Nearly two years after OpenAI launched ChatGPT, tech and business leaders are debating whether AI is … [+]
For all its promise to transform industries and change how we live and work, there’s still debate in tech and business circles around whether AI’s financial impact matches its ambitious claims. As is with most technological innovations, the argument is polarized.
On one hand are the business leaders certain that we are in an AI bubble, much like the dot-com bubble of the 1990s and its subsequent market crash in 2000. Baidu CEO Robin Li echoed this exact sentiment at the Harvard Business Review’s Future of Business conference.
“I think like many other technology waves, bubbles are kind of inevitable when you pass the stage of initial excitement. People would be disappointed that the technology doesn’t meet the high expectations generated through the initial excitement.”
Li predicted that once this hype fades, only about 1% of companies will endure, ultimately achieving significant growth and creating substantial, lasting value.
On the other hand, venture investors like OurCrowd CEO Jon Medved — managing $2.3 billion in assets and over 460 portfolio companies — see real financial promise in AI. Medved argues that AI, like the dot-com era, may indeed be experiencing inflated valuations, but the underlying business potential is genuine.
“Yes, valuations during the dot-com era were ‘bubblelicious’ but the business opportunity was real indeed and huge companies like Amazon, Google, and so many more were built as a result,” Medved told me.
This contrast sets the stage for examining AI’s true business impact today.
State of the AI Market
The AI market in 2024 is valued at an estimated $214 billion in revenue and is expected to continue growing at a double-digit rate over the coming years, with a projection of $1,339 billion by 2030. Such estimations and projections are backed by tangible metrics like expanding enterprise budgets for AI tools, rising investments in AI startups and the increasing number of AI-powered applications being embedded in nearly every business function across every sector known to man.
In 2023 alone, privately owned AI startups raised nearly $50 billion in funding and have taken a large chunk of startup funding for this year already, according to Crunchbase. This “AI gold rush,” as Forbes’ Phoebe Liu called it earlier this year, has made many tech giants even richer and pushed “more than a dozen new AI billionaires on to Forbes’ World’s Billionaires List.”
But while funding is great, how exactly do AI companies make money today?
AI ROI Drivers
When it comes to return on investment for AI, hardware and infrastructure stand out. Companies like Nvidia have capitalized on the demand for advanced processors and computing devices needed to support AI algorithms — specifically graphic processing units, which sell for as much as $40,000 per unit, according to Nvidia CEO Jensen Huang. These chips power some of the greatest AI innovations today, leading to a possible $5 trillion valuation of the hardware company.
There are two prominent examples of how this translates to revenue from the end user. One is offering AI-as-a-service, allowing businesses to access AI capabilities like image recognition, predictive analytics, language processing and more via application programming interfaces without building proprietary infrastructure from scratch.
Case in point is OpenAI’s AIaaS offering, which is available through Microsoft Azure OpenAI Service and gives users access to OpenAI’s language models, including GPT-4, GPT-3-5-Turbo and others. Through the APIs provided, developers can deploy the GPT-4 model to data and adapt it to a variety of tasks. This revenue stream operates a pay-as-you-go payment model, which only charges for used resources.
OpenAI’s most advanced model, GPT-4o, currently costs from $2.50 per 1 million input tokens to $10 per 1 million output tokens — tokens being pieces of words used for natural language processing. With over a million third-party developers using this technology to power their own services, the San Francisco startup expects to make $1 billion from this revenue stream, according to a New York Times’ review of recent OpenAI financial documents.
Consider also Google Cloud AI, another suite of AIaaS products and services that businesses can deploy for specific needs. Some of those solutions include pre-trained machine learning models for image processing, preconfigured chat platforms using retrieval-augmented generation (known more widely as RAG), and a preconfigured solution for extracting text and summarizing large documents. These offerings, among other things, have contributed significantly to the 35% increase in cloud revenue Alphabet reported for the third quarter of 2024.
