On Artificial Intelligence, the Skills Gap and the Challenges of Sustainability
Haebom
2
Created by
Haebom
Created at
In recent years, artificial intelligence (AI) technology has made great strides and is leading innovation in various industries. However, such AI research and development is increasingly becoming the exclusive domain of big tech companies. AI research requires enormous capital, which has led to a situation where only large companies with abundant capital can conduct AI research sustainably. In this article, we will examine the cost structure of AI research and the background of big tech companies dominating this field, and discuss the problems resulting from this and the future direction.
First, let’s look at the financials of OpenAI and Anthropic to understand why only big tech companies can sustain AI research. OpenAI has grown rapidly, reaching $340 million in annual recurring revenue (ARR) as of May 2024. However, despite this impressive revenue growth, OpenAI is expected to incur an operating loss of up to $5 billion in 2024. This is because ChatGPT alone costs about $700,000 per day, and AI training and staffing costs are also significant. OpenAI has received about $13 billion in investment from Microsoft, but these huge costs are difficult to sustain without continued financial support.
The same goes for Anthropic. Anthropic is expected to reach $3.16 billion in ARR by the end of 2024, but its projected loss cost in 2024 is expected to exceed $2.7 billion. In particular, computing costs alone are expected to reach $2.5 billion, indicating a huge amount of capital required to develop and operate AI models. Anthropic has raised $7.6 billion and is valued at $18.4 billion by 2024, but it still needs a lot of investment to operate sustainably.
In this way, AI research requires massive capital in many aspects, such as computing resources, data, and expert personnel. For example, in "Calculating the Cost of a Google Deepmind Paper," it was analyzed that the total computational cost of reproducing the experiment of the Google Deepmind paper "Scaling Exponents Across Parameterizations and Optimizers" is approximately$12.9 million (approximately 16.9 billion won) . This is due to the extensive experimental execution, such as more than 10,000 training runs for hyperparameter optimization. These experiments are mainly performed using high-performance computing resources such as TPUs, which further increases the cost.
This high cost structure makes it so that only companies with abundant capital can participate in AI research. This raises the entry barrier for AI research and creates an environment that makes it difficult for small and medium-sized companies or startups to enter the AI market. As can be seen in the cases of OpenAI and Anthropic, big tech companies continue to conduct AI research with enormous capital. This naturally results in companies with insufficient capital being excluded from AI research.
The high cost of AI research and the massive operating losses also pose major challenges to sustainability. Both OpenAI and Anthropic are reporting huge losses despite rapid revenue growth, largely due to the high infrastructure costs required to train and operate AI models. Both companies rely on continuous investment to sustain operations, resulting in complex profit sharing structures that delay long-term profitability. The AI market is also highly competitive, putting increasing pressure on companies to manage costs while innovating. There are also concerns that current pricing models may not be sufficient to cover operating costs.
In my previous article about Nvidia, I said that we should continue to buy Nvidia because there is no substitute for building this kind of infrastructure. When I looked at the blog post at that time, Nvidia was at $238. That was before the liquidation. In the future, big techs announced in their earnings that they will spend more money on CapEx. So what choice should our country make?
📚 Welcome to Haebom's archives. --- I post articles related to IT 💻, economy 💰, and humanities 🎭. If you are curious about my thoughts, perspectives or interests, please subscribe. Would you like to be notified when new articles are posted? 🔔 Yes, that means subscribe. haebom@kakao.com
