Pinned Tweet
Hey everyone 👋
I’m Ash. I’m passionate about investing, personal finance and building wealth. I started this page to document my journey and connect with smart people who enjoy discussing markets, business and investing. Looking forward to learning from you all.
31
58
4,055
Ash retweeted
$NBIS Token Factory has surpassed Baseten and Fireworks on Openrouter today.
TogetherAI is still the top dawg.
Tokens on pace today:
TogetherAI - 1.46T
Token factory - 696B
Fireworks - 452B
Baseten - 240B
Now its obviously worth mentioning that Token factory only accomplished this today because of the 40% discount on Deepseek V4.1 Flash 1017 as of those 675B tokens, around 537B is that model.
But I wanted to call this out as I'm curious how this trends over time.
We know Token Factory is approaching a "Multi hundred million run rate".
Lets see when they pass 1B in ARR
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.
The writing on this has been on the wall for a while and the OpenRouter charts don't lie. Look at the absolute explosion in volume on the $NBIS Token Factory surging to a 139B daily pace with DeepSeek V4.1 Flash alone leading the charge at 67.2B tokens today. Why? Because they are undercutting/ discounting token pricing for the said model (By 60%) while maintaining high throughput, availability and serving quality simply because they can.
Here’s the structural shift happening underneath: End compute which is the actively running chips plugged in within a datacenter is now clearly becoming the tightest of the bottlenecks due to recent capital and Land Power Shell related challenges in the market.
Pure platform players like @modal , @FireworksAI_HQ , @baseten followed by a million more now are are forced to rent compute at higher and higher costs from infrastructure hosts such as the Hyperscalers or the Neoclouds because of the said tightening at the end compute layer with compute owners like $CRWV s, $IREN s and $NBIS s including the Hyperscalers with higher and higher rental prices squeezing their margins straight to the bone.
Meanwhile, vertically integrated players who actually own the physical compute stack ($NBIS & $CRWV) hold the ultimate economic advantage. They have the freedom to slash token prices down to raw compute cost and still make strong profits whereas renters cannot compete without going underwater. In other words tokens seems to be getting commoditized not GPUs.
When a player like $NBIS aggressively undercuts the market pure rental platforms get starved out of the token market entirely. The future belongs to the full stack owners who can win on token distribution while printing margin at the foundation layer. Full Stack ownership wins. $NBIS
Ash retweeted
$NBIS Token factory growth is exploding. Why?
youtu.be/zSdOk3AIUN8?si=5Uot…
Ash retweeted
$NBIS What's behind Nebius's record days on OpenRouter this week:
DeepSeek V4.1 Flash 1017 is the #1 model on the site by a mile. 61.9T tokens in the latest weekly ranking, up 15% and 2.7x the #2 model (my math).
Nebius is selling it 40% below its list price, and it has 5.8% of that model's OpenRouter traffic. That puts it among the six biggest of 40+ providers.
Of those six, Nebius is the fastest at 232 tokens/s, ~1.8x DeepSeek's own API (my math), and it's had 100% uptime over the past 3 days.
What it isn't:
• Cheap. Even after the discount it's $0.18 per million input tokens. DeepSeek charges $0.15, DeepInfra ~$0.10 and Wafer $0.036
• Big money, at least not yet. After cache discounts, OpenRouter shows Nebius's effective input price on this model at ~$0.016 per million. 500B tokens a day at that rate is ~$3M a year (my math). Output tokens pay more, but it's small next to the GPU business
Why it still matters: Token Factory runs on Nebius's own GPUs. It's a second way to sell the same capacity, and it shows inference demand in real time.
Nvidia said on Tuesday's call that token prices are down nearly 60% since May and that's pulling in more usage. Nebius's OpenRouter numbers this week are a live example.
Ash retweeted
$NBIS Nebius is serving more than 4x as much of OpenRouter's #1 model as CoreWeave.
That model is DeepSeek V4.1 Flash 1017: 61.9T tokens last week, up 15%, and 2.7x the #2 model (my math).
Share of its OpenRouter traffic over the past day:
• Nebius: 5.8%, at $0.18 per million input tokens after a 40% discount
• CoreWeave: 1.3%, at the full $0.30
Speed is a wash. Nebius, CoreWeave and Together all run it at roughly 215-240 tokens/s, well ahead of DeepSeek's own API at ~127. Nebius is one of the six biggest of 40+ providers and has had 100% uptime over the past 3 days.
The honest part:
• It's not the cheap option. DeepSeek's own API is $0.15, DeepInfra ~$0.10 and Wafer $0.036
• It's small money for now. After cache discounts, OpenRouter shows Nebius collecting ~$0.016 per million input tokens on this model. Today's ~675B pace at that rate works out to ~$4M a year (my math). Output tokens pay more, but next to the GPU business it's tiny
So why care: Token Factory runs on Nebius's own GPUs. It's a second way to sell the same capacity, and it shows inference demand as it happens.
Kress said on Tuesday's call that token prices are down nearly 60% since May and that's pulling in more usage. Nebius's OpenRouter chart this week shows exactly that.
Ash retweeted
1B+ LLM requests a day. One data foundation.
@OpenRouter uses ClickHouse Cloud to turn request telemetry into usage charts, endpoint health, model rankings and routing decisions.
Build on ClickHouse Cloud: ordnl.link/kuMOqqO
Ash retweeted
$NBIS Nebius Token Factory has done more tokens on OpenRouter today than in all of September.
