The "Old Huang-style Manhattan Project" is coming: 5 billion has been poured into Ilya, and a sky-high 250 billion guarantee for OpenAI has also been approved.

Something huge just dropped!

After two years out of the spotlight, Ilya has finally resurfaced.

The Safe Superintelligence Labs (SSI) he founded has officially announced a long-term strategic partnership with NVIDIA.

The news has sent shockwaves across the entire tech industry.

This is far more than a simple GPU procurement deal.

According to sources, NVIDIA is making a "substantial" $5 billion investment in SSI, and has committed to boosting SSI's available computing power tenfold within the next 12 months.

This means the "sword" of safe superintelligence that Ilya has been forging is finally about to be unsheathed.

All the brilliant ideas the genius scientist had for "safe superintelligence", and all the large-scale verification experiments that were stuck behind compute bottlenecks in the lab, can finally run at full speed. Humanity, hold onto your seats!

And right on the heels of this announcement, another bombshell from NVIDIA has broken.

Foreign outlet WSJ reports that NVIDIA is considering providing OpenAI with a jaw-dropping $250 billion financial guarantee to help it acquire a supercomputing hub worth a total of $500 billion, with a power consumption of up to 10 gigawatts.

Neither the Manhattan Project nor the Apollo moon landing program can compare to the grand scale of Jensen Huang's master plan.

What exactly is the massive strategy that Jensen Huang is orchestrating?

The Return of the King! Ilya Breaks His Silence, SSI Goes Into Overdrive

Today, after nearly two years of lying low, Ilya has finally announced his return to the world.

He stated: "Our research has reached a critical point where it is now worth scaling up."

At the end of 2023, after the dramatic internal power struggle at OpenAI ended with Sam Altman's lightning-fast return, Ilya left OpenAI six months later. With lingering concerns about unaligned AI, he founded SSI.

This mysterious startup has never published a single paper or launched any public product, yet at its founding, it raised $2 billion at a staggering $320 billion valuation.

Ilya has only one single goal: to develop safe superintelligence.

Prior to this partnership, SSI remained a mysterious "black box". Industry insiders only knew that its team consisted of dozens of researchers who primarily relied on Google's TPUs to run their experiments.

But starting today, SSI will never be the same again!

Computing Power Explodes 10x in a Year, Full Transition to NVIDIA's Ecosystem

According to the publicly shared partnership details, NVIDIA has made that "substantial" $5 billion investment in SSI.

SSI will gain priority access to NVIDIA's next-generation Vera Rubin hardware platform.

This means SSI's computing power will achieve an order-of-magnitude leap within the next 12 months!

In an interview, Ilya shared exciting new updates —

We have made some promising research breakthroughs. We have identified a new peak to climb, and this direction is slightly different from the research I have done in the past.

As early as last November, Ilya hinted that the marginal gains of simply scaling up data and compute for LLMs are diminishing, and humanity is "re-entering a research-driven era".

By "research-driven", he does not mean that computing power is no longer needed — far from it!

What Ilya means is that we need to restart massive, storm-like scaled training under a completely new architecture and paradigm.

SSI believes it has already found this new paradigm — most likely a new architecture different from Transformer, or a new path for reinforcement learning and reasoning — and that this paradigm is "worth scaling".

SSI may have already built an early system that possesses these capabilities:

• Learning new skills using only an extremely small amount of experience

• Detecting that the current line of reasoning is failing before reaching the final answer

• Updating itself without overwriting its existing capabilities

• Transferring experience learned from one scenario to a completely different type of problem

• Continuing to learn even after deployment

You can think of it as a uniquely gifted young mind: it may not know everything in every field at first, but it can rapidly master any professional skill it is introduced to.

And all the evidence pointing to this is almost irrefutable —

Ilya once pointed out that the fundamental bottleneck of current AI is its extremely weak generalization ability, with models "far inferior to humans in generalizing across tasks". At the same time, he described SSI's vision as creating a "15-year-old with superintelligence" that can learn any skill required for any profession.

When asked how to achieve this human-like learning ability, he admitted he had strong ideas about a certain machine learning principle, but refused to discuss it publicly.

The WSJ report revealed that SSI's secret research direction is focused on "the long-overlooked aspects of how the human brain operates".

Some netizens have speculated their potential technical roadmap:

Experience → Internal Judgment → Self-Correction → Persistent Learning → Transfer → Repeat.

The "internal judgment" component could be a brain-inspired module: humans do not need to finish 10,000 steps of a task to realize they are failing. Emotions, intuition, and judgment provide continuous real-time feedback.

Ilya argues that AI needs an equivalent internal "value function".

