Can Intel Finally Get AI Right? What SambaNova Means for the Chip Giant's Comeback

Why This Deal Matters More Than the Last One

Intel appears ready to make another major AI bet. The chip manufacturer is in advanced discussions to acquire SambaNova Systems for approximately $1.6 billion—a stark contrast to the company’s previous $5 billion valuation. This potential move comes at a critical juncture for Intel’s AI strategy, but skeptics have every right to question the company’s acquisition playbook.

The proposed target, SambaNova, represents a different kind of AI opportunity than Intel’s previous attempts. The company specializes in custom AI inference solutions rather than training chips. Its Reconfigurable Dataflow Units (RDUs) power the SambaRack system—an integrated hardware, networking, and software solution designed for efficient AI inference at scale.

Learning from Past Mistakes

Intel’s track record tells a cautionary tale. In 2019, the chip giant paid roughly $2 billion for Habana Labs, acquiring what was then a promising AI training processor called Gaudi. The investment came early in the AI cycle, before ChatGPT reshaped the entire industry landscape.

Despite reasonable technical performance, Gaudi failed to gain meaningful market adoption. Nvidia’s dominance proved insurmountable, primarily due to its proprietary CUDA ecosystem—a software platform that has spent nearly two decades becoming the industry standard for accelerated computing. Habana’s unfamiliar architecture and immature software environment couldn’t compete. Adding to the problem, Intel simultaneously pursued an unfocused strategy, developing both Gaudi training chips and separate data center GPUs. This scattered approach left both initiatives without sufficient resources or clear positioning.

The damage was compounded when Intel cancelled Falcon Shores, its data center GPU project that incorporated Gaudi technology, effectively conceding the AI training battlefield to Nvidia.

Why SambaNova Could Be Different

Several factors distinguish this potential acquisition from the Habana Labs experience. First, Intel Capital already invested in SambaNova, and Intel’s CEO Lip-Bu Tan serves as SambaNova’s Chairman—suggesting stronger organizational alignment than existed with Habana.

More importantly, the market dynamic has shifted. The AI inference space remains far more competitive than the training market, where Nvidia’s grip is nearly unbreakable. Inference workloads reward efficiency and specialized solutions, creating openings for custom chip architectures that deliver superior performance-per-watt.

SambaNova has demonstrated early traction. The company recently secured three contracts to power sovereign AI inference clouds across Australia, Europe, and the United Kingdom. French cloud provider OVHcloud selected SambaNova’s systems to complement GPU-based infrastructure for its AI Endpoints offering. These wins suggest the market recognizes value in SambaNova’s integrated approach.

The Strategic Alignment

This acquisition aligns neatly with Intel’s recalibrated AI roadmap. After abandoning the AI training race, Intel shifted focus toward rack-scale solutions—complete integrated systems built for data center deployment. SambaNova’s inference rack platform fits this vision perfectly.

Whether this move will actually succeed remains uncertain. Intel’s acquisition success rate doesn’t inspire confidence, particularly in complex technology areas. Habana Labs proved that having good chips isn’t enough when Nvidia’s ecosystem advantage is overwhelming. However, positioning SambaNova as an inference-focused, rack-scale solution rather than a chip-only offering significantly improves the odds compared to Intel’s previous approach.

The real question isn’t whether SambaNova is a good company—it is. The question is whether Intel can support it effectively and help it compete in a market where efficiency, integration, and customer relationships matter as much as raw performance.