Open AI, which is currently planning its eighth round of funding, has painted a bright future for investors:

According to documents disclosed by The New York Times, both active users and revenue of Open AI have maintained rapid growth. As of August this year, its monthly revenue reached USD 300 million, marking a 1700% increase from the beginning of the year. In June, its monthly active users hit 350 million, more than triple that of three months ago.

It is projected that Open AI's annual revenue will be approximately USD 3.7 billion in 2024, growing to USD 11.6 billion in 2025, and reaching USD 100 billion by 2029.

If the funding round is successfully completed, the valuation of this third-largest unicorn company globally will reach USD 150 billion.

However, behind the soaring valuation, this leading startup, which brings together top talents, is casting a shadow. Chief Technology Officer Mira and Chief Scientist Ilya have successively left, and President Brockman was "forced" to take a leave of absence, leaving Sam Altman as the sole member of Open AI's core four-person leadership team.

Produced by Guangzhui Intelligence. Please indicate the source when reprinting.

Transitioning from a non-profit to a for-profit entity, Open AI is destined to turn its ship amidst question and complete its transformation.

Why are top talents leaving frequently?

"I made a difficult decision to stay at Open AI," an Open AI employee lamented on platform X last Wednesday.

The succession of personnel upheavals at Open AI is shaking the confidence of its employees.

Following two waves of departures in November last year and May this year, Open AI had just launched its GPT-4o model in full force, but last Friday witnessed the departures of Chief Technology Officer Mira Murati, Chief Research Officer Bob McGrew, and Vice President of Research Barret Zoph. In the past two years, over 30 employees have left Open AI, with only 3 of the original 11 founding team members remaining.

"I decided to leave to create time and space for my own exploration," Mira explained her reason for departure.

Former CTO of Open AI, Mira Murati

Although the surface appears peaceful, people's doubts about Open AI persist. Setting aside issues like 996 work culture and employee benefits, Open AI faces two major controversies that hang like a sword of Damocles over the team:

First is the concern for safety alignment.

'Safety culture and processes have given way to shinier products,' remarked Jan Leike, a scientist from Open AI's safety alignment team, upon announcing his departure in May. In a series of tweets, Leike expressed his dissatisfaction with Open AI's disregard for safety issues.

It appears that Open AI's neglect of its safety team persists and has gradually become a public controversy:

Last Wednesday, Open AI once again staged a 'sniper attack' on its powerful rival Google: On the same day that Google unveiled its Gemini family, Open AI officially launched its previously small-scale tested product, the voice version of 'GPT-4o,' allowing Plus and subscription members to use it. Unsurprisingly, Open AI successfully stole the spotlight from its competitor.

Yet behind this precisely timed and successful marketing campaign, Open AI's safety alignment team suffered immensely.

According to The Wall Street Journal, the launch of GPT-4o was a spur-of-the-moment 'benchmarking action.' Within the nine days leading up to the launch, Open AI's safety team was forced to rush through testing, working up to 20 hours a day.

Even so, after the full launch of GPT-4o, new issues emerged during internal evaluations. Insiders revealed that subsequent analysis found the model's persuasive abilities exceeded OpenAI's internal standards—i.e., the ability to create content that convinces people to change their beliefs and engage in potentially dangerous or illegal activities.

Moreover, Open AI's decision to turn the ship and embrace profitability has unsettled its employees.

According to The Wall Street Journal, before leaving, co-founder John expressed his concerns to colleagues: Key member Ilya had already departed, and Open AI's shift towards profitability made him question the company's purity.

Open AI is transforming into an early-stage DeepMind, nurturing talent for the industry. Currently, some former Open AI employees have chosen to start their own ventures, while others have joined Open AI's formidable competitors. Statistics show that multiple employees have joined Anthropic and Google.

After core team members like Jan Leike and John Schulman joined Anthropic, the latter, which has long ranked second, is gradually catching up to and even surpassing Open AI.

