Mirendil signed a multi-year Google Cloud partnership valued at over $100 million to secure compute resources for self-improving AI research, according to an announcement made August 6.
The deal grants Mirendil access to Google Cloud's tensor processing units and graphics processing units for large-scale model training. Google Cloud has become a primary infrastructure provider for AI labs scaling compute-intensive workloads, competing with Amazon Web Services and Microsoft Azure for long-term enterprise AI contracts.
Mirendil is an AI research firm focused on self-improving systems. The company uses TPUs and GPUs for large-scale model training and deep learning systems. Google's TPU chips are purpose-built for matrix operations central to neural network training, while GPUs offer broader flexibility across different model architectures.
Multi-year cloud partnerships have become standard practice among AI labs as training costs climb. The commitment structure locks in pricing and resource allocation for research teams planning months or years of experimentation. The $100 million valuation places Mirendil's deal among the larger publicly disclosed AI infrastructure agreements, though major tech companies routinely sign confidential contracts with undisclosed terms.
Google Cloud's AI and machine learning business has accelerated since 2023 as demand for compute infrastructure outpaced supply. The company competes directly with Microsoft, which has exclusive cloud agreements with OpenAI and other AI labs, and with custom silicon plays from Meta and xAI. Google's own AI division, DeepMind, also consumes substantial internal compute, creating a dual role as both infrastructure provider and AI developer.
Mirendil's partnership involves a commitment to use Google Cloud services for model training and inference workloads. The deal does not disclose exclusivity terms or whether Mirendil remains free to purchase compute from competing providers. Companies often negotiate hybrid arrangements with multiple cloud vendors to avoid single-vendor dependency on critical infrastructure.
Google Cloud's AI services revenue grew 26 percent year-over-year in the most recent quarter, outpacing the company's overall cloud growth rate. Multi-year enterprise agreements reduce churn and create predictable revenue streams, a priority for Google Cloud as it works to close its profitability gap with Amazon Web Services. If Mirendil's compute needs remain flat over the partnership term, the deal locks in customer spending at current levels; if research demands grow beyond contracted capacity, Mirendil would need to renegotiate or supplement with additional providers.