And after catching a glimpse of SSI's internal research work, Jensen Huang and NVIDIA's technical team immediately decided to bet big on this project.

Jensen Huang even personally endorsed the partnership: "Ilya has driven groundbreaking advances at the foundation of modern AI. We are incredibly excited to see what new breakthroughs SSI will discover using our Vera Rubin platform!"

Starting today, Ilya's lab has officially obtained the world's most premium unlimited computing power "experience pass".

Jensen Huang's Underlying Logic

Not Just a Shovels Supplier — Aiming to Be the "Creator"

With a market cap already soaring to $4.7 trillion, why would NVIDIA step in directly to invest $5 billion in SSI, a startup that has not even launched a single product yet?

Dan McAteer, a well-known Silicon Valley investor, offered a sharp, incisive analysis on this move.

SSI to NVIDIA is what DeepMind is to Google, OpenAI to Microsoft, Anthropic to AWS, and Cursor to SpaceXAI. Every tech giant with massive computing power must have its own forward-edge lab to place strategic bets on.

This observation lays bare the core logic of this capital game.

In the age of artificial intelligence, there will be one simple core metric that defines success: How much intelligence can be generated per unit of computing power? Whoever wins this battle will own the future.

Over the past year, Jensen Huang has been on a relentless "shopping spree". He did not just invest in SSI — he also placed a massive bet on Thinking Machines Lab, founded by OpenAI's former CTO.

Jensen Huang understands better than anyone that NVIDIA's competitive moat cannot rely on hardware alone.

Just recently, AMD announced it would invest up to $5 billion in Anthropic.

If all top-tier AI labs end up being locked down by Microsoft (with OpenAI), Google (with DeepMind), and Amazon (with Anthropic), NVIDIA will be reduced to a pure hardware contract manufacturer, at constant risk of being pushed out by in-house chips from these giants, such as Google's TPU and Microsoft's Maia.

That is why NVIDIA must deeply align itself with the brightest minds on the planet.

Investing in SSI is NVIDIA's ultimate ticket to the era of Artificial Superintelligence.

$250 Billion Sky-High Guarantee! The Unstoppable "AI Manhattan Project"

And NVIDIA's recent shocking revelations do not end there.

According to recent reports from WSJ and other foreign outlets, NVIDIA is in secret negotiations with OpenAI, planning to provide up to $250 billion in financing guarantees to help OpenAI "lock in" the world's most powerful supercomputing center!

How grand is this project?

Total Cost: Over $500 billion (roughly 3.6 trillion RMB)

Power Demand: 10 gigawatts

Developer: The energy subsidiary of SoftBank Group

Location: Southern Ohio, USA

This earth-shattering plan is essentially pushing the limits of human engineering history.

10 gigawatts of electricity is enough to meet the total power consumption of an entire megacity with a population of 10 million, like New York City.

To meet this demand, several nuclear power plants may need to be purpose-built, which will also require government-level regulatory approval.

What does $500 billion even mean?

The Manhattan Project cost roughly the equivalent of $300 billion; the Apollo moon landing program cost over $2500 billion in today's dollars.

But this supercomputing center, a joint project between OpenAI, SoftBank, and NVIDIA, costs more than ten times as much as the Manhattan Project!

What is interesting is that the role NVIDIA plays here is very subtle.

This $250 billion is not paid directly out of NVIDIA's pocket. Instead, NVIDIA is acting as a "guarantor", using its own balance sheet and credit rating to back the project.

If SoftBank seeks bank financing, and banks are worried that OpenAI cannot afford the rent, NVIDIA steps in to cover the risk. With a single word from Jensen Huang, hundreds of billions of dollars in funds can be instantly unlocked.

In addition, rumors say NVIDIA is also in separate negotiations to provide $350 billion in financing support for OpenAI to purchase its chips.

This computing power closed-loop game is absolutely masterfully designed.

1. NVIDIA provides a guarantee to OpenAI, enabling SoftBank to build the data center.

2. OpenAI borrows money to buy chips to fill this massive data center.

3. Whose chips do they buy? NVIDIA's, of course!

4. Hundreds of billions of dollars flow through the full cycle, and eventually turn into eye-popping revenue and profits on NVIDIA's financial statements.

As a result, NVIDIA not only locks in OpenAI, the world's biggest "capital guzzler", but also firmly ties SoftBank and Wall Street capital to its own chariot.

Fighting on Two Fronts, Jensen Huang's Grand Master Plan

If you look at both the $5 billion SSI investment and the $250 billion OpenAI guarantee together, you will see just how visionary Jensen Huang really is.

Current AI development is at a crossroads between two distinct paths:

Path 1: The