Meta acquired the chip startup Rivos for an undisclosed amount, potentially valuing the deal at around $2 billion, to accelerate its internal AI chip development and reduce its dependency on Nvidia. The acquisition of Rivos, a company specializing in chips using the open-source RISC-V architecture, brings valuable talent and technology to Meta's custom silicon program, the Meta Training and Inference Accelerator (MTIA).
Background on the acquisition
- Need for speed: Despite having its MTIA chip program, sources close to the matter say Meta CEO Mark Zuckerberg has been dissatisfied with the pace of internal chip development. This led Meta to search for talent and expertise externally to "bring in reinforcements" and speed up the process. The deal with Rivos comes after Meta's failed attempt to acquire the South Korean chip startup FuriosaAI in March 2025.
- Rivos's technology: Rivos specializes in designing chips for both AI training and inference. It uses the RISC-V architecture, an open-standard instruction set, to develop its systems. This approach offers flexibility and avoids the licensing fees associated with proprietary architectures like those from Arm.
- Common ground: Meta's MTIA chips also utilize the RISC-V architecture, suggesting a smooth integration of Rivos's technology and expertise into Meta's existing program.
- Manufacturing plans: Rivos had previously worked with Taiwan Semiconductor Manufacturing Co. (TSMC) to produce its chips. Meta also uses TSMC for manufacturing its MTIA chips.
Strategic implications of the deal
- Reduced reliance on Nvidia: Meta spends billions of dollars annually on Nvidia's graphics processing units (GPUs) for its AI infrastructure, even while developing its own chips. By strengthening its in-house chip capabilities with Rivos, Meta can decrease this reliance, optimize its hardware for specific workloads, and achieve long-term cost savings.
- Acceleration of AI ambitions: The acquisition will bolster Meta's push for "superintelligence"—AI models that perform as well or better than humans. Access to Rivos's full-stack AI system expertise will help power Meta's generative AI products, personalized experiences, and long-term metaverse initiatives.
- Escalation of the "custom silicon war": Meta joins other tech giants like Google (with its TPUs) and Amazon (with its Trainium and Inferentia chips) in developing custom hardware. This intensifies the competition for AI hardware and could compel other companies to accelerate their own efforts.
- Validation of RISC-V: By betting on a startup centered on the RISC-V architecture, Meta further validates the open-source platform as a viable alternative for high-performance computing.
Challenges ahead for Meta
- Technical integration: Meta must successfully integrate Rivos's team, technology, and culture into its existing hardware division.
- Scaling production: Designing chips is complex and expensive. Meta will need to successfully scale up the production of its custom silicon with manufacturing partners like TSMC.
- Retaining talent: Competition for top semiconductor talent is fierce. Meta must ensure it retains the key engineers from Rivos to execute its plans.