• Latest
  • Trending
  • All
Answer card stating that Bloomberg reported on 26 August 2026 that Anthropic agreed to pay Nscale roughly 45 billion dollars over six years for about 460 megawatts at the Monarch campus in Mason County, West Virginia, on NVIDIA Vera Rubin systems, with CNBC matching the report, Anthropic declining to comment, and the facility expected online at the end of 2027.

Anthropic’s Nscale deal buys 460 MW for late 2027

3 September 2026
Answer card stating that Ternary Bonsai 2 27B, released by PrismML on 17 September 2026 under Apache 2.0, packs Qwen3.8 27B into 5.95 gigabytes at 1.72 bits per weight, keeps 98.2 percent of the 14-benchmark average, about 75 percent on SWE-bench Verified and Terminal-Bench 2.1, and needs PrismML's llama.cpp fork to run.

Does Bonsai 2 27B really keep 98% of Qwen3.8 in 5.95 GB?

20 September 2026
Answer card stating that Jev 1.13 from TypeSafe AI is a decision model in early access since 15 September 2026 that returns typed probabilities instead of text, priced at 42 dollars per billion input tokens with output tokens free, answering in 70 to 500 milliseconds, with a 64K token request budget, text input only, and a documented list of things it does badly, including counting and dates.

Jev 1.13 bills $42 a billion tokens, and it can’t count

19 September 2026
Answer card stating that Qwen3.8-Omni-Flash launched on 17 September 2026 as an API only model on Alibaba Cloud Model Studio, taking text, images, audio and video in a 1M token context and returning text only, priced at 0.15 dollars per million input tokens for every modality and 0.47 dollars per million output tokens in the international regions, with no open weights published and the Qwen-Live Harness GitHub repository returning 404.

Qwen3.8-Omni-Flash bills audio at $0.15 and ships no weights

18 September 2026
Answer card stating that on 15 September 2026 AWS said it is unable to restore access to resources and data hosted exclusively in the Middle East Bahrain region me-south-1 and in the mec1-az2 zone of the UAE region, because the damage spanned multiple Availability Zones and exceeded what multi-AZ services are designed to withstand.

AWS can’t restore me-south-1, six months after the drone strikes

17 September 2026
Answer card stating that Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on 15 September 2026 at 3 dollars per million audio input tokens and 12 dollars out, that the thinking model requires asynchronous tools, and that Artificial Analysis scores it 82.6 on its Speech to Speech Quality Index.

Gemini 3.8 Live Extended Thinking rejects any tool that blocks

16 September 2026
Answer card summarising the Atria Dawn Preview release: 744B GLM-5.2 base, MIT licence, 1.5 TB BF16 and 756 GB FP8 checkpoints, 256K context, top on five of sixteen benchmark rows and trailing on SWE-bench Pro.

Atria Dawn Preview is 744B under MIT, and the BF16 weighs 1.5 TB

15 September 2026
Answer card stating that OpenAI released the Agents API in public beta on 10 September 2026 with no separate fee, billed through model tokens, tool calls and hosted sandbox time, with a choice of OpenAI hosted, self hosted or partner sandboxes, US only data residency and no Zero Data Retention support.

OpenAI’s Agents API has no fee, no ZDR and a one hour sandbox clock

14 September 2026
Answer card: Sakana Fugu Max at $2 and $6 per million tokens, Fugu Ultra v2 unchanged at $5 and $30, and Sakana saying Ultra v2 scores without Fable 5 or GPT-6 Astra in its pool.

Fugu Max costs $2 and $6 while Fugu Ultra v2 runs without Fable 5

13 September 2026
Answer card stating that DeepSeek released DeepSeek-V4.1-Flash on 10 September 2026 as a 552 billion parameter mixture of experts model with a new causal encoder decoder architecture that activates 8 billion parameters on input and 16 billion on output, with native vision, a one million token context and MIT licensed weights, that the API model name is now deepseek-flash at 0.15 dollars per million input tokens and 0.60 dollars per million output tokens off peak, and that DeepSeek announced V4 Pro would be routed to V4.1-Flash from 14 September and reversed that on 11 September.

DeepSeek V4.1-Flash arrived, and the V4 Pro retirement lasted a day

12 September 2026
Answer card stating that Cognition released SWE-2 on 10 September 2026, a coding model post-trained from Kimi K3, scoring 50.0 percent on FrontierCode 1.1 Main against 50.9 percent for Claude Fable 5.1 and 27.3 percent on Terminal-Bench 4 against 55.8 percent, available only inside Devin.

SWE-2 trails Fable 5.1 by one point, and by 28 on Terminal-Bench 4

11 September 2026
Answer card for Meta Muse, free to 100 million tokens a week then $20 a month, launched 8 September 2026 for United States adults only, running in a dedicated per user virtual machine.

