J.P. Morgan calls out AI spend, $650B required to deliver 10% return (www.tomshardware.com)

J.P. Morgan warned that the AI industry must generate roughly $650 billion in annual revenue to justify the capital companies are expected to deploy through 2030 and still deliver a 10% return. The bank framed that sum as roughly $34.72 per month from every iPhone user or $180 from every Netflix subscriber to show the scale — a useful shorthand given ~1.5B active iPhone users and ~300M paid Netflix accounts — but the burden will actually be spread across consumers, enterprises and governments. The report notes that while firms like OpenAI and Anthropic report large run‑rates or revenue targets, that hasn’t yet translated into durable profits across the sector. The significance for AI/ML is twofold: financing and demand risk. J.P. Morgan cautions the industry could repeat the telecom/fiber buildout mistake where infrastructure outpaced revenue, creating long‑lived compute overcapacity (idle, billion‑dollar data centers and GPUs). It also highlights the upside/downside dynamics: spectacular winners are likely, but so are spectacular losers if monetization and product‑market fit lag. For researchers and engineers the takeaway is clear — algorithmic efficiency (e.g., linear attention and other compute‑saving advances), careful CAPEX planning, and business‑aligned ML products matter as much as model scale, because a single breakthrough in efficiency could materially shrink compute demand and reshape the market economics.