Three months ago, Anthropic introduced the Claude 3.5 Sonnet model. According to Anthropic's official website, Claude scored higher than GPT-4o in metrics such as programming, mathematics, and visual understanding.

Claude 3.5 Sonnet outperforms GPT-4o in multiple ability metrics. Source: Anthropic

A Necessary Transformation

Open AI is a top-tier tech company that is also a major cash-burner.

To date, Open AI has completed seven funding rounds, raising over USD 11.5 billion. Microsoft has consistently been Open AI's largest financial backer, participating in three funding rounds, including one that injected USD 10 billion, setting a record for the largest AI funding round.

However, Open AI's cash-burning pace is also accelerating. Even with an estimated annual revenue of USD 3.7 billion, Open AI remains a voracious 'gold-guzzler'.

First, Open AI's workforce has rapidly expanded from 770 employees last year to 1,700 this year.

Additionally, in terms of model training, from GPT-1 to GPT-4, as model parameters have gradually expanded from hundreds of millions to tens of billions, Open AI's training costs have snowballed. In public forums, Sam Altman has shared that the training cost of GPT-4 exceeded USD 100 million.

Amidst the accumulation of high R&D and training costs, Open AI's revenue remains paltry.

According to recent internal OpenAI documents disclosed by The New York Times, insiders estimate that losses, excluding equity and compensation, will still exceed USD 5 billion.

As early as July this year, foreign media outlets made a grim prediction—if the current pace continues, Open AI's cash reserves will be depleted within the next year.

With funds running low, Open AI has chosen to initiate another funding round.

Last week, Open AI announced its new funding plan, aiming to raise USD 6.5 billion, surpassing the USD 6 billion raised by Elon Musk's AI platform, xAI, this year. Based on current information, this round is expected to involve Microsoft, NVIDIA, Thrive Capital, Tiger Global, Khosla Ventures, and the Abu Dhabi sovereign wealth fund.

If this funding round is successfully completed, Open AI's valuation will exceed USD 150 billion, maintaining its position as the world's third-most valuable unicorn company, after ByteDance and SpaceX.

Top 5 on the Hurun Global Unicorn Index 2024, with Open AI ranking third. Source: Hurun Report

However, facing a tech company that consistently burns cash with limited economic returns for the time being, investor sentiment is starting to waver. Apple has already made its stance clear. Last Sunday, foreign media reported that Apple has opted out of this funding round.

Investor uncertainty has intensified doubts about Open AI. After multiple funding rounds, Open AI faces two expectations—it must maintain its global leadership and present investors with a more certain future.

Moreover, the agreement previously signed between Open AI and investors already planted the urgency of profitability. According to Reuters, if OpenAI insists on its non-profit plan, investors will have the right to demand a refund of their previous investments. Currently, Open AI is developing a plan to restructure its core business into a for-profit company that will no longer be controlled by its non-profit board.

Under immense pressure, Sam Altman, Open AI's sole remaining core member, has had no choice but to turn the ship, transitioning from a purely research-oriented company to one pursuing profitability.

From a commercial perspective, Open AI's most lucrative business remains its C-end subscription service. With the rapid expansion of its user base, Open AI's revenue has also increased. It is projected that Open AI's revenue will reach USD 3.7 billion in 2024.

The New York Times disclosed a document sent by Open AI to investors, revealing that as of August, Open AI's monthly revenue reached USD 300 million, marking a 1700% increase from the beginning of the year. As of June, its monthly active users hit 350 million, more than triple that of three months ago.

While specific revenue sources were not disclosed, based on the reported 11 million subscription users, each contributing USD 20 per month, ChatGPT's monthly subscription revenue reaches USD 200 million, which constitutes Open AI's primary source of income.

The remaining revenue primarily comes from third-party developers and corporate partnerships. The document reveals that over 1 million third-party collaborators currently use Open AI's technology.

