Do OpenAI's multibillion-dollar deals mean exuberance has got out of hand? (www.theguardian.com)

OpenAI’s recent multibillion-dollar arrangements with chipmakers Nvidia and AMD — where OpenAI pays cash for chips while Nvidia takes non‑controlling equity and OpenAI commits to buy hundreds of thousands of AMD chips with an option to acquire 10% of AMD — have intensified debate about whether AI investment has become overheated. Critics warn these deals resemble vendor‑financing (supplier funding a customer’s buy), a dynamic that inflamed the dotcom bubble, and point to runaway private valuations (OpenAI ~$500bn) and large operating losses (reported H1 revenue $4.3bn, operating loss $7.8bn) as red flags. Markets have shown extreme swings: chip and infrastructure stocks spiked on the announcements, and hyperscalers are forecast to spend roughly $325bn on capex this year, raising the risk that hardware could be outdated before returns materialize. For the AI/ML community the implications are concrete. On the positive side, guaranteed chip supply accelerates model training and deployment at scale, supporting advances and broad user growth (ChatGPT ~800M weekly users). But the deals also concentrate infrastructure control, may distort competitive dynamics, and could lock capital into specific hardware generations. Empirical adoption data complicates the hype: an MIT study found 95% of organizations see zero ROI from generative AI deployments, and McKinsey reports widespread use with limited bottom‑line impact. The net effect: faster technical progress backed by huge capital, but heightened systemic and allocation risks that practitioners and investors should monitor closely.