Does Meta Muse do enough to earn your inbox and a card on file?

9 September 2026
Answer card stating that the public download pages for the VMware Virtual Disk Development Kit on developer.broadcom.com began returning 404 errors on 25 August 2026 with no announcement or deprecation notice, that Broadcom support tells customers the kit is no longer available for use or download, and that release lines 7.0.3.1, 8.x and 9.x are all affected.

Broadcom pulled VDDK 8.0 and 9.0, and the 404 is the only notice

8 September 2026
  • About
  • Contact
  • Privacy
  • Legal
Sunday, September 20, 2026
  • Login
Packet Nebula
  • Home
  • Articles
    • Security
    • Network
    • Dev
    • Sysadmin
    • SEO
    • Email & DNS
  • Tools
    • Network tools: free, fast, no signup
    • Security tools: free, fast, no signup
    • Developer tools: free, fast, no signup
    • Sysadmin tools: free, fast, no signup
    • SEO tools: free, fast, no signup
    • Email & DNS tools: free, fast, no signup
  • Download
  • About
No Result
View All Result
Packet Nebula
No Result
View All Result
Home Dev

Anthropic’s Nscale deal buys 460 MW for late 2027

by stephane
3 September 2026
in Dev
0
Answer card stating that Bloomberg reported on 26 August 2026 that Anthropic agreed to pay Nscale roughly 45 billion dollars over six years for about 460 megawatts at the Monarch campus in Mason County, West Virginia, on NVIDIA Vera Rubin systems, with CNBC matching the report, Anthropic declining to comment, and the facility expected online at the end of 2027.
493
SHARES
1.4k
VIEWS
Share on FacebookShare on Twitter

Two people who would not be named, a British infrastructure company founded in 2024, and a field in Mason County, West Virginia. That's the sourcing on the largest compute contract of the month. On 26 August Bloomberg reported that Anthropic agreed to pay Nscale roughly $45 billion over six years for about 460 megawatts at its Monarch campus, running NVIDIA Vera Rubin systems. CNBC matched it the same day. Anthropic declined to comment, Nscale has published nothing, and neither company has put a line on its newsroom. The number that matters if you're actually paying for Claude today isn't the 45. It's the date.

The short answer

Bloomberg reported on 26 August 2026 that Anthropic will pay Nscale about $45 billion over six years for roughly 460 megawatts at the Monarch campus in Mason County, West Virginia, on NVIDIA Vera Rubin systems. CNBC matched it. Neither company has confirmed anything. The site is due online at the end of 2027, so nothing about your rate limits, your prices or your regions changes because of this. What it tells you is where Anthropic thinks its capacity problem gets solved, and how late.

$45Breported, over six years
460 MWcontracted capacity
late 2027when the site opens
Answer card stating that Bloomberg reported on 26 August 2026 that Anthropic agreed to pay Nscale roughly 45 billion dollars over six years for about 460 megawatts at the Monarch campus in Mason County, West Virginia, on NVIDIA Vera Rubin systems, with CNBC matching the report the same day, Anthropic declining to comment, and the facility expected online at the end of 2027.
The deal in one card. The stat is the part people skip.

What was actually reported

Bloomberg went first on 26 August. Anthropic agreed to pay Nscale, a British AI infrastructure company founded in 2024, roughly $45 billion across six years for compute capacity. CNBC ran its own version the same day, with the detail that the capacity is about 460 megawatts at an Nscale development in West Virginia, drawing on Nvidia’s Vera Rubin chips, expected online at the end of 2027. TechCrunch and others followed.

Anthropic declined to comment. Nscale hasn’t issued a release. So every number in the paragraph above traces back to unnamed sources at two outlets that both have good records on this beat, which is a decent place to be but is not the same as terms in a filing. We’d flag that harder if the numbers were doing something surprising. They aren’t.

Worth saying plainly: $45 billion over six years is $7.5 billion a year, from a company whose run rate reportedly passed $65 billion at the end of July. It’s a large cheque, not an implausible one.

The campus is the more interesting half

Monarch isn’t new, and Anthropic isn’t its anchor. Nscale announced the site on 16 March 2026 at GTC, on up to 2,250 acres in Mason County, with a Microsoft letter of intent for 1.35 gigawatts of Vera Rubin NVL72 capacity in phase one. Nscale puts the total power runway beyond 8 gigawatts.

Where does the power come from? Not the grid, at least not at first. Nscale is building what it calls America’s first state certified AI microgrid, running independently of the local network, with Caterpillar G3500 natural gas generator sets scaling to 2 gigawatts of generation by the first half of 2028. Grid tie-in comes later, and Nscale frames it as export capability rather than a dependency.

