Intel plans to axe Nervana in favour of Habana, fundamentally altering their AI strategy

As the Intel enters this new decade, it sees itself as a silicon company, not as a CPU company. Looking at Intel’s hardware roadmap, it is easy to see why this is the case, especially after so many acquisitions. Intel plans to innovate in the memory market, AI market, graphics market and FPGA markets; expanding well beyond the world of x86 CPUs. 

In December 2019, Intel announced that they were acquiring Habana Labs, an Israeli AI startup. This came despite Intel’s existing Nervana roadmap, signalling some major changes to the company’s AI plans. 

Now, Intel has confirmed that they are making “strategic updates” to its AI roadmap, stating that it will “deliver on current customer commitments” by releasing its Intel NNP-I “Spring Hill” inference accelerator. The development of Intel’s planned Intel NNP-T “Spring Crest” accelerator has now ceased, marking an end of the company’s Nervana product stack. This information comes via WikiChip, who have discussed Intel’s plans for AI in depth. 

Let’s be clear; Intel has effectively killed off Nervana, with the company’s Habana acquisition acting as an alternative route to the company’s AI ambitions. As commented by Forbes, Intel likely received unfavourable feedback on Nervana from its engineers and customers. From Intel’s perspective, Habana has better prospects within the AI market. It is possible that Intel’s Nervana IP and software could be reused with Habana’s technology, though it is worth noting that Habana’s technology is fundamentally different to Nervana. Whether or not this was a smart move remains to be seen.