Cheaper Chinese and open source AI models threaten infrastructure spending returns
Jefferies warns that falling prices for AI models from Chinese developers and open source projects could force a major reset in expectations for the AI industry. The cheaper competition may limit revenue for OpenAI and Anthropic, raising doubts about whether cloud providers can earn returns on the vast sums they are spending on data centres.
Chinese models processed 76.12 trillion tokens in the week ending 27 September, compared with 14.2 trillion for the four largest US providers. Jefferies says roughly half of the $2.3 trillion in contracted backlog at Microsoft, Alphabet, Amazon and Oracle is linked to OpenAI and Anthropic, while the hyperscalers plan to spend $600 billion to $800 billion on data centres this year. Separately, Bain estimates the AI industry would need $6 trillion in annual revenue by 2031 to fund expected infrastructure demand, leaving a substantial gap even if new AI applications emerge.
- Cheaper models are putting pressure on AI pricing.
- Hyperscalers face questions over returns on data centre spending.
- Bain says AI needs $6 trillion in yearly revenue by 2031.
New here? Start with this
AI models are the software systems behind tools that can generate text, images and code. They run on data centres filled with specialised computers, and companies such as Microsoft, Google, Amazon and Oracle provide much of the computing capacity, while OpenAI and Anthropic develop models and services that use it.
Some Chinese developers and open source projects offer models at lower prices. Open source models can be used and adapted by others, which may increase competition and put pressure on what model makers can charge. Cloud providers are investing heavily in data centres and rely on demand from AI companies and their customers to earn returns.
The central question is whether AI services will bring in enough money to justify the cost of building and running that infrastructure. Estimates of the revenue needed depend on how quickly useful applications spread and how much customers are willing to pay.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Cheaper Chinese and open-source models could make AI useful to more businesses and consumers, accelerating adoption and creating demand for cloud computing and new applications. If lower prices expand the market, cloud providers may still earn returns through greater usage, while competition could encourage more efficient infrastructure investment.
The case against
Falling model prices could make it harder for OpenAI and Anthropic to generate enough revenue to support the large cloud commitments linked to them. If demand and new applications do not grow enough to offset lower prices, hyperscalers may struggle to earn adequate returns on their substantial data-centre spending.
Read the full article at the source →
Originally published by The Register as “Cheap Chinese and open source LLMs could trigger trillion-dollar AI infrastructure ‘reset’”.