On October 3, OpenAI announced on its official website that it had successfully raised $6.6 billion (approximately RMB 46.69 billion) in funding. After this round of funding, OpenAI's valuation exceeded $150 billion.

However, despite the significant amount of $6.6 billion, for OpenAI, it amounts to just over a year's worth of losses. OpenAI estimates that its revenue for this year could reach $3.7 billion, with losses of $5 billion.

(Source: OpenAI)

This is not even the worst moment for OpenAI's losses. According to US media outlet The Information, financial data analysis suggests that OpenAI will achieve profitability in 2029, with revenue reaching $100 billion (approximately RMB 707.48 billion). However, before then, OpenAI will continue to experience significant losses, with projected losses of $14 billion (approximately RMB 99.05 billion) in 2026.

As a global leader in AI technology, OpenAI has been favored by many investors. However, after several years, instead of turning a profit, OpenAI's losses have continued to grow. With limited liquidity, in addition to the $6.6 billion in funding, OpenAI has also raised $4 billion through a revolving credit facility from banks.

On the other hand, global internet giants are embarking on the AI journey, with various AI large models emerging. OpenAI faces an increasing number of competitors. How can OpenAI break through in this crowded field?

AI Large Models: A 'Burning Money' Game

In recent years, OpenAI has successively launched AI large models such as GPT-4/4o and the smaller GPT-4o mini, as well as introduced the o1-preview and o1-mini models, enabling them to achieve general reasoning and learn to think like humans. With technological iterations, ChatGPT has indeed become more user-friendly and has explored viable directions for other peers. For example, Bloomberg reported that Google is also developing human-like reasoning AI, attempting to catch up with the o1 model.

(Source: OpenAI)

Behind OpenAI's leadership lies significant financial investment. According to The Information, OpenAI currently bears three major costs: inference, training, and labor. Among them, labor costs are the lowest, requiring only about $1.5 billion annually. Renting Microsoft servers to power the inference costs for ChatGPT and the underlying LLM costs about $4 billion per year. Training costs, including data expenses, amount to approximately $3 billion annually.

The problem lies in the computational power required for the training and inference tasks of large models, especially for training tasks, where the required computational power scales exponentially in the later stages. Due to the increasing demand for computational power in the AI field, NVIDIA, a global GPU giant, has reached a market value of $325 billion, surpassing Microsoft to become the second-largest company in the world by market value.

Dario Amodei, CEO of OpenAI competitor Anthropic, stated that the current training cost of large AI models under development is approximately $1 billion, and this figure is expected to soar to $10 billion or even $100 billion within the next three years. The same applies to OpenAI, where future training costs for large models are projected to exceed $10 billion. As a result, The Information estimates that OpenAI's losses will reach $14 billion in 2026.

(Source: OpenAI)

Losses are not unique to OpenAI but are a common challenge faced by all enterprises in the AI industry. The industry cycle comprises four stages: start-up, growth, maturity, and decline. Although the concept of AI originated in the 1950s, it is currently transitioning from the mid to late start-up stage, with high industry risks and increasingly clear profit points, attracting more enterprises.

However, due to the nature of AI technology, the money-burning capability of large AI models may surpass that of any industry in the past. Essentially, the entire AI industry has become a money-burning game. In February this year, The Wall Street Journal reported that OpenAI CEO Sam Altman was seeking $7 trillion (approximately RMB 49.5 trillion) to reshape the global semiconductor industry and address the computational demands of the AI industry.

It is worth noting that according to National Bureau of Statistics data, China's total investment in research and development (R&D) exceeded RMB 3.3 trillion in 2023, which is only one-fifteenth of the funds required by OpenAI. Therefore, Altman's request is highly unlikely to be fulfilled.

Additionally, The New York Times revealed that if OpenAI fails to transform into a profitable enterprise within two years, all investments in it will be converted into debt. Rapid transformation and profitability are top priorities for OpenAI.

Internal and External Challenges: Is OpenAI Worried About Survival?

ChatGPT's emergence sparked an AI craze globally, with countless enterprises jumping on the bandwagon. According to Backlinko's June 2024 statistics, ChatGPT accounts for 60% of the monthly web traffic among the top 50 most commonly used generative AI web products, firmly establishing its position as the industry leader. Despite significant external pressure, OpenAI's position remains unshaken in the short term. However, OpenAI faces more pressing internal issues.

In February and August this year, Tesla CEO Elon Musk twice sued OpenAI and its CEO Sam Altman, repeatedly mocking him on the X platform for only engaging in internal strife. Musk's hostility towards OpenAI and Altman stems from his status as a co-founder of OpenAI.

(Source: X Platform Screenshot)

OpenAI originally had eleven founders. Today, besides Altman, only Wojciech Zaremba, the head of the language and code generation team, remains at OpenAI. Insider revelations also suggest that after Chief Scientist Ilya Sutskever left in May, OpenAI prepared for potential closure. The departure of core technical personnel and the underperformance of new hires often precede corporate collapse. Perhaps due to these pressures, OpenAI is eager to transform into a for-profit company and plans to raise more funds through an IPO.

