• Latest
  • Trending
  • All
Official co-branded announcement image for the partnership, the ssi wordmark on a dark card beside the green Nvidia eye logo and wordmark, separated by a vertical rule on a black background.

Nvidia SSI deal: $5B reported, 10x compute on Vera Rubin

29 July 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
Answer card stating that OpenAI published its research acceleration measurements on 6 September 2026, that as of mid August 2026 its research organisation logged 3.1 agent workdays of coding agent runtime for every workday of human labour normalised to a standard eight hour day, and that OpenAI states this should not be read as a 3.1 times productivity gain because it measures runtime rather than delivered output.

OpenAI’s 3.1 agent-workdays per human day is not a 3.1x gain

7 September 2026
Answer card stating that Mullvad announced on 3 September 2026 that it is shutting down its public encrypted domain name system servers on 2 November 2026 and sponsoring the Quad9 Foundation instead, with 194.242.2.2 and its five sibling addresses all going away, and virtual private network customers unaffected.

Mullvad’s DNS servers go dark on 2 November, and Quad9 blocks no ads

5 September 2026
OpenAI announcement image for GPT-6 Astra, a spiral galaxy of white, blue and amber points of light curling around a bright core on a near black star field.

GPT-6 Astra lists at $10 and $50, 2.5x what GPT-5.6 Sol costs

6 September 2026
Google's official announcement image for the release, reading Introducing Gemini 3.8 Flash and 3.8 Flash Cyber in black type over a pale blue background with a blurred white chevron and the four colour Gemini spark below.

Gemini 3.8 Flash keeps the price and the 1 January cliff

3 September 2026
Answer card stating that Anthropic announced Enterprise Frontier Safeguards on 1 September 2026, that activity data used for misuse monitoring moves into cloud storage the customer controls under the customer own encryption keys, that Anthropic charges nothing for the feature while the cloud provider bills storage and egress, and that the phased rollout starts later in autumn 2026 with interim zero data retention on Fable 5 and Fable 5.1 for eligible customers.

Anthropic moves retention into your own cloud, for 30 days

3 September 2026
Official Google diagram of a client connection in three numbered steps: a DNS lookup with a query and an address, a TLS ClientHello and ServerHello, then a content exchange with a website. A callout on the DNS step reads 25% of global web traffic is now protected by encrypted DNS, and a callout beside an Android phone on the ClientHello step reads Android 17 supports ECH GREASE by default.

Android 17 hides the SNI, not your DNS or destination

3 September 2026
  • About
  • Contact
  • Privacy
  • Legal
Tuesday, September 15, 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

Nvidia SSI deal: $5B reported, 10x compute on Vera Rubin

by stephane
29 July 2026
in Dev
0
Official co-branded announcement image for the partnership, the ssi wordmark on a dark card beside the green Nvidia eye logo and wordmark, separated by a vertical rule on a black background.
491
SHARES
1.4k
VIEWS
Share on FacebookShare on Twitter

A four-line post on X, from an account that had barely spoken in two years. That's how the biggest number in this week's AI coverage got loose. On 27 July, Nvidia and Ilya Sutskever's Safe Superintelligence announced a long-term strategic partnership: SSI gets access to the Vera Rubin platform and, by its own account, will 10x its compute in the next 12 months. Both parties called the investment substantial. Neither put a figure on it. The $5 billion now welded to every headline came from Bloomberg citing people familiar with the matter, and nobody with a signature on the deal has confirmed it. So we have a genuinely interesting announcement, wrapped in a number nobody stated.

The short answer

Nvidia and Safe Superintelligence announced a long-term strategic partnership on 27 July 2026. SSI gets access to the Vera Rubin platform and says that will 10x its compute within a year. The $5 billion in the headlines is Bloomberg’s reporting from unnamed sources, not a stated figure. SSI has never said what it runs on today, so the multiplier has no base under it.

10xcompute in 12 months, per SSI
$0disclosed by either party
$32BSSI valuation, Feb 2025 round
Answer card: Nvidia and SSI announced a long-term strategic partnership on 27 July 2026 giving SSI access to the Vera Rubin platform with an order-of-magnitude compute increase, SSI says 10x within 12 months, and neither party disclosed the investment amount despite the $5 billion figure in the coverage.
The platform is on the record. The money is somebody's sourcing.

What Nvidia and SSI actually said

The release went up on Nvidia’s newsroom on 27 July, and it’s short. A long-term strategic partnership, SSI getting access to Vera Rubin systems, technical collaboration on current and future Nvidia compute platforms, and compute rising by an order of magnitude.

Jensen Huang’s line: “Ilya has pioneered fundamental breakthroughs at the foundation of modern AI, beginning with AlexNet. We are excited to see what new breakthroughs SSI will discover powered by our Vera Rubin platform.”

Sutskever’s: “We have research that is worthy of scaling up, and having access to a big NVIDIA computer will let us do so.”

A big NVIDIA computer. I like that phrasing more than I expected to. It’s the least marketing sentence in any GPU press release this year.

The clock came from SSI’s own post, not the release:

“We are announcing a long-term strategic partnership with NVIDIA. NVIDIA is making a substantial investment in SSI that will let us 10x our compute in the next 12 months. We reached the point where our research is worth scaling and with this partnership we will be able to. We are honored by NVIDIA’s conviction.”

SSI Inc. (@ssi on X), 27 July 2026.

One line in Nvidia’s release runs the other direction and mostly got skipped in the coverage: Nvidia says it obtained rare access into SSI’s closely guarded research. For a company that has published nothing since 2024, that isn’t a small concession. Whatever the cash number turns out to be, research access was part of the price.

The official co-branded image released with the announcement: the ssi wordmark on a dark card at left, the green Nvidia eye logo and wordmark at right, split by a vertical rule on black.