• September total: 464B (my math)
• Old daily record: 99B
• Tuesday: 141B
• Today: 478B so far, ~675B pace
421B of it is DeepSeek V4.1 Flash 1017, the #1 model.
Five startups, one per layer of the physical AI stack. The 2026 Nebius Physical AI Awards winners are Archetype AI, @Outsight_tech, @Luminary, @ChefRobotics, and @lancedb. Each receives $150,000 in Nebius AI Cloud compute credits. Here’s what they’re building: bit.ly/4t3210o
Ash retweeted
$NBIS: It all comes down to active power.
Full site-by-site breakdown of capacity, timelines, and conservative assumptions.
Ash retweeted
🚨 $NBIS BULL CASE JUST GOT STRONGER — AND THE NUMBERS ARE HARD TO IGNORE
Northland just reaffirmed its OUTPERFORM rating on Nebius, with a new $380 price target
Yes, the price target was lowered. But there's a much bigger story here that investors shouldn't overlook:
💰 ~$60 BILLION in reported funding capacity.
🔥 ~$56 BILLION in estimated capital expenditures through CY2027.
Think about that.
Nebius may have secured enough financial firepower to support one of the most aggressive AI infrastructure buildouts in the market.
And this isn't just about spending billions.
It's about building the infrastructure that could power the next wave of AI growth:
⚡ Massive GPU capacity
☁️ Scalable AI cloud infrastructure
🔥 High-performance AI inference
🚀 Next-generation computing at scale
THE BIGGER QUESTION:
What happens when all that infrastructure is deployed, operational, and generating revenue?
That's where the long-term opportunity could become significant.
The market may be focused on the cost of building today.
I'm watching the revenue potential of what Nebius is building for tomorrow.
I'm still BULLISH on $NBIS.
Are you accumulating, holding, or waiting for a better entry?
Follow for more AI infrastructure and high-growth stock opportunities. 🚀
Nebius puts AI agents to work on physical AI pipelines.
@evanhelda, Head of Physical AI, describes how bringing infrastructure, tools, and shared context into one framework that agents can operate in helps teams reduce the engineering work of connecting each stage.
See the recap on Nebius LIVE: youtube.com/watch?v=r0nxG1Zk…
Ash retweeted
大多数中文区投资者可能没意识到:
$NBIS 有点像AI的 @ycombinator !
昨天研究了NBIS的客户结构,几乎所有AI初创公司都会找NBIS寻求算力资源。
NBIS正在构建谷歌那样的AI生态系统,Agent AI,Enterprise AI,Physical AI, 医疗AI,科学发现AI。
我依然相信 Nebius 有很大潜力成为一个千亿级别、万亿级别市值的公司。
Ash retweeted
The first number we’re watching at Nebius’s planned 1.2 GW Pennsylvania campus is 260 MW. That’s how much power $NBIS is targeting for October 2027 at Highridge.
The map shows where the first two phases sit along I-81 and where Nebius wants to expand. Butler Township expanded the area zoned for data centers, and Pennsylvania approved the transfer of two existing environmental permits. A stormwater amendment is still under review.
If power arrives in October, how many halls can Nebius finish and hand over to customers before year-end? That’s what will determine how much Highridge can contribute in 2027. We traced the power schedule and the approvals by phase in the updated report.
Ash retweeted
robinhood:0x9d9c6684f596f66a64c030b93a886d51fd4d7931 Northland PT $380. 📈
Northland on $NBIS: Reiterating Outperform rating & lowering PT to $380
Estimates Nebius has access to ~$60B of funds to cover estimated $56B of capex through end of CY27.
Thoughts? 🤔
Ash retweeted
$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…
Ash retweeted
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.
Ash retweeted
Kress basically says $NVDA still wants to be the bank of AI, and the company will continue to finance datacenter builds.
Ash retweeted
🚨 BULLISH $NVDA: THE COMPANY IS EXPANDING BEYOND SELLING GPUS. $NBIS $CRWV $IREN
NVIDIA disclosed $32B in additional conditional neocloud agreements involving upfront infrastructure sales and potential revenue sharing.
If deployed successfully, NVIDIA could participate in revenue generated by third-party cloud customers rather than relying exclusively on hardware purchases.
That's a potentially important extension of its business model.
🚨 $NVDA DOUBT 3: WHAT I'M WORRING ABOUT AND WHY ?
NVIDIA NOW HAS $118B OF MAXIMUM GROSS GUARANTEE EXPOSURE.
This includes:
- $105B supporting SB Energy's infrastructure associated with OpenAI.
- $9.5B in power and systems guarantees.
- $3.5B supporting neocloud land, power and shell development.
The exposure is conditional and phases in over time. NVIDIA says it hasn't made payments under these guarantees.
Still, there's a difference between selling GPUs to customers and supporting the financial obligations behind their data centers.
Ash retweeted
$NBIS
GPU pricing is insane
$NVDA says GPU rental prices are rising across generations.
Hopper GPUs have recontracted at up to 2x their original price under multi-year agreements, and B300 rates have risen more than 60% since their introduction.
Token prices have fallen nearly 60% since May, accelerating usage and demand for compute. Nvidia compute "remains fully utilized across every hyperscaler and neocloud we serve," she said.
Ash retweeted
New record for $NBIS on OpenRouter. On pace for 100B+ tokens today.






































