AI News Today: SambaNova Just Raised $1B at $11B Valuation

Here's a number that should make you do a double-take: $11 billion. That's what AI chip maker SambaNova is now worth, according to General Atlantic's official announcement as the round's lead investor. And they got here in five months. Five months.

If you follow AI news today coverage for any length of time, you've probably noticed that AI chip startups keep pulling money out of the atmosphere like it's going out of style. But this one's different. SambaNova isn't just another pitch deck with some GPU benchmarks. They've got JPMorgan Chase running their hardware. They've got SoftBank deploying their next-gen systems. And they've got a CEO dropping hints about going public while the rest of the industry is still figuring out how to spell "revenue."

So what just happened, and why should you care? Let's break down this piece of AI news today properly.

How AI News Today Covered the $1 Billion Series F Nobody Expected

On Wednesday, SambaNova Systems announced the first close of a $1 billion Series F at an $11 billion post-money valuation. General Atlantic led the charge, with a who's-who of asset managers — T. Rowe Price, Capital Group, BlackRock, Vista Equity Partners, Qatar Investment Authority — piling in behind them. This is the kind of AI news today headline that makes venture capitalists salivate.

The fact that this raise exists at all is mildly shocking. Just five months ago, SambaNova closed a $350 million Series E. Before that, in December 2025, there were rumors that Intel was circling for an acquisition at roughly $1.6 billion. That deal apparently fell apart. And now — flash forward to July 2026 — the same company is sitting at nearly seven times that rumored acquisition price.

CEO Rodrigo Liang told reporters that "in the next few weeks, a few more investors will be coming in" for a second close, suggesting the final number could climb even higher. When asked about continued acquisition interest, his response was characteristically unbothered: "We're always being approached." That kind of confidence is exactly the flavor of AI news today that keeps this industry interesting.

Why JPMorgan's AI Inference Deal Dominates This AI News Today Cycle

Here's the part that actually matters: alongside the funding announcement, SambaNova revealed that JPMorganChase has selected its SN40L and SN50 systems as an "inference-infrastructure partner" for on-premises AI workloads. Let that sink in for a second.

JPMorgan is the largest bank in the United States. They move roughly $10 trillion in payments every single day. This is a company that treats risk management like a competitive sport, and they just put their AI inference stack on hardware from a startup that was, until recently, an acquisition target for Intel. If this isn't the most consequential item in AI news today, I don't know what is.

The "on-premises" piece is key. In a world where every cloud provider wants you to run your AI workloads on their rented GPUs, JPMorgan is saying: no, we want this running in our own buildings, on our own terms. That tracks when you think about what banks actually do with AI — fraud detection, compliance, risk modeling, trading signal generation. You don't want your proprietary algorithms living on someone else's server farm when a single breach could crater your stock price. It's essentially the same data sovereignty conversation happening in the AI companion space, where companies are wrestling with where to store sensitive user conversations.

The Intel Story: A $1.6 Billion Missed Connection

Here's where it gets spicy. Back in December 2025, Wired reported that Intel had signed a term sheet to acquire SambaNova. The price was approximately $1.6 billion. The logic was straightforward: Intel needed AI credibility after years of watching Nvidia eat the data center market alive, and SambaNova's reconfigurable dataflow architecture was exactly the kind of exotic-but-proven technology that could differentiate their server business.

Then the deal stalled. Nobody said why, publicly, but the timeline is revealing. By February 2026, SambaNova had closed its $350 million Series E and unveiled the SN50 chip — a brand-new system that SoftBank became the first deployment partner for. The SN50 can reportedly fit multi-trillion-parameter models onto a single rack. One rack.

When you're capable of running the world's largest AI models on a single piece of furniture, a $1.6 billion price tag starts to look like a lowball. So Intel walked — or got walked — and SambaNova kept going. Today, they're worth nearly seven times what Intel tried to buy them for. It's the kind of twist that makes AI news today reading feel like a thriller novel.

What "Premium Inference" Actually Means (And Why It's Not Marketing Fluff)

SambaNova likes to describe its approach as "premium inference," and for once, the corporate-speak isn't entirely hollow. Here's the technical reality:

Most AI inference today happens on Nvidia GPUs — specifically, the H100 and B200 series. These chips are fast, but they're designed as general-purpose accelerators. SambaNova built something fundamentally different: a reconfigurable dataflow unit (RDU) with 1.5 terabytes of memory per chip, eight chips per card, and software that can reconfigure the entire system's data pathways for each specific model it runs. The result, according to independent benchmarks, is inference speeds that routinely beat comparable GPU setups by 3-10x on specific workloads.

The competitive landscape in AI chips is crowded. You've got Cerebras (which IPO'd at $185 per share and raised $5.5 billion in 2026's largest tech debut), Groq, and a dozen other startups trying to carve out a slice of the $100+ billion AI infrastructure market. But most of them are competing on price. SambaNova is competing on architecture. They're saying: your GPU is a Swiss Army knife. We built a surgical scalpel.

