Google's $17 Billion on AI, Intel vs Nvidia, Meta's AI App, Lightrun Raises $70 Million, & DeepMind's Music AI
Google spending $17 billion on AI "supply constraints" not in the numbers, Intel's building homegrown AI chips, Meta launched an AI app, Lightrun raised $70 million, and DeepMind expands music AI
Google Spends $17 Billion on Cloud AI, Growth is Not Reflecting “Capacity Constraints”
Google invested over $17 billion in servers and data centers to enhance their AI and cloud services, announced during their Q1 2025 earnings call.
The upgrades support faster model training and inference for enterprise clients, positioning Google Cloud as the vertically integrated leader in AI.
Google justified the investment by claiming they are still capacity constrained, but as Ben Thompson noted, it seems like much of Google’s growth came from Workspace price increases, not from a surge in Cloud demand.
Despite the one-time Workspace price increases, cloud growth slowed from 30% to 28% YoY, while margins held steady. If it wasn’t for these price increases, margins and growth would have both been far lower.
Slowing Google Cloud revenue has been a theme over the past 3 quarters. Which begs the question: if revenue is slowing down, where are the supply constraints Google is claiming?
Breaking out Workspace would answer this question directly. Google Cloud growing >30% over the next two quarters would also validate their capacity claims.
Intel Challenges Nvidia with Homegrown AI Chips
Intel is developing a new AI processor to compete with Nvidia’s, following years of missed opportunities. Set for testing in 2025, the chip targets data centers and generative AI workloads, which would reduce reliance on Nvidia’s costly GPUs.
Intel’s success could lower AI deployment costs, especially for cloud providers. Intel’s integrated manufacturing gives it an edge in supply chain control, but catching up to Nvidia’s performance remains uncertain.
Even if Intel catches up in performance, Nvidia has a dominant position with their CUDA platform, which is what developers use for building applications on GPUs.
Ultimately, the question is whether Intel’s performance or cost is worth switching from CUDA to a new platform. That switch is already being forced on some developers with chip export controls.
If fewer developers build on CUDA, Nvidia’s network effect dissipates, giving competitors like Intel more of an opportunity.
Meta Launches Standalone AI App Powered by Llama 4
Meta debuted a standalone AI app powered by their Llama 4 model. The app targets consumers with features like content creation and virtual assistance, competing with ChatGPT.
For developers, Meta’s new LlamaCon event and API previews signal stronger support for building on Llama 4, potentially expanding AI-driven apps. The app’s free access could disrupt the market, and could attract users who don’t pay for ChatGPT.
With nearly a billion users, will enough ChatGPT free users move over to Meta’s offering? Or is it too little, too late for seizing this market?
Lightrun Secures $70M to Fix AI-Generated Code Bugs
Lightrun raised $70 million to enhance its AI debugging platform, which identifies and resolves bugs in AI-generated code. The tool targets developers working with generative AI models, streamlining workflows for coding projects.
For software firms, this could reduce development time and improve code reliability, a critical need as AI-generated code grows common. The funding signals strong demand for AI-adjacent tools, and with Cursor crossing $200 million ARR, a complimentary tool like Lightrun could see strong demand as well.
DeepMind Expands Music AI Sandbox for Creative Applications
Google’s DeepMind announced enhancements to their Music AI Sandbox, enabling musicians to generate and refine compositions. The platform creates melodies and harmonies with generative AI, offering tools for both amateurs and professionals.
For the music industry, this could lower barriers to entry, fostering creativity and reducing production costs. However, concerns arise over copyright, as AI-generated works may inadvertently replicate existing songs, risking legal disputes.
The expansion highlights AI’s growing role in creative fields, but balancing innovation with intellectual property protections remains a challenge.
too much expense to develop junk AI = m economic collapse, economic recession (example 2008).