpenRouter represents only 1-3% of a provider’s total serving volume as we've seen from most of these platform players. That ratio isn’t publicly established for $NBIS, but you will get some crazy numbers if you multiply the current Openrouter volumes/revenues of $NBIS TF by that broader tokens to Open Router multiple. The thesis: owned compute → optimized inference → a larger share of the value per token with easier scalability. I will attach my last estimate of total Token Factory ARR (172 Million) estimate from September which turned out to be right near their disclosed September ARR with $NBIS last earnings stating it at near multi hundred million ARR by September 2026 thus likely below $200 Million. Once we have atleast a months worth of stable OpenRouter Token Volumes for $NBIS I will do a re-estimate as to when we would likely hit 1 Billion ARR on Token Factory alone. Stay Tuned. Long $NBIS here with this being the strongest component of my current $NBIS Thesis. I added significantly today. NFA
The story of $NBIS Token Factory suddenly pacing towards ~700 Billion tokens today (Most of that from Deepseek V4.1 Flash) is not just a price discount. It is the serving performance holding up through the volume explosion even with the discount holding token pricing significantly above most of the inference players on OpenRouter. This to be honest is the first time I was genuinely impressed at $NBIS Token Factory as a Developer and a Shareholder. After the 40% cut for the model, $NBIS pricing for DeepSeek V4.1 Flash 1017 is still above 66% of the inference providers on Open Router. So for people who aren't familiar with Token Prices this is not a "Crazy Discount" At 0.18 input and $0.72 output per million tokens, this is not at all traffic chasing the lowest price. The impressive part is availability maintained at a 100% throughout the Token Explosion, with remarkably high throughput and low latency on OpenRouter’s most popular model and arguably the most popular open weight in the market right now. $NBIS is delivering ~250 tokens/sec at P50. Above every regular offering slightly above even Morph's Fast (Premium) Tier at 246 and behind only Baseten Fast at 475 which are Premium Offerings with significantly high Token Prices. Both Fast tiers charge $0.60 input and $2.40 output per million tokens. Over 3x the $NBIS pricing. At the same time, $NBIS is maintaining ~0.6s latency, roughly a 93% cache hit rate and a displayed 100.0% uptime on DeepSeek. Maintaining that combination while absorbing this much traffic is an achivement most will fail to comprehent. The volume charts I'm tracking put $NBIS Token Factory on pace for ~700 billion tokens today. Most of the incremental volume looks like DeepSeek token share shifting over from other providers. They took on this sudden surge while maintaining uptime and throughput all in a matter of hours which is more than worth my glaze here. And the growth is not confined to the discounted model. I'm seeing nearly doubled token volumes and roughly doubled serving capacity on other popular open weights without a single $NBIS discount. MiniMax M3.1 Flash, Kimi K3 and Kimi K3.1 Pro, which I'd currently call the most powerful open weight LLMs in the market, all at regular market pricing (No discounts), with strong throughput across the broader lineup which is also suddenly multiplying the traffic $NBIS was receiving to these models prior to this week. As someone who works in inference, this is the first time I've been genuinely impressed by $NBIS Token Factory and if anyone ever wants to point to an inflection point for Token Factory I believe it will be the last 2 weeks. A discount can attract traffic. Scaling capacity and serving that traffic reliably while staying at the top of the chain for quality (Token Throughput in this case) is what makes this worth paying attention to.
7
4
1
56
5,575
$QQQ today feels like Jason's stash after Chapter 1
365
$NBIS New record for UK south. $13.70
1
1
25
1,047
Happy unofficial, international men's day everyone! No stock market in the game unfortunately. Making money is definitely a grind.
1
445
HOLY SHIT $NBIS! That record from yesterday may be beaten 5-fold today. Yesterday we breached 140B tokens for the first time. Someone already spent 228B in Deepseek tokens alone today and the day just started. On pace for 600B+ tokens today! Unreal.
New record for $NBIS on OpenRouter. On pace for 100B+ tokens today.
16
26
4
375
50,853
$NBIS $NVDA $IREN $CRWV Major news that you probably missed! Nebius gets more out of 64 Vera Rubin GPUs than others get out of 72. Vera Rubin is the flagship of the entire AI trade. No wonder that Nvidia 💚Nebius. In MLPerf training on gpt-oss 20B, Nebius's 64 GPUs beat every single-rack Rubin submission, including full 72-GPU racks. The other 8? Kept as hot spares for customers. Source: @nebiusai, Nov 16, 2026 nebius.com/blog/posts/mlperf…
2
9
1
76
3,959
$NVDA $IREN Now what is collettes definition of a quality Neocloud? Let me take a guess... Maybe the one that can get the most out of their vera rubins? Because that's objectively Nebius.
Our MLPerf® Training v6.1 results: 1st on @NVIDIA Vera Rubin at rack scale, in both pre-training and post-training. NVIDIA Vera Rubin NVL72 on Nebius trained up to 1.63x faster than NVIDIA Grace Blackwell Ultra NVL72 at matched scale. Plus updated Inference v6.1 results 👇 nebius.com/blog/posts/mlperf…
1
2
14
1,101
$NBIS $NVDA Colette just said business a usual for Neocloud financing. She literally said nothing's changing.