Insights into OpenAI's development projects also reveal some clues. In February 2024, OpenAI officially launched Sora, a large text-to-video model. Many believe that the era of mass video creation, where everyone can be a director and screenwriter, is approaching. However, to date, Kuaishou's text-to-video model Keling has been online for several months, while Sora has yet to release a public beta version. Even Sora team leader Tim Brooks has jumped ship to DeepMind, a subsidiary of Google, raising the possibility of Sora's cancellation.

(Source: OpenAI)

Amidst increasing external pressure and severe internal talent loss, OpenAI urgently needs to open source and cut costs.

In both To B and To C scenarios, OpenAI excels globally and has forged partnerships with leading global internet companies such as Microsoft and Apple. In the future, OpenAI must collaborate with its partners to explore more usage scenarios and integrate large AI models into everyday life.

Cost-cutting efforts should focus on small models. According to AIGCRank statistics, OpenAI's large GPT-4o model is priced at $5 per million input tokens and $15 per million output tokens. In contrast, GPT-4o mini is priced at 15 cents per million input tokens and 60 cents per million output tokens, representing a significant economic difference.

Under the premise of meeting individual or corporate needs, sacrificing some performance to reduce usage costs and making AI large models affordable and desirable for everyone is crucial for their widespread adoption. The development of small models has gradually become an industry consensus, with executives from Alibaba, Google, Baidu, and other internet companies advocating for them, arguing that fine-tuned small models can offer comparable experiences to large models in specific scenarios.

Building upon GPT-4o mini, OpenAI can also explore developing end-side large models with fewer parameters, shifting the resources and costs required for inference tasks to local environments. By pursuing both open-sourcing and cost-cutting initiatives, OpenAI can offer investors a glimmer of hope for profitability, despite the Severe situation it currently faces.

Recently, AI luminary Geoffrey E. Hinton's receipt of the 2024 Nobel Prize in Physics sparked heated debates. During a press event, Hinton publicly criticized OpenAI CEO Sam Altman, stating, "One of my students fired Sam Altman," and expressing pride in this action. Hinton admitted that when OpenAI was founded, its primary goals were to develop general AI and ensure its safety. However, over time, it has become evident that Altman is more concerned with profits than safety.

From OpenAI's perspective, it is understandable that Sam Altman currently prioritizes profits, as an entrepreneur responsible for the company's survival. While Hinton is correct, Altman is also constrained by circumstances.

In fact, the setbacks currently faced by OpenAI are inevitable in the industry's development trajectory. Many companies fare worse than OpenAI. For instance, Anthropic's total costs for this year are projected to reach $2.7 billion, with revenues of only $0.4 to $0.6 billion, approximately one-tenth of OpenAI's.

Unlike blockchain and the metaverse, AI's impact on daily life is tangible. I frequently use AI to refine articles and generate images for my work, and my programmer friends rely on ChatGPT as a valuable work assistant. As technology continues to advance, the AI industry may undergo some changes, but OpenAI has a strong chance of emerging as one of the dominant players.

Setbacks Are But Refinements; OpenAI's Future Is Promising

Annual losses of nearly RMB 100 billion undoubtedly hang like the sword of Damocles over OpenAI, forcing it to continuously raise funds and accelerate its transformation. These pressures also serve as a driving force for OpenAI's continuous development, urging it to enhance ChatGPT's capabilities.

From an investor's perspective, while short-term gains are important, long-term returns are even more enticing. Investors are aware of OpenAI's temporary losses but are focused on its potential to deliver substantial benefits in the future. Each GPT iteration has led the industry, and OpenAI's partnerships with giants like Apple and Microsoft, coupled with the rapid growth of the AI industry, have infiltrated various sectors. Amidst this prosperity, investors are willing to continue funding OpenAI in anticipation of future returns.

(Source: OpenAI)

As the industry leader, OpenAI may encounter setbacks but will not lack investment. If it can transform into a for-profit enterprise within two years and complete an IPO, it is likely to attract even more funding. Simultaneously, OpenAI's revenue-generating capabilities are continually improving, with projected significant reductions in losses after the peak in 2026.

The coming years represent a period of growing pains for OpenAI's development. With strong revenue capabilities and support from investors like Microsoft, Thrive Capital, NVIDIA, and others, OpenAI can navigate the difficult times ahead, provided it abandons unrealistic plans to reshape the global semiconductor industry.

By 2030, the AI industry is expected to enter its growth stage, with an increasing number of high-quality AI large model products available. Competition among enterprises will drive improvements in AI tool experiences and price reductions, ultimately benefiting consumers with AI's convenience.

Similar to the development of the new energy vehicle industry, many enterprises may falter due to inadequate financial reserves or lack of investment. Only those that overcome significant obstacles will emerge victorious. Currently leading the AI industry, OpenAI has the potential to become its equivalent of Tesla.

Source: Leitech