The May-25 edition mentioned a silicon renaissance. It turns out it is shopping season in Semiland with four acquisitions in two weeks. Earlier this month, Qualcomm announced the $2.4bn acquisition of Alphawave, a UK firm that specializes in chip design IP, especially silicon photonics and chiplets. AMD acquired 3 firms: Untether AI - an AI edge / inference chip designer, Brium - an AI software and compiler firm, and Enosemi - a silicon photonics and co-packaged optics firm.

‍

With Alphawave, Qualcomm hopes to offset its weakness in HPC and AI accelerators by offering an alternative to Nividia’s tightly integrated monolithic dies and NVLink. Together, Qualcomm + Alphawave could deliver a disaggregated, scalable and power efficient alternative. As for AMD, its acquisitions suggest it is trying to replicate Nividia’s integration moat (GPU, interconnect, software) via M&A focused on inference + photonics + software stack.

‍

Among the potentially threatened companies, Lambda’s current edge in offering customizable GPU infrastructure for AI researchers will erode if AMD/Qualcomm offer better software + hardware integration. Nvidia may continue to secure demand via CUDA and DGX, making it harder for Lambda to differentiate. As for EdgeQ, Qualcomm’s strength in edge + 5G, combined with Alphawave's IP, make it a formidable player. Finally, if AMD successfully integrates its recent stack, RISC-V competitors like Tenstorrent will face greater challenges in attracting adoption unless they can offer equally strong software and infra support.

‍

SiMa.ai (MLSoC) and d-Matrix (digital in-memory compute) are now in the spotlight and are likely the next targets for acquisition or strategic partnership. Companies like Tenstorrent (RISC-V inference/training chips) might attract more strategic investment as AMD and others diversify ISA and execution models. Finally, companies like Brium are gaining as compiler/software innovation becomes acquisition-worthy. Expect rising valuations and more activity around firms building alternatives to Nvidia’s CUDA ecosystem.