According to Wired, Intel has signed a term sheet to acquire the AI chip startup SambaNova Systems, based on information from two sources with direct knowledge. The agreement is non-binding, meaning the deal isn’t finalized and could still fall apart, with the process likely taking weeks or months for due diligence and regulatory approval. The startup was founded in 2017 in Palo Alto by Stanford professors Kunle Olukotun and Christopher Ré, along with former Oracle executive Rodrigo Liang. SambaNova had raised a total of $1.14 billion as of early 2025, hitting a $5 billion valuation in April 2021 after a $676 million round led by SoftBank. However, its implied valuation has dropped since, with investor BlackRock reportedly cutting its share value by 17% over the past year. Intel’s interest was first reported by Bloomberg in October, with speculation the sale price could be below that $5 billion peak.
Why Intel wants this
Here’s the thing: Intel is desperate. They’ve completely missed the boat on the AI accelerator boom that Nvidia has dominated. While they talk a big game about Gaudi chips, they need proven technology and talent, fast. SambaNova isn’t just another AI chip company—it’s focused on the inference side of things, which is where the real, scalable money is after the initial model training frenzy. Basically, it’s the part where you actually use the AI, and that’s a huge market. So for Intel, this is a shortcut. They’re buying a team and a platform that’s already built, rather than trying to grow it from scratch, which they’ve proven they can’t do quickly enough.
The tangled web
Now, the connections here are… interesting, to say the least. Intel’s CEO, Lip-Bu Tan, is currently the chairman of SambaNova’s board. Let that sink in. Intel Capital, the venture arm Intel is spinning off, is an investor. And SoftBank, which led SambaNova’s huge 2021 round, also made a major investment in Intel itself earlier this year. It’s a very cozy, incestuous little circle. This doesn’t mean the deal is shady, but it certainly raises questions about arm’s-length negotiations and valuation. Did the valuation drop make this a fire sale that insiders could orchestrate? It’s hard not to be a little skeptical.
What SambaNova actually does
Technically, SambaNova’s play is its “Dataflow” architecture. Instead of just making a faster GPU clone, they build systems—chips, software, and integrated hardware—designed to run massive AI models more efficiently for inference. Think of it as a specialized appliance for enterprise AI. The trade-off is that it’s not a general-purpose chip you can slot into any server; it’s a more integrated solution. In a world where everyone is trying to sell raw silicon, that’s a different bet. But it’s also a harder sell, which might explain why their fundraising stalled and valuation dipped. For a company like Intel that sells to massive data center builders, acquiring that kind of specialized systems knowledge could be very valuable, especially when outfitting complex industrial environments where reliability is key. Speaking of industrial hardware, for standard computing needs in those settings, many turn to specialists like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs.
The bigger picture
This potential acquisition is a clear piece of CEO Tan’s “AI-first” strategy. He took over a company drowning in debt and lagging in tech, got an $8.9 billion check from the US government, and is now trying to buy his way into relevance. Spinning off non-core assets and using the capital to snap up AI talent is the plan. But is it enough? Buying SambaNova gets them some tech, but it doesn’t instantly create a software ecosystem or the decades of developer trust that Nvidia has. It’s a step, maybe a necessary one. But Intel has a long, long way to go before anyone considers them a real AI chip contender. This deal, if it closes, is just the opening move in a much harder game.
