OpenAI President Greg Brockman predicted that computing capacity constraints will persist in the AI industry during a forum held in New York City on the 23rd. Brockman stated that companies face difficult decisions on what to train models with and how much to scale products due to ongoing computing shortages. AI firms including OpenAI and Anthropic, along with big tech companies like Microsoft, Amazon, and Meta, are struggling to meet surging AI demand while simultaneously training new models and serving customers, requiring massive computational equipment such as NVIDIA GPUs and continued heavy datacenter spending.
According to Yahoo Finance on the 25th (local time), Brockman said at the forum on the 23rd: "My strong prediction is that no matter what happens, we will continue to remain in this computing shortage phenomenon. Right now we have to make difficult decisions about what we actually train models with and how much we scale products."
Big Tech Companies Struggle With AI Computing Demand
AI companies like OpenAI and Anthropic, as well as big tech firms including Microsoft, Amazon, and Meta, are experiencing difficulties meeting exploding AI demand while balancing new AI model training and customer service provision. This requires large-scale computing equipment such as NVIDIA GPUs. Despite continued massive spending on datacenters, companies struggle to meet demand.
Brockman explained: "Occasionally, seeing OpenAI's capacity grow rapidly, people think there will be a surplus of computing capacity, but as the company develops new models and new usage methods, demand increases further." He added: "I think this situation will continue going forward because we are transitioning to a compute-powered economy."
Alphabet Announces Third-Party Capacity Contracts and Google-SpaceX Deal
At Alphabet's Q2 earnings announcement on the 22nd, the company's Chief Financial Officer announced that despite the entire industry making large-scale spending on datacenters, the company will contract with third-party vendors to meet its computing capacity in Q3.
Google signed an AI capacity lease contract worth $920 million per month with SpaceX in June.
Brockman Rejects Token Maxing Concept
Brockman dismissed the concept of 'token maxing' - using AI computing to maximum capacity purely for show-off purposes - noting that customers are beginning to see return on investment as they utilize software to improve productivity.
He emphasized: "If customers use AI well, they will start to see return on investment. If you just maximize token usage, you only get maximum token usage, but if you maximize value, you can apply these technologies to solve problems and obtain actual benefits."
FAQ
What did OpenAI's Greg Brockman predict about AI computing capacity?
Greg Brockman predicted at a New York City forum on the 23rd that computing capacity constraints will persist in the AI industry. He stated that companies will continue to face difficult decisions on model training and product scaling due to ongoing computing shortages.
Why did Google sign a contract with SpaceX for AI capacity?
Google signed a $920 million per month AI capacity lease contract with SpaceX in June. This followed Alphabet's CFO announcement on the 22nd that the company would contract with third-party vendors to meet its Q3 computing capacity needs, despite industry-wide massive datacenter spending.
What is token maxing and why did Brockman dismiss it?
Token maxing refers to using AI computing to maximum capacity purely for show-off purposes. Brockman dismissed this concept, stating that customers should focus on maximizing value rather than token usage. He emphasized that applying AI technologies to solve problems and obtain actual benefits delivers real return on investment, not just maximum token consumption.
