Nvidia-backed AI startup wants to solve AI’s ‘hard problem’

Jensen Huang is betting on a new AI lab that thinks the current chatbot frenzy is just the beginning.

Nvidia (NVDA) has already emerged as one of the greatest winners of the artificial intelligence boom.

Its chips now make up the backbone of most of the infrastructure powering modern AI, helping businesses like Microsoft (MSFT), Amazon (AMZN) and Google parent Alphabet (GOOGL) compete to construct bigger and more powerful systems.

That growing demand has turned Nvidia from a gaming-chip specialist into the engine of Silicon Valley’s AI economy.

But discreetly, Nvidia CEO Jensen Huang appears to be contemplating beyond the current chatbot craze.

Now the chip giant is supporting a new British firm that believes the current generation of artificial intelligence has only addressed the “easy problem.”

The business, Ineffable Intelligence, said it is working withNvidia to develop AI systems that learn continually from experience rather than through mostly human-generated data.

The business was founded in late 2025 by former Google DeepMind reinforcement learning scientist David Silver, one of the industry’s top researchers in machine learning and AI reasoning systems.

For investors in Nvidia, the agreement may be far more than just about one startup.

If reinforcement learning becomes a greater part of the next AI wave, the systems will need huge computational power, just the kind of demand that has driven Nvidia’s spectacular ascent over the past two years.

Nvidia backs a new approach to AI training

Unlike many of today’s leading artificial intelligence systems, which mainly train on internet content and human-created datasets, Ineffable Intelligence focuses on reinforcement learning.

This strategy allows AI systems to learn by trying different approaches and adjusting their strategies based on the results and feedback they receive over time.

The technique could one day enable AI systems to advance beyond simply duplicating human knowledge, researchers believe.

Instead, future models might learn to discover new ways and solve issues on their own.

“The next frontier of AI is superlearners — systems that learn continuously from experience,” Huang said.

Silver said much of the current AI industry has focused on teaching systems what humans already know.

“But now we need to solve the harder problem of AI: how to build systems that discover new knowledge for themselves,” Silver said in a company statement.

Nvidia and Ineffable announced the engineers from both businesses will work together on infrastructure capable of supporting enormous-scale reinforcement learning.

The agreement will pair Nvidia’s Grace Blackwell chips with its new Vera Rubin AI platform.

That’s critical to Nvidia’s long-term strategy.

The business is rapidly aiming to position itself not as a chip supplier, but as the core infrastructure provider for the next phase of AI research.

Related: Nvidia gets unexpected China opening as chip fight intensifies

Ineffable Intelligence launched its $1.1 billion seed investment round in April co-led by Sequoia and Lightspeed.

Joining the round were Alphabet, DST Global, Index Ventures and the U.K.’s Sovereign AI Fund, as well as Nvidia.

Massive investment rounds are a testament to how rapidly investors are pouring money into next-generation AI laboratories created by former researchers from OpenAI, DeepMind, Anthropic and Meta Platforms (META).

Related: Nvidia gets unexpected China opening as chip fight intensifies

The AI investment race has heated up substantially in the past year as corporations strive to grab talent and computing resources.

Key takeaways from Nvidia’s AI partnershipNvidia is partnering with Ineffable Intelligence on reinforcement learning infrastructure.The startup was founded by former Google DeepMind scientist David Silver.Ineffable raised $1.1 billion in seed funding earlier this year.The partnership will use Nvidia’s Grace Blackwell chips and Vera Rubin platform.Nvidia is positioning itself for the next phase of AI development beyond chatbots.

This approach has allowed Nvidia to command huge pricing power for AI chips and enhance its clout across the larger AI ecosystem.

Nvidia’s AI dominance keeps attracting new partners

The transaction also underscores the increasing dominance of Nvidia in almost every corner of the AI sector.

Huang has overseen Nvidia’s evolution from a graphics chip maker to the biggest supplier of AI computer infrastructure.

Their H100 and Blackwell chips are now at the heart of data-center expansion plans across Silicon Valley.

That demand has helped Nvidia deliver record financial performance in recent quarters, and momentarily pushed its valuation past $4 trillion earlier this year.

And meanwhile, the race of AI is fast going.

In the past year, an increasing number of companies created by former Big Tech experts have been funded by investors pouring billions into companies pursuing artificial general intelligence and sophisticated reasoning systems.

Another business, Recursive Superintelligence, created by former DeepMind engineer Tim Rocktäschel, is said to have raised $650 million.

Also, ex-OpenAI and Anthropic researchers have started new companies building next-gen AI systems.

More AI:

Micron sits at the center of a red-hot chip rallyIBM CEO sends blunt message on AI and quantum computingAnthropic CEO makes shocking admission about AI

It’s partnerships like this one that can help Nvidia lock its chips into the heart of the next generation of AI development no matter how much the underlying technology may shift.

The company’s strategy is moving away from just selling gear.

As it tries to become the foundational platform of the future of artificial intelligence, Nvidia is integrating itself right into the infrastructure and research stack of budding AI labs.

Jensen Huang arrives in Beijing as Nvidia seeks China chip breakthrough.

Photo by China News Service on Getty Images

Nvidia wants a front-row seat in the next AI shift

It is not apparent whether reinforcement learning will become the key route for artificial intelligence.

The AI explosion today is in great part due to large language models, trained on massive data sets and deployed on huge cloud-computing networks.

That recipe has already produced some of the market’s greatest winners, including Nvidia.

But more and more academics think future AI systems would need to do more than just respond to suggestions.

Eventually, sophisticated models could be able to test tactics, learn from the results, and improve themselves, rather than relying solely on static information produced by humans.

That’s the opportunity firms like Ineffable Intelligence are seeking to tap into.

And certainly, Nvidia doesn’t want to be left behind when the industry pivots again.

Partnership also gives Nvidia another avenue to cement its grip on the larger AI ecosystem as rivals are still trying to threaten its supremacy.

Nvidia faces stiff competition from rivals such as Advanced Micro Devices (AMD) and Intel (INTC), while authorities keep an eye on the chipmaker’s expanding market power.

Still, Nvidia is heavily integrated in almost every significant AI effort in Silicon Valley.

Whether reinforcement learning is the next AI gold rush is yet to be seen.

One thing is becoming increasingly clear about Nvidia, though: Huang has no intention of letting the company sit on the sidelines if artificial intelligence goes through another huge change.

Nvidia’s AI expansion timeline2022: OpenAI’s ChatGPT launch ignites a global AI infrastructure boom.2023: Nvidia’s AI-chip demand surges as cloud companies ramp up spending.2024: Nvidia becomes one of Wall Street’s most valuable companies during the AI rally.2025: Nvidia launches Blackwell AI chips to support next-generation AI workloads.2026: Nvidia partners with Ineffable Intelligence to expand into reinforcement learning infrastructure.

Related: Nvidia’s AI dominance faces uncomfortable new reality

#Nvidiabacked #startup #solve #AIs #hard #problem