In Big Tech's knotty game of AI Twister, what happens when a giant slips? (www.businessinsider.com)

Big tech’s AI race has produced an odd mix of frenemies: firms are simultaneously competing to build the best models and knitting themselves into deep commercial dependencies to get the compute and talent they need fast. Recent headline deals — OpenAI’s reported $300 billion compute agreement with Oracle, Nvidia’s roughly $100 billion investment tying it to OpenAI (and plans to support at least 10 GW of datacenter capacity), Meta’s $10 billion Google Cloud pact, Microsoft reselling Anthropic models on others’ clouds, and cloud/backer relationships around Anthropic and CoreWeave — illustrate how rivals are exchanging capital, chips and cloud services even as they try to outpace one another. OpenAI is buying GPUs from Nvidia and AMD, training on Google TPUs, and exploring in‑house chips, creating overlapping supply chains and multi‑cloud footprints. That closeness accelerates AI infrastructure buildout but also creates systemic and strategic risks: vendor financing and “round‑tripping” can blur revenue quality and inflate demand, while a failure or strategic shift by one giant (or a mispriced bet like heavy buybacks of unsold capacity) could cascade across startups, cloud providers and chip suppliers. Technically, the era is marked by gigawatt‑scale datacenter deals, GPU/TPU dependency, and nascent vertical integration (in‑house silicon and proprietary data centers). The net effect: faster progress and bigger markets, but heightened concentration, financial entanglement and the potential for disruptive shocks if a major player slips.