Intel buys deep-learning startup Vertex.AI to join its Movidius unit; Vertex says Intel will continue to develop its PlaidML engine as an open source project
Intel wants to bring more artificial intelligence technology into all aspects of its business, and today it is stepping up its game a little in the area with an acquisition.
Context & Ripple Effects
Vertex.AI is the third piece in Intel's two-year AI acquisition run: after buying deep-learning company Nervana Systems and then Movidius with its drone and VR vision chips, Intel had detailed an AI chip vision that EE Times noted remained incomplete until the Movidius deal closed. Folding Vertex into Movidius adds the missing software layer — the PlaidML engine that compiles deep learning onto hardware — directly on top of the vision silicon.
The commitment to keep PlaidML open source matters as much as the buy itself: it lets Intel court developers who won't lock into proprietary stacks, a playbook it extended two years later when it acquired Cnvrg.io's data-science platform while promising independence.
First-order effects
- The Vertex team joins Movidius, giving the vision-chip unit an in-house deep-learning engine instead of relying on external frameworks to make its hardware usable by ML developers.
- Existing PlaidML users get continuity rather than a shutdown — Intel inherits an open-source community and its contributor base along with the code.
Second-order effects
- An open-source compiler tuned for Intel silicon lowers the switching cost of building on Movidius and Nervana hardware, pressuring rival accelerator vendors whose stacks are closed to match on developer accessibility.
- Each acquisition narrows what Intel still has to license or partner for, shifting spend from external framework vendors toward internal integration across the units it has bought.
Third-order effects
- If the Nervana-to-Movidius-to-Vertex sequence holds, Intel is assembling a vertically integrated AI offering — chips, compilers, and eventually data-science tooling — where the moat is the stack, not any single product.
- Open-sourcing the software layer while owning the hardware beneath it becomes a template for how large chipmakers absorb startups without scaring off the developer ecosystems those startups depend on.
The trend: Chipmakers are acquiring their way down the AI stack from silicon into compilers and developer tools, using open source to hold the software layer while they own the hardware.