That’s the detail we’d actually keep. A hyperscale AI site with its own gas plant is a different risk object from one hanging off a utility. Fuel supply becomes an operational input. Local air permits become a schedule item. And the state politics get louder rather than quieter as the megawatts climb, because $80 million a year of projected state tax revenue buys a lot of patience right up until it doesn’t.

Bar chart on a logarithmic scale of compute capacity in megawatts across three Anthropic infrastructure agreements: up to 5,000 megawatts under the Amazon expansion announced 20 April 2026, about 460 megawatts under the Nscale agreement reported 26 August 2026, and 133 megawatts under the Volta agreement in Norway reported 4 August 2026.
Log scale, because otherwise you can't see Volta at all.

460 MW is a diversification move, not a step change

Line it up against what Anthropic already has and the size stops looking dramatic. The Amazon expansion announced on 20 April 2026 covers up to 5 gigawatts. This is 460 megawatts. Under a tenth of that ceiling, and about a third of what Microsoft signed a letter of intent for at the same campus.

Run the arithmetic on the reported numbers and you get about $97.8 million per megawatt across the six years. The Volta deal in Norway, reported on 4 August, works out around $75.2 million per megawatt on $10 billion and 133 megawatts. Roughly 30 percent apart. I wouldn’t read too much into that gap, honestly, because we have no idea what each contract actually includes, and Vera Rubin density in 2027 is not Vera Rubin density in 2026. But if someone tells you contracted compute is getting cheaper per watt, those two datapoints don’t say so.

What the deal does buy is supplier count. Volta, AMD, SpaceX, Google, Broadcom, Amazon, and the Theseus platform with Macquarie and GIC that landed on 10 August with no numbers attached at all. Nscale is one more name that isn’t AWS. For a company that filed confidentially for an IPO in June, a supply chain with seven counterparties reads better than one with two.

What it changes for you, today

Nothing. That’s not a snarky answer, it’s the whole practical takeaway.

Checklist of what the reported Anthropic and Nscale agreement changes for a team paying for Claude in August 2026: no price, rate limit or region moved on 27 August 2026; the West Virginia site is not due online until the end of 2027; neither company confirmed the deal; Anthropic has acknowledged peak hour reliability problems; and the Monarch campus runs on a natural gas microgrid rather than the local grid.
Four of these five rows are things to file, not act on.

The reason to care is the shape of the answer. Anthropic put the problem in writing back in April: unprecedented consumer growth had impacted reliability and performance for free, Pro, Max and Team users, especially during peak hours. That was the company’s own language in its Amazon announcement, not a critic’s. The fixes it has bought since split into two piles. Amazon capacity that lands in months. Purpose built sites that land in years.

Monarch is firmly in the second pile. So is most of what has been announced since spring, because Vera Rubin itself is only now reaching its first named customers and a 2,250 acre campus with its own gas plant takes as long as it takes.

If you’re writing a 2027 capacity assumption into a plan, this is a reasonable input. If you’re trying to work out why your agent run stalled at 11am on a Tuesday, it isn’t one.

Sources

Bloomberg, 26 August 2026, first report of the $45 billion figure. CNBC for the 460 megawatts, the six year term, the West Virginia location and the late 2027 date. TechCrunch for the Vera Rubin detail and the run of prior Anthropic compute agreements. Campus specifications, the microgrid, the Caterpillar generation figures and the Microsoft letter of intent come from Nscale’s own Monarch announcement of 16 March 2026. The peak hour reliability quote is Anthropic’s, from its 20 April 2026 Amazon compute announcement. The Volta terms are from TechCrunch, 4 August 2026. Run rate and IPO filing details are from Forbes, 18 August 2026.

Frequently asked questions

Did Anthropic and Nscale confirm the $45 billion deal?

No. Bloomberg reported it on 26 August 2026 citing people with knowledge of the matter who asked not to be named because the details are confidential, and CNBC matched the report the same day. Anthropic declined to comment. Nscale has not published a press release about it, and there is nothing on either company newsroom as of 27 August 2026. Treat the figures as well sourced reporting rather than disclosed terms.

How much compute is 460 megawatts?

Enough electricity for roughly 345,000 US homes at any one time, by the comparison used in the coverage. In Anthropic terms it is modest: the Amazon expansion announced on 20 April 2026 covers up to 5 gigawatts, so this contract is under a tenth of that ceiling. It is also about a third of the 1.35 gigawatt first phase Microsoft signed a letter of intent for at the same campus.

When does this capacity actually come online?

The end of 2027 for the West Virginia site, which is also when Nscale said Vera Rubin deployment at Monarch begins, phased across multiple tranches. Nothing in this contract adds a single megawatt to Claude before then. If your 2026 problem is rate limits at 10am Eastern, this deal is not the fix.

What is the Monarch Compute Campus?