Beyond AIaaS, another revenue driver lies in AI-enabled applications — software that incorporates AI for specific use cases. These applications pair AI models with custom industry data or language models to provide solutions tailored to niche needs. This is the umbrella under which the famous ChatGPT falls.
OpenAI hit $300 million in monthly revenue this August and it mostly came from ChatGPT. For this chatbot, OpenAI runs a flat-rate subscription model for its Plus, Team and Enterprise packages, starting from $20. But only about 10 million of its 350 million monthly users (just over 2%) are subscribed to this service.
Still, the company’s revenue projection from ChatGPT is $2.7 billion this year, nearly a 300% increase from last year’s $700 million.
Then there is Notion AI, which charges an extra $8 monthly to access its AI features that help users summarize notes and identify action items from a meeting. This integration got the productivity app a spot on Forbes’ sixth annual AI 50 list. Still on this AI 50 list are applications like automated medical documentation app, Abridge, worth $850 million; Harvey, AI assistant for law, tax and finance professionals valued at $1.5 billion; and Synthesia, AI avatar and video generator with over a $1 billion valuation.
However, even with all the head-swinging numbers, the costs of running these platforms are substantial. The research and development costs of Nvidia’s Blackwell chip, for example, were about $10 billion, according to Huang.
Getting the Right ROI on AI
As the needs for AI chips and AI platforms increase, their associated costs will likely rise, pushing companies to seek new revenue models or even adjust pricing to cover operational demands. This is true especially for AI-enabled applications that directly interact with end users and need to keep pricing within reach to maintain a solid user base. They face the unique challenge of balancing affordability for customers with the high costs associated with AI processing and data management.
“We’re still in the early days of monetizing AI-enabled applications, as folks are trying to figure out which AI-enabled workflows create tangible business value for customers. Regardless of how the AI-enabled apps monetize, they’re often licensing AI-as-a-Service and paying real costs for it (along with paying real costs for the extra compute),” Tremont cofounder Kyle Poyar said in an interview with Mostly Metrics.
Adding to this challenge is a degree of uncertainty about the ROI for AI-driven solutions. As Medved explained, “There are real questions about ROI in the AI space, given that even the trillion-dollar tech leaders are speaking about 15-year recovery of these investments they are now making. However, they would rather take the risk of investing here than miss this unprecedented opportunity.”
For AI companies, the path to sustainable profitability will require navigating the tricky landscape of the high operational costs of the technology and the pressure to demonstrate real value quickly. The road to a sustainable business model may then hinge on factors like efficient cost and pricing control.
As Poyar noted, “With their costs scaling on the basis of usage, AI application vendors may wish to charge their own customers based on usage and thereby protect against their costs ballooning on account of a few heavy users.”
The most pricing power, he said, will be around positioning AI as a means to increase revenue rather than merely a time-saving solution. “This is because time savings isn’t differentiated, doesn’t create urgency, and doesn’t capture real dollars,” he added.
Another factor is prioritizing human capital to maximize the return on substantial investments.
“Spend your money first on hiring the best AI people you can, and let them help you make these investment decisions. Safe Superintelligence reportedly got a seed investment of $1 billion at a $5 billion valuation with just 10 employees — thus implying a value of nearly $1 billion per employee,” Medved said.
Gilles Thonet, deputy-secretary general of the International Electrotechnical Commission, admits he’s unsure about the AI bubble bursting but said AI is pretty much ingrained in our everyday lives and that’s not going to change.
“Risk is in the nature of startups: Some succeed and unfortunately, others fail,” he said. “But following global regulatory frameworks like the ISO/IEC AI standards can eliminate the need for businesses to start from scratch, guiding them to follow the best practices in designing algorithms and ensuring data quality to reduce AI bias.”
Ultimately, the business leaders who combine great talent with a robust vision for AI’s business applications are more likely to find themselves in the successful 1% of AI companies that endure beyond the current hype.