Meta is set to acquire chip startup Rivos with the aim of accelerating its autonomous chip development and ultimately reduce its reliance on Nvidia’s GPUs.
The social media giant will be buying Rivos, a Santra Clara-based chip startup known for building and designing chips on the RISC-V architecture, an open-source set and alternative to chip designs from Arm, Intel, and AMD.
This architecture allows Rivos to develop both custom CPUs and GPUs accelerators that are optimized for large-scale AI training and inference. Its programmable, energy-efficient server solutions also support today’s LLM demands and future AI innovations.
Meta’s acquisition of Rivos also adds to the company’s growing portfolio of custom chip projects designed to optimize large-scale AI training and inference workloads. An example is the social media giant’s existing Meta Training and Inference Accelerator (MTIA) chips, which are designed and engineered specifically for AI tasks.
For Meta, the rationale behind this acquisition is to reduce the overreliance on Nvidia as its default GPU provider.
By acquiring Rivos, Meta gains more control over hardware design, which may enable it to tailor and design chips specifically for its workload demands, potentially reducing costs and improving efficiency. This strategy also offers Meta a leeway by diversifying away from a supply chain that is heavily dominated by Nvidia.
Why This Acquisition Matters Beyond Meta Itself
Meta’s acquisition of Rivos highlights a growing trend among tech giants that own and customize their AI infrastructure end-to-end. Companies like OpenAI, Google and Amazon, striving for performance gains and independence from third-party providers, have continued investing in designing proprietary chips that are tailored to their AI platforms.
Meta’s acquisition in its bid to stand alone has also followed closely in this competitive AI arms race, signaling that it aims to stand shoulder to shoulder with other tech giants by further developing its own chips for the company’s Superintelligence Labs.
From a broader perspective, the AI hardware race is speeding up, where it is driven by explosive demand for large-scale training and inference capacity. Building in-house chips helps Meta reduce its exposure to the volatility and supply constraints that has riddled the mass-market for GPUs.
As such, the acquisition may help Meta further align its hardware development intimately with its software stack and AI models, which can improve the performance of tasks being rendered by its GenAI portfolio.
However, integrating Rivos into Meta’s existing AI chip division comes with unique challenges. Ensuring seamless compatibility between Rivos’ open-hardware designs with the MTIA program, managing manufacturing scale-up with other partners like the Taiwan Semiconductor Manufacturing Company (TSMC), and retaining important AI engineers amid a competitive semiconductor and AI labor market will be high priorities for the social media company.
The acquisition also raises the question of how it will quickly contribute to the reshaping and restructuring of a market that is traditionally dominated by Nvidia.
Still, if successful, this acquisition could cut Meta’s AI infrastructure costs by billions annually and further free up more capital for research and development.