Nscale's flagship US site in Mason County, West Virginia, on up to 2,250 acres. Nscale announced it on 16 March 2026 alongside a Microsoft letter of intent for 1.35 gigawatts of NVIDIA Vera Rubin NVL72 capacity, with Caterpillar G3500 natural gas generator sets scaling to 2 gigawatts of generation by the first half of 2028. Nscale calls the power setup America's first state certified AI microgrid, running independently of the local grid, with a runway beyond 8 gigawatts.

Why is Anthropic signing so many of these?

Because it has said, in its own words, that growth broke things. In the 20 April 2026 Amazon announcement it wrote that unprecedented consumer growth had impacted reliability and performance for free, Pro, Max and Team users, especially during peak hours. Since then it has stacked agreements with Volta, AMD, SpaceX, Google and Broadcom, plus the Theseus platform with Macquarie and GIC. It also filed confidentially for an IPO in June 2026, which is the other reason contracted capacity keeps getting announced.

Tags: aianthropicbig-techdata-centerinfrastructurenews
Share197Tweet123
stephane

stephane

  • Trending
  • Comments
  • Latest
Answer card: Proton Lumo 2.0 is private by policy, not by locality. Saved history is locked so even Proton cannot read it, but the prompt is decrypted on a Proton EU server to answer it, then forgotten.

Proton Lumo 2.0 review: how private is it, really?

3 September 2026
The Agentic Coding section of the official Hy4 preview benchmark appendix published by Tencent, a table comparing Hy3 and Hy4 preview against DeepSeek V4 Pro 0813, Qwen 3.8 Max, GLM 5.3, Kimi K3, GPT 5.6 Sol and Claude Opus 5 across SWE-bench Multilingual, SWE-bench Pro, DeepSWE, three SWE Atlas tasks, SWE-Marathon, Terminal-Bench 2.1, NL2Repo-Bench, CyberGym, ProgramBench, PostTrainBench and Harbor-Index.

Tencent’s 770B Hy4 tops one benchmark row in 46

3 September 2026
Answer card: Qwen 3.7 Max is API-only and cannot run locally yet; the open Qwen models (Qwen 3.6 27B, qwen3:8b to 32b) run offline via Ollama.

Qwen 3.7 local: what you can actually run offline

22 June 2026
Answer card: JWTs are not encrypted, anyone can read them; the signature proves who issued the token, not who may read it.

Are JWTs encrypted? No, and the difference will bite you

0
Answer card: a random 8 character password falls in under 2 hours offline, while 16 random characters hold for 1.4 trillion years at the same speed.

How long does it take to crack a password in 2026?

0
Answer card: three DNS records decide if your mail lands or bounces; SPF lists allowed senders, DKIM signs messages, DMARC sets the failure policy.

SPF, DKIM and DMARC explained: the records your email needs

0
Answer card stating that Ternary Bonsai 2 27B, released by PrismML on 17 September 2026 under Apache 2.0, packs Qwen3.8 27B into 5.95 gigabytes at 1.72 bits per weight, keeps 98.2 percent of the 14-benchmark average, about 75 percent on SWE-bench Verified and Terminal-Bench 2.1, and needs PrismML's llama.cpp fork to run.

Does Bonsai 2 27B really keep 98% of Qwen3.8 in 5.95 GB?

20 September 2026
Answer card stating that Jev 1.13 from TypeSafe AI is a decision model in early access since 15 September 2026 that returns typed probabilities instead of text, priced at 42 dollars per billion input tokens with output tokens free, answering in 70 to 500 milliseconds, with a 64K token request budget, text input only, and a documented list of things it does badly, including counting and dates.

Jev 1.13 bills $42 a billion tokens, and it can’t count

19 September 2026
Answer card stating that Qwen3.8-Omni-Flash launched on 17 September 2026 as an API only model on Alibaba Cloud Model Studio, taking text, images, audio and video in a 1M token context and returning text only, priced at 0.15 dollars per million input tokens for every modality and 0.47 dollars per million output tokens in the international regions, with no open weights published and the Qwen-Live Harness GitHub repository returning 404.

Qwen3.8-Omni-Flash bills audio at $0.15 and ships no weights

18 September 2026
  • About
  • Contact
  • Privacy
  • Legal

Copyright © 2026 Stephane Cardon.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Home
  • Articles
    • Security
    • Network
    • Dev
    • Sysadmin
    • SEO
    • Email & DNS
  • Tools
    • Network tools: free, fast, no signup
    • Security tools: free, fast, no signup
    • Developer tools: free, fast, no signup
    • Sysadmin tools: free, fast, no signup
    • SEO tools: free, fast, no signup
    • Email & DNS tools: free, fast, no signup
  • Download
  • About

Copyright © 2026 Stephane Cardon.