Cloud Courts AI Startups

Google Cloud used its Next conference in Las Vegas to show that winning the artificial intelligence race now means winning the startups building on top of it. (techcrunch.com) On April 22, TechCrunch reported that Google had set aside a new $750 million budget to help cloud partners sell artificial intelligence agents to enterprise customers. The money can cover Gemini proof-of-concept projects, forward-deployed engineers, cloud credits, and deployment rebates. (techcrunch.com) Google also used the event to spotlight startups already running on its infrastructure, including Notion, Hebbia, Safe Superintelligence, Lovable, and Anysphere’s Cursor. The pitch was not just model access but a full stack of hosting, chips, and sales support. (techcrunch.com) The clearest sign of that strategy came from Thinking Machines Lab, the startup founded by former OpenAI chief technology officer Mira Murati. TechCrunch reported that the company signed a multibillion-dollar agreement to expand its use of Google Cloud’s artificial intelligence infrastructure. (techcrunch.com) Google said the new agreement gives Thinking Machines access to A4X Max virtual machines powered by Nvidia GB300 graphics processing units, plus Google Kubernetes Engine, Spanner, Cluster Director, Cloud Storage, and Anywhere Cache. Google said the setup is meant to support frontier-model training and production workloads at scale. (googlecloudpresscorner.com) The deal is non-exclusive, according to reporting on the agreement, which means Thinking Machines can still buy computing power elsewhere. That matters in a market where large model labs routinely spread workloads across multiple cloud providers and direct chip suppliers. (techcrunch.com; siliconangle.com) Google paired that startup push with a large enterprise customer announcement. Merck said on April 22 that it would make a multi-year investment valued at up to $1 billion to use Gemini Enterprise on Google Cloud across research and development, manufacturing, and corporate functions. (merck.com) Merck said the partnership will build what it called an “agentic” ecosystem, meaning software agents that can complete multi-step work with human oversight instead of only answering prompts. Reuters reported that the spending will fund artificial intelligence infrastructure, software, and employee training over a number of years. (merck.com; msn.com) Cloud providers have spent the past year competing on more than raw computing power. They are bundling chips, proprietary models, financing, and go-to-market help as startups search for scarce capacity and large companies try to turn artificial intelligence pilots into operating systems for daily work. (techcrunch.com; merck.com; googlecloudpresscorner.com) Google’s message in Las Vegas was that it wants both ends of that market at once: early-stage model makers that need huge clusters now, and established companies willing to spend nine or ten figures to put those systems to work. (techcrunch.com; techcrunch.com; merck.com)