OpenAI has emerged as one of the fastest-growing tech companies of all time, boasting billions in yearly revenue and hundreds of millions of users. Despite its success, OpenAI faces a difficult financial reality.. Creating and operating artificial intelligence is much more expensive than creating and operating traditional applications. The challenge is not that consumers will not use AI; rather, the cost of making AI smarter is colossal.
Current industry projections and internal estimates suggest that OpenAI will experience losses of tens of billions of dollars in the coming years. The cause is a major change in how technology grows. Unlike traditional applications, the performance of AI models does not depend on human-built code, but rather on raw computing power (which is extremely costly).
AI scaling laws are driving explosive costs
Traditional applications improve due to improved engineering. AI, on the other hand, improves by following what researchers call AI scaling laws—a mathematical relationship that states that better performance will require exponentially more computing resources, data, and energy to achieve that performance.
Reports indicate that it cost about $100 million to train the GPT-4 model. However, companies estimate that current and future frontier models will cost over $1 billion to train. Training will not remain a one-time effort; AI firms will need to continuously retrain and update their models to stay competitive.
Rising Costs vs Revenue
The cost of hardware has become an increasing burden on OpenAI. As an example, Nvidia’s advanced AI chips can run between $30,000 and $40,000 each, and a company must use tens of thousands to train large models in huge data centres that require massive amounts of power.
These chips also become obsolete at much quicker rates than traditional factory equipment, which adds to OpenAI’s need to be able to replace them frequently in order to stay competitive.
Cloud dependence and Microsoft deal add financial pressure
OpenAI has grown quickly due in part to its relationship with Microsoft. Microsoft has invested billions into OpenAI via cloud computing (through the Azure cloud platform).
Most of Microsoft’s investment, however, is made up of cloud credits, which cannot be used to pay for OpenAI’s salary, rent and other operational expenses. Therefore, in order to pay for these types of expenses continuously, OpenAI must constantly find new cash from investors.
Additionally, competition continues to heat up as companies such as Meta, Google and various startups, including Anthropic and Mistral AI, continue to build competing models. Some competitors even offer their AI models for free, placing additional pressure on OpenAI’s ability to charge for its products.
This creates a very difficult business challenge with costs continuing to rise, while OpenAI’s ability to sell products continues to be limited.
Energy, infrastructure, and regulation add new risks
The cost of AI goes beyond semiconductors, and electric power has become an expensive item for large data centres that host AI applications and consume as much energy as large metropolitan areas.
To obtain the energy needed to power these types of data centres, companies are building large scale solar farms and nuclear plants.
At the same time, regulators are scrutinizing how companies interact with each other. Agencies, like FTC and the EU, are looking into the nature of partnerships and the dominance of companies in the AI space.
The cost associated with compliance to safety regulations and legalities adds additional costs.
ALSO READ: India Forex Reserves Record High as Currency Markets Shift
OpenAI’s future depends on achieving a breakthrough
Despite these challenges, investors have continued to support OpenAI because of OpenAI long term opportunity to develop AGI, which is AI that can perform complex tasks similar to those performed by humans.
If successful at achieving AGI, these types of systems would change the way whole industries operate and generate substantial revenues.
However, OpenAI faces a harsh business reality. In contrast to traditional technology companies, AI businesses are finding that for every enhancement they deliver, they will need to build massive amounts of physical infrastructure (data centres), require vast amounts of energy, and raise a tremendous amount of capital.
The technology industry has gone through a major shift. The future of AI companies will be determined not solely by the smartest algorithms, but also by their ability to pay for the tremendous cost of operating a company that develops and deploys AI.
FOR MORE: https://civiclens.in/category/category-business-economy/
About the Author