Open AI is not satisfied with its current pace of monetization. In the document, Open AI mentions specific plans and projections for the future, such as increasing subscription prices, raising ChatGPT's monthly subscription fee from USD 20 to USD 44. It expects revenue to reach USD 11.6 billion by 2025.

However, with the subscription price increase, user willingness to pay and the growth in subscriber numbers remain unknowns. Nevertheless, Open AI is destined to endure this prolonged period of pain.

The Next DeepMind?

A decade ago, two prominent figures in the tech world—Elon Musk and Sam Altman—gathered at a cocktail party and discussed the then-thriving AI company, DeepMind.

Sam Altman (left) and Elon Musk (right)

That year, DeepMind was acquired by internet giant Google, which remains a leading player in the AI field. Both individuals were concerned that smarter-than-human machines, if created by profit-driven companies, could be dangerous.

Sharing the same insight, the two joined forces to establish a non-profit AI company, and Open AI was born.

Two years after its acquisition by Google, DeepMind's AlphaGo shone on the world chessboard, defeating world Go champion Lee Sedol 4-1, garnering widespread attention.

On the path towards AGI, the acquired DeepMind did not have to worry about funding sources.

AlphaGo's creation was built upon Google's vast financial resources. Although DeepMind did not disclose relevant cost information, during its eight years with Google, from 2014 to 2019, it incurred losses totaling GBP 1.435 billion (approximately USD 13.4 billion).

At that time, Open AI, adhering to its non-profit approach, had to find funding on its own to advance towards AGI. After Elon Musk left Open AI and revoked his USD 1 billion funding commitment, Sam Altman found Open AI's largest financial backer—Microsoft, which filled the USD 1 billion gap.

Without the need to worry about profitability, DeepMind also lost its freedom to make independent decisions.

Before joining Google, DeepMind's founders sought to remain neutral and requested that Google sign an 'Ethics and Safety Review Agreement,' which mandated that DeepMind's core technologies undergo ethical review by a committee. However, after one meeting, this committee seemingly ceased to exercise its authority.

According to foreign media reports, after joining Google, DeepMind sought independence multiple times internally but was unsuccessful. During this phase, the once united DeepMind began to lose a significant number of employees, earning the nickname 'Huangpu Military Academy' in the AI community, as some departing employees subsequently founded various startups.

Ultimately, last year, DeepMind officially merged with Google's internal team, Google Brain, to form Google DeepMind. The merged department is jointly refining the Gemini large model, supporting Google's plan to catch up with Open AI.

After developing ChatGPT, Open AI's choices began to diverge from its past—adopting closed-source technology, prioritizing products, and pursuing profitability.

Open AI's ambitions are no longer limited to exploring cutting-edge technology. Insiders comment that Sam is now focused on product development.

Since the previous year, Open AI has expanded beyond GPT1-4 series development, with plans for 'shiny products' already on the agenda. Sam Altman is attempting to transform Open AI into a 'product-oriented' company. ChatGPT marked a successful debut, becoming the fastest product to reach one million users worldwide in just five days.

In the context where the exploration of large model technology routes takes time, Open AI has been constantly active on the product side, having successively launched products such as GPTs Store, AI search engine, and end-to-end model GPT-4o.

In terms of product promotion strategy, Open AI's method of sniping at Google is simple yet effective—by releasing new products just before or on the same day as competitors' key launches.

Over the past six months, Open AI has been fiercely competitive with Google in terms of product launches: first, it released the text-to-video model Sora on the same day as the launch of Gemini 1.5 Pro in February, and then released the end-to-end model GPT-4o the day before Google's annual developer conference, stealing Google's thunder twice in a row.

However, in its efforts to compete with Google, Open AI has also paid a price. Although it has not been acquired by Microsoft, Open AI's transformation into a commercial organization has led to internal conflicts and talent loss.

DeepMind, which was acquired, and Open AI, which independently transitioned into a for-profit organization, have ultimately not escaped the pressure of funding.

But for Open AI, which is at least maintaining independent operations for now, will its fate be better than that of DeepMind?