Image: Nvidia, from the 27 July partnership announcement.

Ten times what, exactly

Here’s where it stops being clean.

SSI has never disclosed its compute. Not GPU counts, not megawatts. It partnered with Google Cloud last year to power its research, and that arrangement came with no published capacity either. So 10x fixes a ratio against a number nobody outside the company knows. Ten times a modest research cluster and ten times a serious training fleet are different announcements, and we can’t tell which one this is.

Same problem with the money. Bloomberg put it at $5 billion, sourced to people familiar with the matter, and other outlets have credited the same figure elsewhere. It might be exactly right. It just isn’t a disclosure, and it’s already being quoted as though it were, which is how a sourced estimate becomes a fact in about three weeks.

What we can size is the history. SSI raised $1 billion at a $5 billion valuation in 2024, then $2 billion at $32 billion in February 2025. Against that, a multi-billion strategic investment isn’t a rounding error. It’s most of what the company has ever raised, arriving from its hardware vendor.

Which is worth naming plainly, without getting dramatic about it. Nvidia invests in SSI, SSI buys Nvidia hardware. That structure is all over 2026 AI capex, it isn’t fraud, and it does mean “Nvidia invested” and “Nvidia sold” stop being independent signals about demand. Read them as one event.

Checklist separating what the Nvidia SSI announcement confirmed from what it did not: the partnership, the Vera Rubin platform, the research access clause and SSI's own 10x wording are on the record, while the $5 billion figure, SSI's current compute base, the delivery schedule and any evaluable output are not.
Four things stated, four things everyone is quoting anyway.

The part that matters if you buy GPUs

Forget the valuation for a second. The useful fact here is the platform.

Vera Rubin is Nvidia’s generation after Blackwell, second-half-of-2026 silicon that’s only now reaching its first named customers. We went through that when BMS committed to eight DGX Vera Rubin NVL72 systems, and again when SK signed a 2 GW build in Korea. SSI now joins that list.

None of these are clusters running workloads. They’re allocation. And the pattern across July is that Rubin capacity keeps getting spoken for by named partners before the parts ship in volume, which is the thing to model if you’re budgeting Rubin-class capacity for 2027. You aren’t first in the queue, and the queue is being published one press release at a time.

The softer signal: SSI ran on TPUs and is now building on Nvidia. That doesn’t tell you TPUs are worse. Honestly I think it says more about tooling gravity and about who was willing to write the cheque than about silicon. But a frontier lab with no revenue picking the expensive ecosystem because somebody else is funding the purchase is exactly the dynamic that makes GPU demand look sturdier than the businesses underneath it.

Timeline of SSI's public compute arrangements: a Google Cloud TPU partnership in 2025 with no published capacity, the Nvidia Vera Rubin partnership announced 27 July 2026 with a substantial but undisclosed investment, and a stated 10x by mid 2027.
Two public arrangements, no published capacity behind either one.

My read, and I might be wrong: the research access clause is the most informative sentence in the whole release, and it got almost no coverage. Nvidia doesn’t need SSI’s revenue. What it bought was a window into what a lab that hasn’t published in two years is actually doing. That’s a strange thing to pay for, and a fairly strong hint that somebody at Nvidia has already seen something.

Sources: the partnership terms, the Vera Rubin platform access, the order-of-magnitude framing, the research access clause and both executive quotes come from the official Nvidia newsroom announcement of 27 July 2026. The 10x-in-12-months wording is SSI’s own, posted to @ssi on X. SSI’s funding history, the $32 billion valuation, the earlier Google Cloud arrangement and the absence of any commercial product are from TechCrunch’s report on the partnership. The $5 billion figure is Bloomberg’s, cited to people familiar with the matter and relayed by Tech Startups. No investment amount is asserted here as confirmed, because neither party disclosed one.

Frequently asked questions

How much did Nvidia invest in Safe Superintelligence?

Neither company said. The 27 July announcement describes a substantial investment and gives no figure, and SSI's own post uses the same word. The $5 billion quoted across the coverage comes from Bloomberg citing people familiar with the matter, with other outlets attributing the same number elsewhere. It may well be accurate, but it is reporting from unnamed sources rather than a disclosure, and it should be quoted that way.

What does 10x compute actually mean for SSI?

As an absolute quantity, nothing you can check. SSI posted that the investment "will let us 10x our compute in the next 12 months", and Nvidia's release describes an order of magnitude increase. But SSI has never published what it runs on today, so the multiplier sets a ratio against an unknown base. Ten times a modest research cluster and ten times a large training fleet are very different announcements, and there is no public way to tell them apart.

Is SSI switching from Google TPUs to Nvidia GPUs?

It is adding an Nvidia path, and how much of the TPU work moves is not public. SSI partnered with Google Cloud last year to power its research, and that arrangement came with no published capacity either. The new deal names the Vera Rubin platform and commits both sides to technical collaboration on current and future Nvidia compute platforms. Neither company has said the Google arrangement ends.

When will SSI actually get Vera Rubin systems?

No schedule was announced. Vera Rubin is Nvidia's generation after Blackwell, second-half-of-2026 silicon that is only now reaching its first named customers, so this is a commitment to future capacity rather than a cluster running workloads. No system count, no site and no delivery date appear in the release.

Has SSI released anything since it launched?

No product and no published research. The company launched in 2024 on a stated straight-shot approach to safe superintelligence, deliberately skipping commercial releases and short-term revenue, and has operated in stealth since. That is what makes one clause in Nvidia's release notable: Nvidia says it obtained rare access into SSI's closely guarded research as part of the deal.

Tags: aifundinggpuhardwarenewsnvidiavera-rubin
Share196Tweet123
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 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
  • 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.