11
814
Colette quite validating for $NBIS and $IREN "our neoclouds have been so instrumental for supporting our business, and we continue to be working with them" She loves the n-word
1
4
65
4,847
$NBIS $NVDA NVIDIA named 5 clouds already running Vera Rubin racks: CoreWeave. Google Cloud. Microsoft Azure. Oracle. Nebius. Nebius is also on NVIDIA's list of early Vera CPU builders, next to Meta and OpenAI. On its own Rubin setup, Nebius posted the top per-GPU MLPerf inference result on DeepSeek R1. NVIDIA holds a 9.3% stake in Nebius.
3
40
2,835
$NBIS $IREN $NVDA NVIDIA's non-hyperscaler sales nearly tripled in a year. Data center revenue, Q3 vs a year ago: → Hyperscalers: $35.1B → $60.1B (+71%) → AI clouds, industrial & enterprise: $16.1B → $47.5B (+195%) That bucket is now 44% of NVIDIA's data center sales, up from 31%. It's the one Nebius and IREN sit in.
7
1,069
Bottleneck Investor retweeted
$NBIS I’ve been enjoying writing on Substack a lot, as it feels more aligned with what you need as someone who writes articles, but I know that many people don’t like that platform. Given that Nebius is by far my most commented stock, and a lot of people care about my thoughts on this stock here. Here is my Substack article, so everyone can read it directly here! If you want to subscribe/follow my Substack, you can do it here: funmentalist.substack.com
1
3
15
1,753
New record for $NBIS on OpenRouter. On pace for 100B+ tokens today.
6
3
82
54,656
See you on the earnings call tonight! jk. Will take a bit, if it happens.
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
1
356
Bottleneck Investor retweeted
The detail most people missed in NBIS MLPerf results isn’t just that Vera Rubin trained 1.63x faster than Blackwell Ultra it’s how Nebius configured the rack. Nebius ran on 64 GPUs (keeping 2 nodes as live hot spares just like a real customer cluster) and still beat every 72-GPU Vera Rubin submission in both pre-training and post-training. You don't beat full racks with 11% fewer GPUs if you're just a commodity hardware landlord. That comes from cluster software and orchestration and cutting training times by 33–39% means Nebius can push far more billable work through every megawatt it owns. The stock is trading like compute is a low-margin commodity, while $NBIS is quietly proving it has a real software and execution edge.
Our MLPerf® Training v6.1 results: 1st on @NVIDIA Vera Rubin at rack scale, in both pre-training and post-training. NVIDIA Vera Rubin NVL72 on Nebius trained up to 1.63x faster than NVIDIA Grace Blackwell Ultra NVL72 at matched scale. Plus updated Inference v6.1 results 👇 nebius.com/blog/posts/mlperf…
1
5
50
3,608
Bottleneck Investor retweeted
$NBIS Direct message to management @nebiusai @romanchernin @wienglischmann @marcboroditsky @demian_ai The last thing I am going to post here about the Q3 earnings is my direct recommendations to management, as a long-term believer in the company. 1) Reinstate the 2027 ARR guidance I understand that it will get progressively more difficult to provide a consistent estimate that won’t be volatile throughout the year, but there are two ways to deal with this. Either be conservative, or (in my opinion) the better solution is to split the business into segments and provide an ARR floor for the AI cloud segment. 2) Reinstate the active/connected capacity reporting I don’t understand the exclusion of these two metrics lately. I don’t care if the number might not be what everyone expects, and it might disappoint some people. Transparency is the #1 priority when it comes to investing, and there is no point in hiding how many MW are currently connected and how many are generating revenue. 3) Change the call organization I don’t want anyone in management to change their personality because that is going to do more harm than good. What I want from the management is the ability to react to live questions and allow investors to ask what actually comes to their mind, not scripted questions. It also allows investors to ask follow-ups, which are often more useful than the original question. 4) Token Factory/Asset-light model I believe these two segments are reaching a point where they deserve more than a comment or two in the report on their performance. Token Factory revenue/ARR could very well be reported along with its margins and some outlook. The same goes for the asset-light model, because it would help investors understand what actually happened during the quarter and how to model revenue estimates. Overall, I am asking for maximum transparency. The stock is facing one of the worst drawdowns in the history of the company and shareholders deserve to know as much as possible. Before anyone starts commenting, I think that Nebius has an incredible management team and I believe in their ability to execute even in this difficult environment. I just believe that adressing these 4 points directly would improve the overall sentiment around the stock and mainly its earnings reports. Feel free to add your comments Nebius bulls: @Sandeman52 @babyfolio @daniel_koss @SMASIMHO @AndersInvests @MktMavPro @ai_canaries @FunOfInvesting @TylerNorthwise @ren_stocks @itsalasdairmann @mvcinvesting @BKad2005 @BlackPantherCap @borex_investing @Futurenvesting @EndicottInvests
19
12
197
11,679
Trump says he "wouldn't mind" Jensen Huang being reinstated as $NVDA CEO.
1
7
758
This. Simple, irrefutable math.
Good heavens. Doomers are fleeing what they believe is an AI bubble, only to build a magnificent safe-haven bubble of their own. The gentlemen are now paying 50x earnings for businesses growing at 10%. And if they can pump one of these to 100x earnings by Christmas, some of these value investors might just become immortal.
3
585