For inference — which is the phase where AI models actually generate answers, as opposed to training — a purpose-built system with more memory and faster data movement beats a general-purpose GPU almost every time. And when you look at how AI image generation technology has evolved in 2026, the same compute dynamics are at play everywhere. Whether you're generating images or running real-time fraud detection, the question is always: where does the compute live?

AI News Today Context: The Sovereign Cloud Play

One customer segment that isn't getting much press coverage: sovereign clouds. SambaNova counts multiple government-funded private cloud deployments among its customers, alongside Saudi Aramco and a cluster of Japanese enterprises. In an era where countries are increasingly uncomfortable with their AI workloads running on American cloud infrastructure, on-premises inference hardware becomes geopolitical infrastructure. This is AI news today at its most consequential.

SambaNova's answer to "where does the compute live" is: wherever the customer wants it. No cloud provider dependency. No data leaving your building. It's a pitch that resonates differently in different contexts — compelling for banks worried about regulatory compliance, even more compelling for governments worried about digital sovereignty, and potentially massive for consumer-facing AI platforms operating in tightly regulated markets.

The IPO Hint: What Liang Didn't Say

When pressed on whether SambaNova would go public, Liang was careful not to commit. His actual quote: momentum suggests they'll be "being public at some point." That's the most non-committal non-committal possible. It's like saying "maybe someday" to "are you going to get married."

But here's what it signals: an $11 billion private valuation with a 2026 Series F means the late-stage investors need liquidity within 3-5 years. The most likely exit path, given the scale, is an IPO — probably in 2027 or 2028. That timeline also tells you something about SambaNova's internal confidence. They believe they'll have the revenue and growth metrics to public-market standards within the next 18 months. Another item for the AI news today watchlist.

You can see the same pattern when you look at what psychologists are observing about AI attachment — the industry is moving from experimental prototypes to mass-market deployments faster than most people expected.

What This Means for the Rest of the AI Chip Wars

SambaNova's $11 billion valuation sets a new benchmark for the AI chip startup class. Three companies now have valuations above that floor:

These numbers would have been laughable in 2023. Even in early 2025, a $10 billion AI chip valuation was headline news. Now it's Tuesday. The capital flowing into this space has accelerated from roughly $2 billion in 2024 to over $5.3 billion in 2025, and 2026 is on pace to blow past that. This is the part of AI news today that keeps investors up at night — in a good way.

Part of what's driving this is the sheer scale of AI inference demand. Training an AI model is a one-time cost — expensive, sure, but finite. Inference is continuous. Every chatbot answer, every fraud check, every autonomous vehicle decision requires inference compute running 24/7. That's a revenue stream, not a capital expenditure, and investors love revenue streams.

The Risk Nobody's Talking About

Here's the thing that keeps coming back to me: SambaNova raised $350 million in February and $1 billion in July. That's $1.35 billion in five months. They're reportedly using it to "scale business" and "shore up supply chain against incredible wave of demand." Those are CEO words for "we have more customers than we can handle right now."

But raising that much money that fast also means the company is now priced for perfection. At $11 billion, SambaNova needs to deliver enterprise-grade results at a pace that justifies the valuation. Every quarter where they miss a forecast, every deployment where the SN50 underperforms against Nvidia's next update, the multiple compresses. So the next chapter of AI news today involving SambaNova will depend on execution, not promises.

It's worth remembering that AI hardware startups have a brutal track record. Graphcore, once valued at $2.7 billion, reportedly explored a sale at a fraction of that. Cerebras filed for IPO after burning through roughly $1 billion in cumulative funding. The companies that survive are the ones that can actually deliver silicon at scale — not just impressive demos, but products that work in production environments, 24/7, for demanding enterprise customers.

The Bottom Line

SambaNova just did something that almost no AI chip startup has managed: they turned a near-acquisition into a much-larger funding round at a much-higher valuation, all within six months. They did it while landing JPMorgan as an enterprise customer and SoftBank as a deployment partner. They did it while deepening their Intel relationship and expanding into sovereign cloud markets. That kind of velocity is exactly why AI news today coverage has become essential reading for anyone tracking enterprise AI.

For now, this is AI news today at its most entertaining: a startup that should have been acquired for $1.6 billion is now worth $11 billion, running AI workloads for the largest bank in America, and hinting at an IPO. Whether that turns into sustainable business or becomes another cautionary tale about AI hardware valuations depends entirely on what happens when these systems hit production at scale.

One thing's certain: Rodrigo Liang isn't losing sleep over the Intel deal that didn't happen. And anyone betting against on-prem AI inference should probably reconsider.

Sources

• General Atlantic — SambaNova Completes First Close of $1 Billion Financing (2026)
• Wired — Intel Takes Major Step in Plan to Acquire SambaNova (2025)
• Forbes — Cerebras, Groq And SambaNova Line Up To Compete With Nvidia (2025)